From jamesalexanderlevy at gmail.com Tue Sep 1 01:02:32 2009 From: jamesalexanderlevy at gmail.com (James Alexander Levy) Date: Mon, 31 Aug 2009 18:02:32 -0700 Subject: [Developers] Problems with uri_submit method In-Reply-To: <8212a4320908281347k4ce598dbj77be408e89f3491c@mail.gmail.com> References: <8212a4320908281347k4ce598dbj77be408e89f3491c@mail.gmail.com> Message-ID: <8212a4320908311802v246031d5r8e8e826b4e79f924@mail.gmail.com> Apologies about replying with the full mailing-list content before. I'm still getting a hang of this whole e-mail thing. Here's my reply to Alec Flett's response from earlier today: On Aug 28, 2009, at 1:47 PM, James Alexander Levy wrote: > fb_session.uri_submit('http://freebaselibs.com') > > This URL is a 404 - there is no document at that URL, and the http > status code says so > Oh man, I think I meant to paste http://freebaselibs.com/static/suggest/0.9.4/. In any case, the uri_submit method seems to be working much better now than it was on Thursday and Friday of last week. > > What exactly are you trying to do? The above action tries to retrieve > the data at the specified url, and upload the contents into the blob/ > trans store. Is that what you're trying to do? > > You're right, I don't need to use uri_submit in this case. I'm trying to simply add a weblink for a given topic, and I've already tried both the methods of trying to set the webpage in an MQLwrite call as a 'webpage' or '/common/topic/webpage' property without explicitly creating the webpage topic first, as well as explicitly creating the weblink in advance, but the former approach returns errors and the latter approach thinks I'm doing an MQLread instead of an MQLwrite. Here is an example of the type of query I'm attempting to use: http://bit.ly/15baBL [{ "/common/topic/webpage": { "/common/webpage/uri": "http://www.mytesturi.com", "/common/webpage/description": "Test Page" }, "id": null, "create": "unless_exists" }] I can't find any examples of setting a weblink property in the developer docs, so perhaps one example would be a useful reference. Alec -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090831/60e91fd6/attachment.htm From jg at metaweb.com Tue Sep 1 08:05:33 2009 From: jg at metaweb.com (John Giannandrea) Date: Tue, 1 Sep 2009 01:05:33 -0700 Subject: [Developers] Problems with uri_submit method In-Reply-To: <8212a4320908311802v246031d5r8e8e826b4e79f924@mail.gmail.com> References: <8212a4320908281347k4ce598dbj77be408e89f3491c@mail.gmail.com> <8212a4320908311802v246031d5r8e8e826b4e79f924@mail.gmail.com> Message-ID: James Alexander Levy wrote: > Here is an example of the type of query I'm attempting to use: http://bit.ly/15baBL you want to put the create_unless clause inside the webpage clause since you are making a new webpage object. you need to add a "type" link in there too to make it a /common/ webpage e.g. "type":"/common/webpage", "uri": "http://www.mytesturi.com" "description":"Test Page" and you need to not use "id":null since you want to connect that new webpage object to something specific, like a topic. See http://tinyurl.com/kkt7ye and http://www.sandbox-freebase.com/tools/explore/guid/9202a8c04000641f800000000e5e536d -jg From ganguly.sarthak at gmail.com Tue Sep 1 21:32:26 2009 From: ganguly.sarthak at gmail.com (SArTHAK..... gangopadhyay) Date: Tue, 1 Sep 2009 14:32:26 -0700 (PDT) Subject: [Developers] Invitation to connect on LinkedIn Message-ID: <92156789.10200106.1251840746014.JavaMail.app@ech3-cdn05.prod> LinkedIn ------------ I'd like to add you to my professional network on LinkedIn. - SArTHAK..... Learn more: https://www.linkedin.com/e/isd/712601192/EyqgVQI5/ ------ (c) 2009, LinkedIn Corporation -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090901/f0d16708/attachment-0001.htm From thelevybreaks at gmail.com Wed Sep 2 04:42:47 2009 From: thelevybreaks at gmail.com (James Alexander Levy) Date: Tue, 1 Sep 2009 21:42:47 -0700 Subject: [Developers] Problems with uri_submit method In-Reply-To: <8212a4320908311802v246031d5r8e8e826b4e79f924@mail.gmail.com> References: <8212a4320908281347k4ce598dbj77be408e89f3491c@mail.gmail.com> <8212a4320908311802v246031d5r8e8e826b4e79f924@mail.gmail.com> Message-ID: <8212a4320909012142k527f8f49r2896a346be88474a@mail.gmail.com> > > you need to add a "type" link in there too to make it a /common/ > webpage e.g. > > "type":"/common/webpage", > "uri": "http://www.mytesturi.com" > "description":"Test Page" > > and you need to not use "id":null since you want to connect that new > webpage object to something specific, like a topic. > > See http://tinyurl.com/kkt7ye and > http://www.sandbox-freebase.com/tools/explore/guid/9202a8c04000641f800000000e5e536d > > Great! I got it to work. It seems so easy once someone else tells me how to do it ^^. > -jg > > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090901/4bcf3ab4/attachment.htm From thelevybreaks at gmail.com Wed Sep 2 05:02:20 2009 From: thelevybreaks at gmail.com (James Alexander Levy) Date: Tue, 1 Sep 2009 22:02:20 -0700 Subject: [Developers] speeding up fb-suggest using incoming property mqlfilter Message-ID: <8212a4320909012202t4d8944aeu4ced12855eddfaf9@mail.gmail.com> > > > On Mon, 31 Aug 2009, James Alexander Levy wrote: > > > It seems like if I set the type argument to /common/topic then it will > only > > return topics that are only /common/topic. > > Err, no, it should returns only topics, not things that are only topics. > In other words, these topics could be typed as other things as well. > For example: > > http://www.freebase.com/api/service/search?query=gore&type=/common/topic&indent=1 > You can also condition that type match with the type_strict parameter. > See http://www.freebase.com/api/service/search?help#limiting > > Thanks for the help. To be more precise about the issue I face, here's a comparison of the expected results from the query editor ( http://bit.ly/32j4J1) and the empty results from both the search service ( http://bit.ly/cOkz2) and the suggest service ( http://bit.ly/3ZGFiR) . From looking at the documentation, I think I'm creating these MQL filters correctly, but the searching part is the part that seems to not be executing as expected. Andi.. > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090901/d0584d63/attachment.htm From vajda at metaweb.com Wed Sep 2 05:33:05 2009 From: vajda at metaweb.com (Andi Vajda) Date: Tue, 1 Sep 2009 22:33:05 -0700 (PDT) Subject: [Developers] speeding up fb-suggest using incoming property mqlfilter In-Reply-To: <8212a4320909012202t4d8944aeu4ced12855eddfaf9@mail.gmail.com> References: <8212a4320909012202t4d8944aeu4ced12855eddfaf9@mail.gmail.com> Message-ID: On Tue, 1 Sep 2009, James Alexander Levy wrote: > Thanks for the help. To be more precise about the issue I face, here's a > comparison of the expected results from the query editor > (http://bit.ly/32j4J1) and the empty results from both the search service > (http://bit.ly/cOkz2) and the suggest service (http://bit.ly/3ZGFiR). From > looking at the documentation, I think I'm creating these MQL filters > correctly, but the searching part is the part that seems to not be executing > as expected. I can't help you with suggest, I don't know anything about it. The api/service/search service returns matches when your filter is left out. http://www.freebase.com/api/service/search?&prefix=Twit&indent=1 Same is true using suggest: http://www.freebase.com/private/suggest?&prefix=Twit I guess there are no matches that pass your filter. Andi.. From daepark at metaweb.com Fri Sep 4 00:15:55 2009 From: daepark at metaweb.com (Dae Park) Date: Thu, 3 Sep 2009 17:15:55 -0700 Subject: [Developers] freebase-suggest 0.9.5 Message-ID: <7BF0CB90-5277-4F3C-807C-6F9B131A32E5@metaweb.com> ... is now available: http://www.freebase.com/docs/suggest Some new features and bug fixes include: - Search by type name. The type parameter now accepts names like 'person', 'politician', etc. instead of the fully qualified id ('/ people/person'). Suggest has also supported a 'type:' constraint inside the input as well. So now you can use type names as constraints as well. Try 'george type:politician'. - Better disambiguation for the various types of flyouts. - Better caching of results (suggestions and flyouts). - Handle mouse paste enjoy! - The Freebase Team From daepark at metaweb.com Fri Sep 4 01:18:09 2009 From: daepark at metaweb.com (Dae Park) Date: Thu, 3 Sep 2009 18:18:09 -0700 Subject: [Developers] freebase-suggest 0.9.5 In-Reply-To: <7BF0CB90-5277-4F3C-807C-6F9B131A32E5@metaweb.com> References: <7BF0CB90-5277-4F3C-807C-6F9B131A32E5@metaweb.com> Message-ID: More on search by type name... Our relevance guru also points out: > Andi Vajda wrote: > > ...type also accepts pseudo type names such as 'volcano', a mountain > type (but not technically a type). > Such type names are stemmed, so you can also use type=volcanos, for > example. I believe you can also treat people's professions as a pseudo type as well (i.e., "johnson type:lawyer"). -dae On Sep 3, 2009, at 5:15 PM, Dae Park wrote: > ... is now available: > > http://www.freebase.com/docs/suggest > > Some new features and bug fixes include: > > - Search by type name. The type parameter now accepts names like > 'person', 'politician', etc. instead of the fully qualified id ('/ > people/person'). Suggest has also supported a 'type:' constraint > inside the input as well. So now you can use type names as constraints > as well. Try 'george type:politician'. > > - Better disambiguation for the various types of flyouts. > > - Better caching of results (suggestions and flyouts). > > - Handle mouse paste > > > enjoy! > > - The Freebase Team > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From vajda at metaweb.com Fri Sep 4 02:08:51 2009 From: vajda at metaweb.com (Andi Vajda) Date: Thu, 3 Sep 2009 19:08:51 -0700 (PDT) Subject: [Developers] freebase-suggest 0.9.5 In-Reply-To: References: <7BF0CB90-5277-4F3C-807C-6F9B131A32E5@metaweb.com> Message-ID: On Thu, 3 Sep 2009, Dae Park wrote: > More on search by type name... > > Our relevance guru also points out: > >> Andi Vajda wrote: >> >> ...type also accepts pseudo type names such as 'volcano', a mountain >> type (but not technically a type). >> Such type names are stemmed, so you can also use type=volcanos, for >> example. > > > I believe you can also treat people's professions as a pseudo type as > well (i.e., "johnson type:lawyer"). The values in the properties returned by [1] are indexed as pseudo type names. For example, "/en/volcano" occurs in "/geography/mountain/mountain_type" values and its name, "volcano" is indexed as the a type name of things that have "/en/volcano" in their "/geography/mountain/mountain_type" property. Andi.. [1] http://tinyurl.com/mz2fnb From alecf at metaweb.com Tue Sep 8 18:49:04 2009 From: alecf at metaweb.com (Alec Flett) Date: Tue, 8 Sep 2009 11:49:04 -0700 Subject: [Developers] test freebase-python trunk? Message-ID: <20E64925-1653-4EF8-81F2-D568DEA0595E@metaweb.com> I just made some minor changes to the way freebase-python deals with unicode and I'm wondering if I can get some folks to test out the freebase-python. I'm about to release version 1.0.3 but before I did I wanted to make sure I don't break anyone. http://code.google.com/p/freebase-python/source/checkout or more specifically: % svn checkout http://freebase-python.googlecode.com/svn/trunk/ freebase-python % cd freebase-python % path-to-your-python-executable/python setup.py develop Specifically I found some cases where it was throwing unicode exceptions just trying to make mqlread() requests with certain data. Theoretically, this should be a transparent change that just fixes random exceptions Alec From raphael.jolivet at gmail.com Tue Sep 8 19:05:07 2009 From: raphael.jolivet at gmail.com (Raphael Jolivet) Date: Tue, 8 Sep 2009 21:05:07 +0200 Subject: [Developers] [mjt] Accessing the current DOM element in mjt.script Message-ID: Hi, In a "mjt.script" tag, I would like to reference the current element dynamically, to attach some user-defined properties to it. I have tried something like
  • Page:$pageId 
                  this.pageIdx=pageIdx;
But it does not work. "this" points to an unrelevant template object. It would be very useful to have it. In combination with jQuery, I could access the parents of the current element and attach some events, properties. Thanks in advance for your help. Kind regards, Raphael -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090908/c8b66804/attachment-0001.htm From tfmorris at gmail.com Tue Sep 8 19:38:57 2009 From: tfmorris at gmail.com (Tom Morris) Date: Tue, 8 Sep 2009 15:38:57 -0400 Subject: [Developers] test freebase-python trunk? In-Reply-To: <20E64925-1653-4EF8-81F2-D568DEA0595E@metaweb.com> References: <20E64925-1653-4EF8-81F2-D568DEA0595E@metaweb.com> Message-ID: On Tue, Sep 8, 2009 at 2:49 PM, Alec Flett wrote: > I just made some minor changes to the way freebase-python deals with > unicode and I'm wondering if I can get some folks to test out the > freebase-python. I'm about to release version 1.0.3 but before I did I > wanted to make sure I don't break anyone. I ran a few of my programs and didn't see any obvious problems. It looks like you've switched to using POST. Will this have performance (or other) implications? Tom From thelevybreaks at gmail.com Tue Sep 8 20:36:03 2009 From: thelevybreaks at gmail.com (James Alexander Levy) Date: Tue, 8 Sep 2009 13:36:03 -0700 Subject: [Developers] test freebase-python trunk? (Alec Flett) Message-ID: <8212a4320909081336t1f64f5d4qf5c069c5ee553eba@mail.gmail.com> > > I just made some minor changes to the way freebase-python deals with > unicode and I'm wondering if I can get some folks to test out the > freebase-python. I'm about to release version 1.0.3 but before I did I > wanted to make sure I don't break anyone. > > http://code.google.com/p/freebase-python/source/checkout > > or more specifically: > > % svn checkout http://freebase-python.googlecode.com/svn/trunk/ > freebase-python > % cd freebase-python > % path-to-your-python-executable/python setup.py develop > > > Specifically I found some cases where it was throwing unicode > exceptions just trying to make mqlread() requests with certain data. > Theoretically, this should be a transparent change that just fixes > random exceptions > > Awesome. This update fixed a KeyError issue I had ( http://code.google.com/p/freebase-python/issues/detail?id=6). -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090908/883068d3/attachment.htm From thelevybreaks at gmail.com Tue Sep 8 20:43:15 2009 From: thelevybreaks at gmail.com (James Alexander Levy) Date: Tue, 8 Sep 2009 13:43:15 -0700 Subject: [Developers] Problems with uri_submit method (James Alexander Levy) Message-ID: <8212a4320909081343r596d6b53h7819e649fd879cbb@mail.gmail.com> > > On Tue, 1 Sep 2009, James Alexander Levy wrote: > > > Thanks for the help. To be more precise about the issue I face, here's a > > comparison of the expected results from the query editor > > (http://bit.ly/32j4J1) and the empty results from both the search > service > > (http://bit.ly/cOkz2) and the suggest service (http://bit.ly/3ZGFiR). > From > > looking at the documentation, I think I'm creating these MQL filters > > correctly, but the searching part is the part that seems to not be > executing > > as expected. > > I can't help you with suggest, I don't know anything about it. > The api/service/search service returns matches when your filter is left > out. > > http://www.freebase.com/api/service/search?&prefix=Twit&indent=1 > > Same is true using suggest: > http://www.freebase.com/private/suggest?&prefix=Twit > > I guess there are no matches that pass your filter. > I used that example because Twitter is one of the results that comes up in the query editor (http://bit.ly/32j4J1) but not from the suggest service ( http://bit.ly/3ZGFiR) and afaik my mql_filter and prefix arguments should get back the topic 'Twitter'. The same problem applies to any prefix, meaning there is likely some query rule or behavior related to reciprocal property queries that I don't grok. james -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090908/fa62941d/attachment.htm From nix at metaweb.com Tue Sep 8 21:35:26 2009 From: nix at metaweb.com (Nick Thompson) Date: Tue, 08 Sep 2009 14:35:26 -0700 Subject: [Developers] [mjt] Accessing the current DOM element in mjt.script In-Reply-To: References: Message-ID: <4AA6CE1E.80108@metaweb.com> Raphael Jolivet wrote: > In a "mjt.script" tag, I would like to reference the current element > dynamically, to attach some user-defined properties to it. this is a little trickier than you would expect, because mjt templates generate innerHTML rather than generating dom elements one-by-one (this is necessary for speed). by default, mjt.script code is run as the html is generated, so the dom element doesn't actually exist yet. however, you can use a mjt.script="ondomready" attribute to delay your script until after the template text has been converted to dom elements. you will need to set up unique id= attributes so that you can find them when running your jquery code. so you want something like this (untested):
  • Page:$pageId 
                    $('#pageitem-' + pageIdx).get(0).pageIdx=pageIdx;
hope that helps, nick Raphael Jolivet wrote: > Hi, > > In a "mjt.script" tag, I would like to reference the current element > dynamically, to attach some user-defined properties to it. > > I have tried something like > >
    >
  • > Page:$pageId > 
    >               this.pageIdx=pageIdx;
>
    >
    • >
> > But it does not work. > "this" points to an unrelevant template object. > > It would be very useful to have it. > In combination with jQuery, I could access the parents of the current > element and attach some events, properties. > > Thanks in advance for your help. > > Kind regards, > > Raphael > > > ------------------------------------------------------------------------ > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From tfmorris at gmail.com Tue Sep 8 22:15:00 2009 From: tfmorris at gmail.com (Tom Morris) Date: Tue, 8 Sep 2009 18:15:00 -0400 Subject: [Developers] Funny type names and MQL Message-ID: This query returns an error saying it's "not a valid type name," yet it's the same name that MQL returned me from a previous query. { "type": "/user/robert/world$0027s_tallest/building", "return": "count" } The explore page seems to think the name is OK http://www.freebase.com/tools/explore/user/robert/world$0027s_tallest/building How do I need to transform the type name to get it in the form that MQL wants? Tom From tfmorris at gmail.com Wed Sep 9 01:48:52 2009 From: tfmorris at gmail.com (Tom Morris) Date: Tue, 8 Sep 2009 21:48:52 -0400 Subject: [Developers] Is 'search' a reserved property name? Message-ID: I'm doing more type inspection work and the following query [{ "type": "/user/bio2rdf/public/bm", "*": [{}] }?] returns "type /user/bio2rdf/public/bm already has a property named 'search'" That's a true statement, it doesn't have a property named 'search,' but I don't see what that has to do with my query. Is there a way to work around this? Tom p.s. The actual query that I want to get to work is { "return": "estimate-count", "type": "/user/bio2rdf/public/bm", "/user/bio2rdf/public/bm/is_replaced_by": [ { "optional": false, "id": null } ] } From vajda at metaweb.com Wed Sep 9 02:54:58 2009 From: vajda at metaweb.com (Andi Vajda) Date: Tue, 8 Sep 2009 19:54:58 -0700 (PDT) Subject: [Developers] Is 'search' a reserved property name? In-Reply-To: References: Message-ID: On Tue, 8 Sep 2009, Tom Morris wrote: > I'm doing more type inspection work and the following query > > [{ "type": "/user/bio2rdf/public/bm", > "*": [{}] > }?] > > returns "type /user/bio2rdf/public/bm already has a property named 'search'" > > That's a true statement, it doesn't have a property named 'search,' > but I don't see what that has to do with my query. 'search' is a property on /type/object/search and as such is a reserved word. Any property on /type/object is reserved and a property by the same name cannot be defined on another type. /type/object/search is an extension property for eMQL [1]. The fact that it shows up in a "*" is an annoyance that needs to be resolved. The fact that the 'bio2rdf' query is failing is a more serious issue that we thought was resolved. > Is there a way to work around this? The owner of the bio2rdf domain had renamed his 'search' property to something else. By the time he did, we had removed the /type/object/search property because of this problem. This happened in June. I guess, seeing this, he re-renamed it back to 'search'. In late August, believing the bio2rdf property was renamed, I added the /type/object/search property back, along with /people/person/age, to see if they caused problems. My bad for not verifying that it hadn't reappeared first. Until the owner of the 'bio2rdf' domain renames his property again, we're having this problem without workaround. He had graciously accepted to rename it in June, I hope he does so again. I hope to find out more about this soon. With apologies, Andi.. [1] http://search.labs.freebase.com/api/service/emql?help > > Tom > > p.s. The actual query that I want to get to work is > > { > "return": "estimate-count", > "type": "/user/bio2rdf/public/bm", > "/user/bio2rdf/public/bm/is_replaced_by": [ > { > "optional": false, > "id": null > } > ] > } > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From tfmorris at gmail.com Wed Sep 9 04:25:49 2009 From: tfmorris at gmail.com (Tom Morris) Date: Wed, 9 Sep 2009 00:25:49 -0400 Subject: [Developers] Is 'search' a reserved property name? In-Reply-To: References: Message-ID: On Tue, Sep 8, 2009 at 10:54 PM, Andi Vajda wrote: > Until the owner of the 'bio2rdf' domain renames his property again, we're > having this problem without workaround. He had graciously accepted to rename > it in June, I hope he does so again. Thanks for the quick reply. I'll manually exclude the type for now. Tom From philip-freebase at shadowmagic.org.uk Wed Sep 9 11:04:12 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Wed, 9 Sep 2009 12:04:12 +0100 Subject: [Developers] Funny type names and MQL In-Reply-To: References: Message-ID: <20090909110412.GD13273@sphinx.mythic-beasts.com> On Tue, Sep 08, 2009 at 06:15:00PM -0400, Tom Morris wrote: > This query returns an error saying it's "not a valid type name," yet > it's the same name that MQL returned me from a previous query. > > { > "type": "/user/robert/world$0027s_tallest/building", > "return": "count" > } > > The explore page seems to think the name is OK > http://www.freebase.com/tools/explore/user/robert/world$0027s_tallest/building > > How do I need to transform the type name to get it in the form that MQL wants? { "type": { "id": "/user/robert/world$0027s_tallest/building", }, "return": "count" } works. MQL bug, I'd say. Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From sm at metaweb.com Wed Sep 9 16:02:25 2009 From: sm at metaweb.com (Scott Meyer) Date: Wed, 09 Sep 2009 09:02:25 -0700 Subject: [Developers] Funny type names and MQL In-Reply-To: <20090909110412.GD13273@sphinx.mythic-beasts.com> References: <20090909110412.GD13273@sphinx.mythic-beasts.com> Message-ID: <4AA7D191.6040405@metaweb.com> Philip Kendall wrote: > On Tue, Sep 08, 2009 at 06:15:00PM -0400, Tom Morris wrote: >> This query returns an error saying it's "not a valid type name," yet >> it's the same name that MQL returned me from a previous query. >> >> { >> "type": "/user/robert/world$0027s_tallest/building", >> "return": "count" >> } >> >> The explore page seems to think the name is OK >> http://www.freebase.com/tools/explore/user/robert/world$0027s_tallest/building >> >> How do I need to transform the type name to get it in the form that MQL wants? > > { > "type": { > "id": "/user/robert/world$0027s_tallest/building", > }, > "return": "count" > } > > works. MQL bug, I'd say. Indeed. I've create MQL-571 to track it. -Scott From tfmorris at gmail.com Wed Sep 9 17:03:18 2009 From: tfmorris at gmail.com (Tom Morris) Date: Wed, 9 Sep 2009 13:03:18 -0400 Subject: [Developers] Funny type names and MQL In-Reply-To: <4AA7D191.6040405@metaweb.com> References: <20090909110412.GD13273@sphinx.mythic-beasts.com> <4AA7D191.6040405@metaweb.com> Message-ID: Thanks for the workaround Phil (and to Scott for reporting the bug for me). Tom From iainsproat at gmail.com Wed Sep 9 19:03:12 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Wed, 9 Sep 2009 23:03:12 +0400 Subject: [Developers] Acre performance with multiple acre.freebase.mqlread Message-ID: I have an Acre app that calls acre.freebase.mqlread multiple times from a single script (potentially to the order of a hundred or so calls) If it was an external app using the Freebase API ( http://api.freebase.com/api/service/mqlread ), I'd try to minimise the number of costly calls between servers by grouping the individual queries and doing one call using the '?queries=' parameter. But as it's an Acre app I was wondering whether the overhead per individual mqlread call might be remarkably less; and if I could get away with being lazy and allow the hundreds of calls? My concern for optimising performance in Acre is predominantly timeout failures - has anybody dealt with anything similar before? Iain From jason at metaweb.com Wed Sep 9 19:21:17 2009 From: jason at metaweb.com (Jason Douglas) Date: Wed, 9 Sep 2009 12:21:17 -0700 Subject: [Developers] Acre performance with multiple acre.freebase.mqlread In-Reply-To: References: Message-ID: <74D35090-5B1C-4D8F-95FD-F36F717CCAF9@metaweb.com> On Sep 9, 2009, at 12:03 PM, Iain Sproat wrote: > If it was an external app using the Freebase API ( > http://api.freebase.com/api/service/mqlread ), I'd try to minimise the > number of costly calls between servers by grouping the individual > queries and doing one call using the '?queries=' parameter. There's actually a convenience method for this in Acre (no idea why it's not documented): acre.freebase.mqlread_multiple() I believe the arguments are queries, envelopes and params. I think the first two are arrays of objects and the last is a dictionary. Not 100% positive, though. > But as it's an Acre app I was wondering whether the overhead per > individual mqlread call might be remarkably less; and if I could get > away with being lazy and allow the hundreds of calls? http calls are probably the biggest overhead in Acre, especially since they're not done serially. So I would try mqlread_multiple. -jason From stefano at metaweb.com Wed Sep 9 19:40:16 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Wed, 09 Sep 2009 12:40:16 -0700 Subject: [Developers] Acre performance with multiple acre.freebase.mqlread In-Reply-To: <74D35090-5B1C-4D8F-95FD-F36F717CCAF9@metaweb.com> References: <74D35090-5B1C-4D8F-95FD-F36F717CCAF9@metaweb.com> Message-ID: <4AA804A0.1040306@metaweb.com> Jason Douglas wrote: > On Sep 9, 2009, at 12:03 PM, Iain Sproat wrote: >> If it was an external app using the Freebase API ( >> http://api.freebase.com/api/service/mqlread ), I'd try to minimise the >> number of costly calls between servers by grouping the individual >> queries and doing one call using the '?queries=' parameter. > > There's actually a convenience method for this in Acre (no idea why > it's not documented): > > acre.freebase.mqlread_multiple() > > I believe the arguments are queries, envelopes and params. I think > the first two are arrays of objects and the last is a dictionary. Not > 100% positive, though. The method signature is the following: acre.freebase.mqlread_multiple(queries,envelopes,params) where both queries and envelopes are two js objects and the names should align, like this: queries = { 'first_query' : ... }; envelopes = { 'first_query' : ... }; acre.freebase.mqlread_multiple(queries,envelopes) Note that envelopes is optional, so you can just do queries = {} queries.a = "..." queries.b = "..." acre.freebase.mqlread_multiple(queries) the reason why we used an object and not an array is that it helps with the results to be able to retrieve the results of the right query We kept this undocumented because we were not sure this was the way to do it, so if you have feedback that'd be great. HTH -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From iainsproat at gmail.com Wed Sep 9 19:51:02 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Wed, 9 Sep 2009 23:51:02 +0400 Subject: [Developers] Acre performance with multiple acre.freebase.mqlread In-Reply-To: <74D35090-5B1C-4D8F-95FD-F36F717CCAF9@metaweb.com> References: <74D35090-5B1C-4D8F-95FD-F36F717CCAF9@metaweb.com> Message-ID: On Wed, Sep 9, 2009 at 11:21 PM, Jason Douglas wrote: > There's actually a convenience method for this in Acre (no idea why > it's not documented): In the acre recipes documentation, http://www.freebase.com/docs/acre_recipes, I saw a recipe/workaround for calling the Freebase API using acre.urlfetch, which made me think there wasn't anything. > ? ? ? ?acre.freebase.mqlread_multiple() Thanks Jason, that really helps. Out of interest, would you happen to know how the performance benchmark for calling 1000 mqlread's compares to passing a single envelope of 1000 queries to mqlread_multiple? Iain From iainsproat at gmail.com Wed Sep 9 20:54:11 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Thu, 10 Sep 2009 00:54:11 +0400 Subject: [Developers] Acre performance with multiple acre.freebase.mqlread In-Reply-To: <4AA804A0.1040306@metaweb.com> References: <74D35090-5B1C-4D8F-95FD-F36F717CCAF9@metaweb.com> <4AA804A0.1040306@metaweb.com> Message-ID: On Wed, Sep 9, 2009 at 11:40 PM, Stefano Mazzocchi wrote: > Jason Douglas wrote: >> On Sep 9, 2009, at 12:03 PM, Iain Sproat wrote: > ? acre.freebase.mqlread_multiple(queries,envelopes,params) > > Note that envelopes is optional, so you can just do > > ?queries = {} > ?queries.a = "..." > ?queries.b = "..." > ?acre.freebase.mqlread_multiple(queries) > > the reason why we used an object and not an array is that it helps with > the results to be able to retrieve the results of the right query > > We kept this undocumented because we were not sure this was the way to > do it, so if you have feedback that'd be great. I'm getting an error message "Type /type/object does not have property query", for a query which was working when used with the Freebase API mqlread?queries= Could I confirm that the query format is the same as the API, e.g. {"q1":{ "query":[{ "id":null, "type":"/people/person" }] }, "q2":{ "query":[{ "id":null, "type":"/location/location" }] }} >From the error message, I'd assume that the additional wrapper for the "query" is not require for mqlread_multiple? I'd prefer consistency with the Freebase API if possible. Iain From amakhan at microsoft.com Wed Sep 9 22:07:29 2009 From: amakhan at microsoft.com (Andrei Makhanov) Date: Wed, 9 Sep 2009 22:07:29 +0000 Subject: [Developers] Getting a list of films an actor appeared in Message-ID: <09490EFB86213A49A21B7041B32194800CD4F3@TK5EX14MBXC120.redmond.corp.microsoft.com> Hello, I'm having some trouble getting a list of films an actor has appeared in. Issuing the MQL [{ type: "/film/actor", guid:"#9202a8c04000641f8000000000137850", film : [] }] gives me: result: [ expand all ? collapse all ? 1-by-1[ 1 element ] [http://freebaselibs.com/static/cuecard/1.2.4/api/images/expanded.png][http://freebaselibs.com/static/cuecard/1.2.4/api/images/collapsed.png]0: {{...} [http://freebaselibs.com/static/cuecard/1.2.4/api/images/expanded.png][http://freebaselibs.com/static/cuecard/1.2.4/api/images/collapsed.png]film: [ expand all ? collapse all ? 1-by-1[ 57 elements ] 0: null 1: null ... ... 56: null guid: "#9202a8c04000641f8000000000137850"xp type: "/film/actor"xp How can I get the films to now show up as 'null'? thanks aM -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090909/ac5036b5/attachment-0001.htm From filipealvesferreira at gmail.com Wed Sep 9 22:18:59 2009 From: filipealvesferreira at gmail.com (Filipe AlvesFerreira) Date: Thu, 10 Sep 2009 00:18:59 +0200 Subject: [Developers] Development in Economy 4G3W Message-ID: I would like to get collaboration to launch the personal webcashmotor for the money datevaluation practice : https://sites.google.com/a/datevalors1.com/mywebjurispersonquality/home Regards/FilipeAlvesFerreira#4(1942) -- FilipeAlvesFerreira#4(1942) -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090910/be5e1e6f/attachment.htm From philip-freebase at shadowmagic.org.uk Wed Sep 9 22:25:04 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Wed, 9 Sep 2009 23:25:04 +0100 Subject: [Developers] Getting a list of films an actor appeared in In-Reply-To: <09490EFB86213A49A21B7041B32194800CD4F3@TK5EX14MBXC120.redmond.corp.microsoft.com> References: <09490EFB86213A49A21B7041B32194800CD4F3@TK5EX14MBXC120.redmond.corp.microsoft.com> Message-ID: <20090909222504.GE13273@sphinx.mythic-beasts.com> On Wed, Sep 09, 2009 at 10:07:29PM +0000, Andrei Makhanov wrote: > > I'm having some trouble getting a list of films an actor has appeared in. Issuing the MQL > [{ > type: "/film/actor", > guid:"#9202a8c04000641f8000000000137850", > film : [] > }] What you're getting back there are not films, but film performances - a CVT (compound value type) which links together an actor and a film, and which doesn't have a name of its own. Try { "type": "/film/actor", "id": "/en/robert_downey_jr", "film": [{ "film": null }] } Hope this helps. (As an aside, you probably want to be using IDs rather than GUIDs - /en/robert_downey_jr will represent Robert Downey Jr. until the end of time, whereas #9202a8c04000641f8000000000137850 is a reference to a particular node in the graph, which happens to correspond to Robert Downey Jr. at the moment. These won't change *too* often, but can if topics are merged or the like) As another aside, do we have a list of FAQs anywhere, as I think this one is probably about top of it? Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From philip-freebase at shadowmagic.org.uk Thu Sep 10 10:41:45 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Thu, 10 Sep 2009 11:41:45 +0100 Subject: [Developers] Client / Freebase Request feature request Message-ID: <20090910104144.GG13273@sphinx.mythic-beasts.com> One small client and/or Freebase Suggest feature request: * Go to a video game * Edit the genre property * Type "shoot em up" into the empty box At this point, we're looking for a Video Game Genre (/cvg/cvg_genre) topic. What we get in this case is three "Shoot 'em up" topics listed as being of types "Award discipline", "Film" and "Musical track". However, the first item is actually already a Video Game Genre - I think it would be nice if this could be indicated as it would reduce the chances of users using the wrong topic. Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From daepark at metaweb.com Thu Sep 10 16:21:55 2009 From: daepark at metaweb.com (Dae Park) Date: Thu, 10 Sep 2009 09:21:55 -0700 Subject: [Developers] Client / Freebase Request feature request In-Reply-To: <20090910104144.GG13273@sphinx.mythic-beasts.com> References: <20090910104144.GG13273@sphinx.mythic-beasts.com> Message-ID: <1C91E0B4-CC88-4423-8107-10E980BD2698@metaweb.com> This is an excellent idea! I will we see what we can do to get this feature out asap. Thanks for the input! -dae On Sep 10, 2009, at 3:41 AM, Philip Kendall wrote: > One small client and/or Freebase Suggest feature request: > > * Go to a video game > * Edit the genre property > * Type "shoot em up" into the empty box > > At this point, we're looking for a Video Game Genre (/cvg/cvg_genre) > topic. What we get in this case is three "Shoot 'em up" topics > listed as being of types "Award discipline", "Film" and "Musical > track". > However, the first item is actually already a Video Game Genre - I > think > it would be nice if this could be indicated as it would reduce the > chances of users using the wrong topic. > > Cheers, > > Phil > > -- > Philip Kendall > http://www.shadowmagic.org.uk/ > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From jeff at metaweb.com Thu Sep 10 20:32:49 2009 From: jeff at metaweb.com (Jeff Prucher) Date: Thu, 10 Sep 2009 13:32:49 -0700 Subject: [Developers] Refactoring physical geography a bit Message-ID: Based on this discussion (), I'd like refactor some types in the Physical Geography domain. 1. The type /geography/geographical_feature is currently a bucket holding different kinds of features (tarn, isthmus, cave, etc.). I'd like to change the semantics of this type so that it instead holds actual features (Lake Geneva, Isthmus of Panama, Mammoth Cave, etc.). A new property will be added to this type ("Class" or something similar) which will connect to a new type "Geographical Feature Type". This new type will have as instances the same sorts of topics that Geographical Feature currently does. (In fact, all will simply be migrated to the new type.) 2. These types, which have no properties, will be deleted, since the is-a relationship with the kind of feature it is will be handled by a property instead: /geography/desert /geography/bay /geography/strait /geography/wetland The JIRA task is here: https://bugs.freebase.com/browse/DA-916 Will this affect anyone's code? Jeff From philip-freebase at shadowmagic.org.uk Fri Sep 11 12:42:29 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Fri, 11 Sep 2009 13:42:29 +0100 Subject: [Developers] Schema Visualisation app Message-ID: <20090911124229.GK13273@sphinx.mythic-beasts.com> Some of you may have already seen Kirrily's tweet, but if not... a little app while people may be interested in: http://schemaviz.freebaseapps.com/ It should give you a visualisation of any domain you're interested in, showing how the types in that domain are linked by properties. * Nodes represent types, edges represent properties. * Reverse properties are shown in parentheses after the master property name. * Grey edges represent hidden properties * Blue nodes are types outside the domain being viewed, but which are linked to from the domain (the edge will be dashed if the master property is from the type outside the domain to that inside the domain) * Red nodes represent undocumented types * All nodes are clickable to view that type Please let me know of any bugs / feature requests. Thanks should obviously go to everyone at Metaweb for this, in particular the Acre developers and Alexander (for his Genealogy app which was the inspiration for this). Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From stefano at metaweb.com Fri Sep 11 16:29:17 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Fri, 11 Sep 2009 09:29:17 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090911124229.GK13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> Message-ID: <4AAA7ADD.3000102@metaweb.com> Philip Kendall wrote: > Some of you may have already seen Kirrily's tweet, but if not... a little > app while people may be interested in: > > http://schemaviz.freebaseapps.com/ > > It should give you a visualisation of any domain you're interested in, > showing how the types in that domain are linked by properties. > > * Nodes represent types, edges represent properties. > * Reverse properties are shown in parentheses after the master property > name. > * Grey edges represent hidden properties > * Blue nodes are types outside the domain being viewed, but which are > linked to from the domain (the edge will be dashed if the master > property is from the type outside the domain to that inside the > domain) > * Red nodes represent undocumented types > * All nodes are clickable to view that type > > Please let me know of any bugs / feature requests. Thanks should > obviously go to everyone at Metaweb for this, in particular the Acre > developers and Alexander (for his Genealogy app which was the > inspiration for this). Phil, goes without saying, outstanding job. -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From jason at metaweb.com Fri Sep 11 17:20:17 2009 From: jason at metaweb.com (Jason Douglas) Date: Fri, 11 Sep 2009 10:20:17 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <4AAA7ADD.3000102@metaweb.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <4AAA7ADD.3000102@metaweb.com> Message-ID: On Sep 11, 2009, at 9:29 AM, Stefano Mazzocchi wrote: > Philip Kendall wrote: >> Some of you may have already seen Kirrily's tweet, but if not... a >> little >> app while people may be interested in: >> >> http://schemaviz.freebaseapps.com/ >> >> It should give you a visualisation of any domain you're interested >> in, >> showing how the types in that domain are linked by properties. >> >> * Nodes represent types, edges represent properties. >> * Reverse properties are shown in parentheses after the master >> property >> name. >> * Grey edges represent hidden properties >> * Blue nodes are types outside the domain being viewed, but which are >> linked to from the domain (the edge will be dashed if the master >> property is from the type outside the domain to that inside the >> domain) >> * Red nodes represent undocumented types >> * All nodes are clickable to view that type >> >> Please let me know of any bugs / feature requests. Thanks should >> obviously go to everyone at Metaweb for this, in particular the Acre >> developers and Alexander (for his Genealogy app which was the >> inspiration for this). > > Phil, goes without saying, outstanding job. Indeed... it's that rare breed that is fun demo-ware and yet, actually useful. It's found a spot in my bookmarks toolbar already. :-) -jason From spatial.db at gmail.com Fri Sep 11 17:56:35 2009 From: spatial.db at gmail.com (Ed Laurent) Date: Fri, 11 Sep 2009 13:56:35 -0400 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <4AAA7ADD.3000102@metaweb.com> Message-ID: I love it! A suggestion is to add a check box that will add included types to the view. -Ed On Fri, Sep 11, 2009 at 1:20 PM, Jason Douglas wrote: > On Sep 11, 2009, at 9:29 AM, Stefano Mazzocchi wrote: > > > Philip Kendall wrote: > >> Some of you may have already seen Kirrily's tweet, but if not... a > >> little > >> app while people may be interested in: > >> > >> http://schemaviz.freebaseapps.com/ > >> > >> It should give you a visualisation of any domain you're interested > >> in, > >> showing how the types in that domain are linked by properties. > >> > >> * Nodes represent types, edges represent properties. > >> * Reverse properties are shown in parentheses after the master > >> property > >> name. > >> * Grey edges represent hidden properties > >> * Blue nodes are types outside the domain being viewed, but which are > >> linked to from the domain (the edge will be dashed if the master > >> property is from the type outside the domain to that inside the > >> domain) > >> * Red nodes represent undocumented types > >> * All nodes are clickable to view that type > >> > >> Please let me know of any bugs / feature requests. Thanks should > >> obviously go to everyone at Metaweb for this, in particular the Acre > >> developers and Alexander (for his Genealogy app which was the > >> inspiration for this). > > > > Phil, goes without saying, outstanding job. > > Indeed... it's that rare breed that is fun demo-ware and yet, actually > useful. It's found a spot in my bookmarks toolbar already. :-) > > -jason > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090911/70198932/attachment-0001.htm From spencerkelly86 at gmail.com Fri Sep 11 23:19:05 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Fri, 11 Sep 2009 19:19:05 -0400 Subject: [Developers] Refactoring physical geography a bit In-Reply-To: References: Message-ID: +1 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090911/6982dee8/attachment.htm From kurt at spaceship.com Sun Sep 13 01:24:55 2009 From: kurt at spaceship.com (Kurt Bollacker) Date: Sat, 12 Sep 2009 18:24:55 -0700 Subject: [Developers] Freebase accounts and Webapp writes Message-ID: <20090913012455.GB20876@spaceship.com> I'm working on a (non-Acre) Freebase application that involves most of its users adding a small amount of topic data to freebase. This app will integrate with a few other existing web sites that each already have their own login systems (e.g. OpenID and the Django builtin account support). When I get a new user, I want to minimize the pain for that user to authenticate in order to write to Freebase. Also, like Acre apps, I need my middle tier (i.e. not the user's Web browser) to do the writing. If the user has no account on any of the systems involved, then I need to create an account (in Freebase) for them. If they already have an account on one of the other sites, I want to leverage that account and not make them create a new one or be forced to log in twice to access the site and write to Freebase. So how do I do this in on the Freebase side? Some proposals that pop into my head: 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios like this for which tools like this were built, so usage is (hopefully) straightforward. 2. (GOOD) Freebase provides anonymous writes and I use the FB attribution mechanism to link each write to one of the users in my app. 3. (BAD) Freebase provides an API for account creation. I create an account in Freebase for each of my users and write on their behalf. It's still their account, but I set it up for them. The big disadvantage is that the user does not have as much control and awareness of their Freebase account. 4. (UGLY) I create a pool of Freebase accounts for my app to use. When a user needs to write, I grab an account from the pool and use the FB attribution mechanism to mark each write to correspond to the user in my app. When I hit the daily write limit for that account, I switch to another account in the pool. This approach is terrible in so many ways I can't count-- yet it's the only one I can implement today. I'm hoping to start testing my app with real users in 2-3 months. I'm really hoping someone has a better and/or more workable proposal than any of the ones I mentioned here. I'm all ears. Kurt :-) From iainsproat at gmail.com Sun Sep 13 06:37:29 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Sun, 13 Sep 2009 10:37:29 +0400 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <20090913012455.GB20876@spaceship.com> References: <20090913012455.GB20876@spaceship.com> Message-ID: There's documentation on Freebase's oAuth support at http://www.freebase.com/docs/acre/oauth Iain On Sun, Sep 13, 2009 at 5:24 AM, Kurt Bollacker wrote: > > > I'm working on a (non-Acre) Freebase application that involves most of > its users adding a small amount of topic data to freebase. ?This app > will integrate with a few other existing web sites that each already > have their own login systems (e.g. OpenID and the Django builtin > account support). ?When I get a new user, I want to minimize the pain > for that user to authenticate in order to write to Freebase. ?Also, > like Acre apps, I need my middle tier (i.e. not the user's Web > browser) to do the writing. If the user has no account on any of the > systems involved, then I need to create an account (in Freebase) for > them. ?If they already have an account on one of the other sites, I > want to leverage that account and not make them create a new one or be > forced to log in twice to access the site and write to Freebase. > > > So how do I do this in on the Freebase side? ?Some proposals that pop > into my head: > > 1. (BEST) Freebase adds support for OpenID and OAuth. ?It's scenarios > ? like this for which tools like this were built, so usage is > ? (hopefully) straightforward. > > 2. (GOOD) Freebase provides anonymous writes and I use the FB > ? attribution mechanism to link each write to one of the users in > ? my app. > > 3. (BAD) Freebase provides an API for account creation. I create an > ? account in Freebase for each of my users and write on their behalf. > ? It's still their account, but I set it up for them. ?The big > ? disadvantage is that the user does not have as much control and > ? awareness of their Freebase account. > > 4. (UGLY) I create a pool of Freebase accounts for my app to use. When > ? a user needs to write, I grab an account from the pool and use the > ? FB attribution mechanism to mark each write to correspond to the > ? user in my app. ?When I hit the daily write limit for that account, > ? I switch to another account in the pool. ?This approach is terrible > ? in so many ways I can't count-- yet it's the only one I can > ? implement today. > > > I'm hoping to start testing my app with real users in 2-3 months. ?I'm > really hoping someone has a better and/or more workable proposal than > any of the ones I mentioned here. ?I'm all ears. > > > ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Kurt :-) > > > > > > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > From iainsproat at gmail.com Sun Sep 13 07:14:04 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Sun, 13 Sep 2009 11:14:04 +0400 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: References: <20090913012455.GB20876@spaceship.com> Message-ID: actually, on reading, that's just for Acre. On Sun, Sep 13, 2009 at 10:37 AM, Iain Sproat wrote: > There's documentation on Freebase's oAuth support at > http://www.freebase.com/docs/acre/oauth > > Iain > > On Sun, Sep 13, 2009 at 5:24 AM, Kurt Bollacker wrote: >> >> >> I'm working on a (non-Acre) Freebase application that involves most of >> its users adding a small amount of topic data to freebase. ?This app >> will integrate with a few other existing web sites that each already >> have their own login systems (e.g. OpenID and the Django builtin >> account support). ?When I get a new user, I want to minimize the pain >> for that user to authenticate in order to write to Freebase. ?Also, >> like Acre apps, I need my middle tier (i.e. not the user's Web >> browser) to do the writing. If the user has no account on any of the >> systems involved, then I need to create an account (in Freebase) for >> them. ?If they already have an account on one of the other sites, I >> want to leverage that account and not make them create a new one or be >> forced to log in twice to access the site and write to Freebase. >> >> >> So how do I do this in on the Freebase side? ?Some proposals that pop >> into my head: >> >> 1. (BEST) Freebase adds support for OpenID and OAuth. ?It's scenarios >> ? like this for which tools like this were built, so usage is >> ? (hopefully) straightforward. >> >> 2. (GOOD) Freebase provides anonymous writes and I use the FB >> ? attribution mechanism to link each write to one of the users in >> ? my app. >> >> 3. (BAD) Freebase provides an API for account creation. I create an >> ? account in Freebase for each of my users and write on their behalf. >> ? It's still their account, but I set it up for them. ?The big >> ? disadvantage is that the user does not have as much control and >> ? awareness of their Freebase account. >> >> 4. (UGLY) I create a pool of Freebase accounts for my app to use. When >> ? a user needs to write, I grab an account from the pool and use the >> ? FB attribution mechanism to mark each write to correspond to the >> ? user in my app. ?When I hit the daily write limit for that account, >> ? I switch to another account in the pool. ?This approach is terrible >> ? in so many ways I can't count-- yet it's the only one I can >> ? implement today. >> >> >> I'm hoping to start testing my app with real users in 2-3 months. ?I'm >> really hoping someone has a better and/or more workable proposal than >> any of the ones I mentioned here. ?I'm all ears. >> >> >> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Kurt :-) >> >> >> >> >> >> >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers >> > From narphorium at gmail.com Sun Sep 13 08:23:41 2009 From: narphorium at gmail.com (Shawn Simister) Date: Sun, 13 Sep 2009 04:23:41 -0400 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <20090913012455.GB20876@spaceship.com> References: <20090913012455.GB20876@spaceship.com> Message-ID: <4AACAC0D.3040507@gmail.com> I've thought about this problem as well and it seems to me that option 4 becomes a lot less ugly if you only use a single account and ask Metaweb to increase the write quota on that account. That way, all the writes that are coming from your site are properly attributed to your website's Freebase account and you (the owner of the site) implicitly take responsibility for the quality of those edits on behalf of your users. You may even choose to pass the edits through your own moderation pipeline before feeding them back into Freebase. I just think that the other options lead to a lot of confusion because your users may not know or care about Freebase so they'll have no incentive to make sure that their edits are useful to the Freebase community at large. Their only concern is (and should be) the sub-community offered by your website. For example, Freebase treats "President & CFO" as two separate JobTitles whereas the majority of your users may agree that it can be a single person's job title. Even if those users have Freebase accounts created for them, it would be difficult for Freebase data modeling experts to convince them to change their behavior on your site so that it doesn't pollute Freebase with duplicate topics for everyone else. In the end, I think that you (the website owner) are probably the only one who has a vested interest in contributing high quality data back to Freebase so you should be the one attributed and responsible (within reason) for those writes. Just my two cents. Shawn Kurt Bollacker wrote: > I'm working on a (non-Acre) Freebase application that involves most of > its users adding a small amount of topic data to freebase. This app > will integrate with a few other existing web sites that each already > have their own login systems (e.g. OpenID and the Django builtin > account support). When I get a new user, I want to minimize the pain > for that user to authenticate in order to write to Freebase. Also, > like Acre apps, I need my middle tier (i.e. not the user's Web > browser) to do the writing. If the user has no account on any of the > systems involved, then I need to create an account (in Freebase) for > them. If they already have an account on one of the other sites, I > want to leverage that account and not make them create a new one or be > forced to log in twice to access the site and write to Freebase. > > > So how do I do this in on the Freebase side? Some proposals that pop > into my head: > > 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios > like this for which tools like this were built, so usage is > (hopefully) straightforward. > > 2. (GOOD) Freebase provides anonymous writes and I use the FB > attribution mechanism to link each write to one of the users in > my app. > > 3. (BAD) Freebase provides an API for account creation. I create an > account in Freebase for each of my users and write on their behalf. > It's still their account, but I set it up for them. The big > disadvantage is that the user does not have as much control and > awareness of their Freebase account. > > 4. (UGLY) I create a pool of Freebase accounts for my app to use. When > a user needs to write, I grab an account from the pool and use the > FB attribution mechanism to mark each write to correspond to the > user in my app. When I hit the daily write limit for that account, > I switch to another account in the pool. This approach is terrible > in so many ways I can't count-- yet it's the only one I can > implement today. > > > I'm hoping to start testing my app with real users in 2-3 months. I'm > really hoping someone has a better and/or more workable proposal than > any of the ones I mentioned here. I'm all ears. > > > Kurt :-) > > > > > > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > From kurt at spaceship.com Sun Sep 13 16:46:33 2009 From: kurt at spaceship.com (Kurt Bollacker) Date: Sun, 13 Sep 2009 09:46:33 -0700 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <4AACAC0D.3040507@gmail.com> References: <20090913012455.GB20876@spaceship.com> <4AACAC0D.3040507@gmail.com> Message-ID: <20090913164633.GA20986@spaceship.com> On Sun, Sep 13, 2009 at 04:23:41AM -0400, Shawn Simister wrote: > I've thought about this problem as well and it seems to me that option 4 > becomes a lot less ugly if you only use a single account and ask Metaweb > to increase the write quota on that account. That way, all the writes > that are coming from your site are properly attributed to your website's > Freebase account and you (the owner of the site) implicitly take > responsibility for the quality of those edits on behalf of your users. My main motivation to use real users rather than the "indirect users" of attribution was to keep Freebase's user model clean. If one of my users wanted to "reclaim" the the account I created for them or use an existing Freebase account, I could do that too. That said, I really like your idea in that it would be easy for both Metaweb and me to implement. It would also not make things too dirty for Freebase down the road. Perhaps this is a good model of webapp driven write attribution. Ironically though, the original model of attribution was exactly the mirror image -- The human was referenced by the user GUID and the app by the attribution. > You may even choose to pass the edits through your own moderation > pipeline before feeding them back into Freebase. Indeed, the site as currently designed does do some automated vetting/cleaning before writes are done. I had even planned my own anonymous writes that would be held back till they are checked. > I just think that the other options lead to a lot of confusion because > your users may not know or care about Freebase so they'll have no > incentive to make sure that their edits are useful to the Freebase > community at large. Their only concern is (and should be) the > sub-community offered by your website. > > For example, Freebase treats "President & CFO" as two separate JobTitles > whereas the majority of your users may agree that it can be a single > person's job title. Even if those users have Freebase accounts created > for them, it would be difficult for Freebase data modeling experts to > convince them to change their behavior on your site so that it doesn't > pollute Freebase with duplicate topics for everyone else. > > In the end, I think that you (the website owner) are probably the only > one who has a vested interest in contributing high quality data back to > Freebase so you should be the one attributed and responsible (within > reason) for those writes. Yup. My app will be worthless if the data written using it is of poor quality. Kurt :-) From invite+kjdmupw1ddjm at facebookmail.com Sun Sep 13 21:12:48 2009 From: invite+kjdmupw1ddjm at facebookmail.com (David Gonzales) Date: Sun, 13 Sep 2009 14:12:48 -0700 Subject: [Developers] =?utf-8?q?Te_agregu=C3=A9_como_amigo_en_Facebook?= Message-ID: <67feae46417956a42053ddc989e8bece@localhost.localdomain> Hola, developers at freebase.com: He creado un perfil en Facebook donde puedo publicar mis fotos, v?deos y eventos, y quiero agregarte a mis amigos para que puedas verlo. Para ello, necesitas registrarte en Facebook y, despu?s, podr?s crear tambi?n tu propio perfil. Gracias, David Para registrarte en Facebook, sigue este enlace: http://www.facebook.com/p.php?i=741190360&k=32D4P5T4UT6G6BD1SFV6UTW2P&r developers at freebase.com ha sido invitado(a) a unirse a Facebook por David Gonzales. Si no quieres recibir este tipo de mensajes de Facebook, haz clic en el enlace siguiente para borrar tu nombre de la lista de personas suscritas. http://www.facebook.com/o.php?k=e2eac2&u=100000249340683&mid=1175d7eG5af31f56e30bG0G8 Las oficinas de Facebook est?n en 1601 S. California Ave., Palo Alto, CA 94304. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090913/5974fb8c/attachment-0001.htm From stefano at metaweb.com Sun Sep 13 23:46:45 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Sun, 13 Sep 2009 16:46:45 -0700 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <20090913012455.GB20876@spaceship.com> References: <20090913012455.GB20876@spaceship.com> Message-ID: <4AAD8465.7050805@metaweb.com> NOTE: these are my personal opinions, not necessarily what Metaweb thinks or will do. Kurt Bollacker wrote: > > I'm working on a (non-Acre) Freebase application that involves most of > its users adding a small amount of topic data to freebase. This app > will integrate with a few other existing web sites that each already > have their own login systems (e.g. OpenID and the Django builtin > account support). When I get a new user, I want to minimize the pain > for that user to authenticate in order to write to Freebase. Also, > like Acre apps, I need my middle tier (i.e. not the user's Web > browser) to do the writing. If the user has no account on any of the > systems involved, then I need to create an account (in Freebase) for > them. If they already have an account on one of the other sites, I > want to leverage that account and not make them create a new one or be > forced to log in twice to access the site and write to Freebase. > > > So how do I do this in on the Freebase side? Some proposals that pop > into my head: > > 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios > like this for which tools like this were built, so usage is > (hopefully) straightforward. OAuth (aka allowing 3rd-party-apps to act on behalf of a freebase user) support is there already and not just for Acre. OpenID (aka allowing users to log into freebase using credentials from another one of their accounts) is *not*, but we are interested in providing the service, at least with a selected number of identity providers. This is because, like you, we would like to minimize the amount of effort a user has to go thru to be able to contribute to freebase. > 2. (GOOD) Freebase provides anonymous writes and I use the FB > attribution mechanism to link each write to one of the users in > my app. Anonymous writes would be really bad: if Metaweb ever received a DMCA take-down notice for copyright infringement today it could easily and automatically comply by reverting all contributions made by a certain user (or at least narrow it down before inspecting it manually), with anonymous users you are effectively pooling all sorts of people into a single bucket and one bad apple would spoil the whole bunch. We could, ideally, just log their IP address and use that as a provenance disambiguator (that's how wikipedia does it) but personally, I'd much rather invest more in #1 above than cut corners on such an important part of the contribution dynamics equation. > 3. (BAD) Freebase provides an API for account creation. I create an > account in Freebase for each of my users and write on their behalf. > It's still their account, but I set it up for them. The big > disadvantage is that the user does not have as much control and > awareness of their Freebase account. Oh boy, this is even worse than #2: spammers could be able to route around #2 by creating a myriad of fake accounts. Sure, we could wrap the service around captcha and the like, but again, the effort required to implement #1 is probably the same as implementing a good #3 and a lot easier to justify. > 4. (UGLY) I create a pool of Freebase accounts for my app to use. When > a user needs to write, I grab an account from the pool and use the > FB attribution mechanism to mark each write to correspond to the > user in my app. When I hit the daily write limit for that account, > I switch to another account in the pool. This approach is terrible > in so many ways I can't count-- yet it's the only one I can > implement today. Agreed is horrible and again, your pooling contributions together so that if one of your users infringes somebody's copyright and they ask us to take it down, all your app's contribution in that pool might go down with it. > I'm hoping to start testing my app with real users in 2-3 months. I'm > really hoping someone has a better and/or more workable proposal than > any of the ones I mentioned here. I'm all ears. AFAIK, there is an hybrid OpenID + OAuth protocol draft in the wild http://www.techcrunch.com/2009/01/29/openid-oauth-two-great-tastes-that-taste-great-together/ http://googledataapis.blogspot.com/2009/01/bringing-openid-and-oauth-together.html but it doesn't seem like it's getting that much traction in the real world (nor in the various expert groups, really) http://wiki.openid.net/OpenID-and-OAuth-Hybrid-Extension although its usability claims are pretty impressive http://www.readwriteweb.com/archives/comcast_property_sees_92_success_rate_openid.php In short: it would be *awesome* if a user could contribute to freebase with an identity that was hard to mechanically falsify (doesn't have to be legally binding, just unique enough) and one that the user already had (and hopefully was already signed into) instead of having to create a new one specifically for Freebase. How to achieve that, well, that's the big question.... one thing is for sure: if we supported just Google, Yahoo and Facebook accounts, that would cover pretty much 99% of our potential contributors population so we might just do that instead of waiting for a shiny protocol to get finalized. -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From kurt at spaceship.com Mon Sep 14 01:22:34 2009 From: kurt at spaceship.com (Kurt Bollacker) Date: Sun, 13 Sep 2009 18:22:34 -0700 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <4AAD8465.7050805@metaweb.com> References: <20090913012455.GB20876@spaceship.com> <4AAD8465.7050805@metaweb.com> Message-ID: <20090914012234.GC20986@spaceship.com> On Sun, Sep 13, 2009 at 04:46:45PM -0700, Stefano Mazzocchi wrote: > NOTE: these are my personal opinions, not necessarily what Metaweb > thinks or will do. > > > > 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios > > like this for which tools like this were built, so usage is > > (hopefully) straightforward. > > OAuth (aka allowing 3rd-party-apps to act on behalf of a freebase user) > support is there already and not just for Acre. That's great! I can find some OAuth docs for acre at: http://www.freebase.com/docs/acre/oauth that make reference to OAuth URLs for freebase.com, but where are the docs of using OAuth to access Freebase? My case is probably more difficult than the standard use case for OAuth because I need my middle tier to do the Freebase writes (just like Acre does for acre apps). I'll have to proxy all of the calls for the "OAuth dance" through my middle tier in order to grab the cookie. I wonder if there is a better way? > OpenID (aka allowing users to log into freebase using credentials from > another one of their accounts) is *not*, but we are interested in > providing the service, at least with a selected number of identity > providers. This is because, like you, we would like to minimize the > amount of effort a user has to go thru to be able to contribute to freebase. That's good. Are there any official Metaweb plans for such? > > 3. (BAD) Freebase provides an API for account creation. I create an > > account in Freebase for each of my users and write on their behalf. > > It's still their account, but I set it up for them. The big > > disadvantage is that the user does not have as much control and > > awareness of their Freebase account. > > Oh boy, this is even worse than #2: spammers could be able to route > around #2 by creating a myriad of fake accounts. Sure, we could wrap the > service around captcha and the like, but again, the effort required to > implement #1 is probably the same as implementing a good #3 and a lot > easier to justify. The API for account creation could be made so that only a few trusted users could use it. A trusted "account partner" could log in and create an account, which would then be marked as having been created by the partner. My guess is that the number of partners would be small enough for some time, making them manageable for some time. That said, this is not my first choice for a solution. I don't want to be in the account creation business. > > 4. (UGLY) I create a pool of Freebase accounts for my app to use. When > > a user needs to write, I grab an account from the pool and use the > > FB attribution mechanism to mark each write to correspond to the > > user in my app. When I hit the daily write limit for that account, > > I switch to another account in the pool. This approach is terrible > > in so many ways I can't count-- yet it's the only one I can > > implement today. > > Agreed is horrible and again, your pooling contributions together so > that if one of your users infringes somebody's copyright and they ask us > to take it down, all your app's contribution in that pool might go down > with it. The attribution mechanism could still be used to pinpoint the particular abusing user, it's just that the queries to filter by "actual user" would be more complex. I have personally investigated this and know for a fact that such filtering could be done. Are there other objections to this approach, especially Shawn's refinement of it? Its main fault is that /user/* accounts would no longer correspond to nice "units of provenance". Its only advantage is that it can be implemented today. > > I'm hoping to start testing my app with real users in 2-3 months. I'm > > really hoping someone has a better and/or more workable proposal than > > any of the ones I mentioned here. I'm all ears. > > AFAIK, there is an hybrid OpenID + OAuth protocol draft in the wild > > http://www.techcrunch.com/2009/01/29/openid-oauth-two-great-tastes-that-taste-great-together/ > http://googledataapis.blogspot.com/2009/01/bringing-openid-and-oauth-together.html > > but it doesn't seem like it's getting that much traction in the real > world (nor in the various expert groups, really) > > http://wiki.openid.net/OpenID-and-OAuth-Hybrid-Extension > > although its usability claims are pretty impressive > > http://www.readwriteweb.com/archives/comcast_property_sees_92_success_rate_openid.php This is the glorious future. I hope it will become the glorious present. > In short: it would be *awesome* if a user could contribute to freebase > with an identity that was hard to mechanically falsify (doesn't have to > be legally binding, just unique enough) and one that the user already > had (and hopefully was already signed into) instead of having to create > a new one specifically for Freebase. > > How to achieve that, well, that's the big question.... one thing is for > sure: if we supported just Google, Yahoo and Facebook accounts, that > would cover pretty much 99% of our potential contributors population so > we might just do that instead of waiting for a shiny protocol to get > finalized. Thanks for the suggestions. I'm still searching for something that can be implemented in a timely fashion and is a natural fit to Freebase. I intend to shepherd this along till I find something. Kurt :-) From iainsproat at gmail.com Mon Sep 14 08:50:50 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Mon, 14 Sep 2009 12:50:50 +0400 Subject: [Developers] accessing Acre from Git? Message-ID: Out of interest, does Acre have a Git interface? Iain From philip-freebase at shadowmagic.org.uk Mon Sep 14 12:16:52 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Mon, 14 Sep 2009 13:16:52 +0100 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090911124229.GK13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> Message-ID: <20090914121651.GU13273@sphinx.mythic-beasts.com> On Fri, Sep 11, 2009 at 01:42:29PM +0100, Philip Kendall wrote: > Some of you may have already seen Kirrily's tweet, but if not... a little > app while people may be interested in: > > http://schemaviz.freebaseapps.com/ Mildly updated version now published: * Node shapes now indicate the display style of the type (ellipse: standard, rectangle: CVT, triangle: enumeration) * New option to show included types via dotted edges (this deliberately excludes /common/topic) * New option to ignore "common types" (those in /type, eg integer, text, etc). This can make the graph significantly simpler and (I think) easier to understand in some cases (try American Football) * One bugfix: if a domain contains more than 100 types (I'm looking at you, /location), then the app would fail as it only fetched half the data. There's still an issue with /location in that the graphviz server times out if given the full graph, but it does work if you turn *off* "show included types" and turn *on" "ignore common types". Thanks to Jeff, Iain and Ed for their suggestions. Any more welcome :-) Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From iainsproat at gmail.com Mon Sep 14 14:10:50 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Mon, 14 Sep 2009 18:10:50 +0400 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090914121651.GU13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> Message-ID: Superb work Phil. I'd really like to see a link to this from the client (freebase.com)! I have a suggestion for a UI tweak - having the same functionality for both checkboxes works better. i.e. both "show" or both "ignore" but not mixed together. And one more request - could you refactor the infrastructure away from the UI (i.e. have the index file call a script which returns the image data)? This would let anyone else call the script just to get the image data, so could use your app as a service. Iain On Mon, Sep 14, 2009 at 4:16 PM, Philip Kendall wrote: > On Fri, Sep 11, 2009 at 01:42:29PM +0100, Philip Kendall wrote: >> Some of you may have already seen Kirrily's tweet, but if not... a little >> app while people may be interested in: >> >> http://schemaviz.freebaseapps.com/ > > Mildly updated version now published: > > * Node shapes now indicate the display style of the type (ellipse: > ?standard, rectangle: CVT, triangle: enumeration) > * New option to show included types via dotted edges (this deliberately > ?excludes /common/topic) > * New option to ignore "common types" (those in /type, eg integer, text, > ?etc). This can make the graph significantly simpler and (I think) > ?easier to understand in some cases (try American Football) > * One bugfix: if a domain contains more than 100 types (I'm looking at > ?you, /location), then the app would fail as it only fetched half the > ?data. > ?There's still an issue with /location in that the graphviz server > ?times out if given the full graph, but it does work if you turn *off* > ?"show included types" and turn *on" "ignore common types". > > Thanks to Jeff, Iain and Ed for their suggestions. Any more welcome :-) > > Cheers, > > Phil > > -- > ?Philip Kendall > ?http://www.shadowmagic.org.uk/ > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > From stefano at metaweb.com Mon Sep 14 17:26:00 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 10:26:00 -0700 Subject: [Developers] accessing Acre from Git? In-Reply-To: References: Message-ID: <4AAE7CA8.9070409@metaweb.com> Iain Sproat wrote: > Out of interest, does Acre have a Git interface? What do you mean by 'git interface'? -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From stefano at metaweb.com Mon Sep 14 17:28:36 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 10:28:36 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> Message-ID: <4AAE7D44.10607@metaweb.com> Iain Sproat wrote: > Superb work Phil. I'd really like to see a link to this from the > client (freebase.com)! We're thinking of linking/embedding schemaviz from the schema explorer domain views, how's that? > I have a suggestion for a UI tweak - having the same functionality for > both checkboxes works better. i.e. both "show" or both "ignore" but > not mixed together. > > And one more request - could you refactor the infrastructure away from > the UI (i.e. have the index file call a script which returns the image > data)? This would let anyone else call the script just to get the > image data, so could use your app as a service. +1, this would make it easier to experiment with alternative visualizations. > > Iain > > > > On Mon, Sep 14, 2009 at 4:16 PM, Philip Kendall > wrote: >> On Fri, Sep 11, 2009 at 01:42:29PM +0100, Philip Kendall wrote: >>> Some of you may have already seen Kirrily's tweet, but if not... a little >>> app while people may be interested in: >>> >>> http://schemaviz.freebaseapps.com/ >> Mildly updated version now published: >> >> * Node shapes now indicate the display style of the type (ellipse: >> standard, rectangle: CVT, triangle: enumeration) >> * New option to show included types via dotted edges (this deliberately >> excludes /common/topic) >> * New option to ignore "common types" (those in /type, eg integer, text, >> etc). This can make the graph significantly simpler and (I think) >> easier to understand in some cases (try American Football) >> * One bugfix: if a domain contains more than 100 types (I'm looking at >> you, /location), then the app would fail as it only fetched half the >> data. >> There's still an issue with /location in that the graphviz server >> times out if given the full graph, but it does work if you turn *off* >> "show included types" and turn *on" "ignore common types". >> >> Thanks to Jeff, Iain and Ed for their suggestions. Any more welcome :-) >> >> Cheers, >> >> Phil >> >> -- >> Philip Kendall >> http://www.shadowmagic.org.uk/ >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers >> > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From rfh at metaweb.com Mon Sep 14 17:40:01 2009 From: rfh at metaweb.com (Reilly Hayes) Date: Mon, 14 Sep 2009 10:40:01 -0700 Subject: [Developers] accessing Acre from Git? In-Reply-To: <4AAE7CA8.9070409@metaweb.com> References: <4AAE7CA8.9070409@metaweb.com> Message-ID: <72446BDD-C59A-4CC6-BB2F-DF673A961812@metaweb.com> Iain, Unless I am wrong, the answer to your question is no. Acre maintains version history in the graph (OTG) and document store (CDB). Stefano, git has a well defined remote repository interface. It allows users to clone the repository, get incremental changes, post new changes, etc. Given that Acre does not implement a full version control system, I doubt we could implement a git interface. -r On Sep 14, 2009, at 10:26 AM, Stefano Mazzocchi wrote: > Iain Sproat wrote: >> Out of interest, does Acre have a Git interface? > > What do you mean by 'git interface'? > > -- > Stefano Mazzocchi Application Catalyst > Metaweb Technologies, Inc. stefano at metaweb.com > ------------------------------------------------------------------- > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 2434 bytes Desc: not available Url : http://lists.freebase.com/pipermail/developers/attachments/20090914/47147959/attachment.bin From spatial.db at gmail.com Mon Sep 14 17:45:17 2009 From: spatial.db at gmail.com (Ed Laurent) Date: Mon, 14 Sep 2009 13:45:17 -0400 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> Message-ID: Love it even more! Here's an interesting comparison of ways to visualize FGDC compliant metadata schema: FGDC image map: http://www.fgdc.gov/csdgmgraphical/index.html Schema viz of MapCentral: http://schemaviz.freebaseapps.com/?domain=%2Fbase%2Fmapcentral&show_includes=on -Ed On Mon, Sep 14, 2009 at 10:10 AM, Iain Sproat wrote: > Superb work Phil. I'd really like to see a link to this from the > client (freebase.com)! > > I have a suggestion for a UI tweak - having the same functionality for > both checkboxes works better. i.e. both "show" or both "ignore" but > not mixed together. > > And one more request - could you refactor the infrastructure away from > the UI (i.e. have the index file call a script which returns the image > data)? This would let anyone else call the script just to get the > image data, so could use your app as a service. > > Iain > > > > On Mon, Sep 14, 2009 at 4:16 PM, Philip Kendall > wrote: > > On Fri, Sep 11, 2009 at 01:42:29PM +0100, Philip Kendall wrote: > >> Some of you may have already seen Kirrily's tweet, but if not... a > little > >> app while people may be interested in: > >> > >> http://schemaviz.freebaseapps.com/ > > > > Mildly updated version now published: > > > > * Node shapes now indicate the display style of the type (ellipse: > > standard, rectangle: CVT, triangle: enumeration) > > * New option to show included types via dotted edges (this deliberately > > excludes /common/topic) > > * New option to ignore "common types" (those in /type, eg integer, text, > > etc). This can make the graph significantly simpler and (I think) > > easier to understand in some cases (try American Football) > > * One bugfix: if a domain contains more than 100 types (I'm looking at > > you, /location), then the app would fail as it only fetched half the > > data. > > There's still an issue with /location in that the graphviz server > > times out if given the full graph, but it does work if you turn *off* > > "show included types" and turn *on" "ignore common types". > > > > Thanks to Jeff, Iain and Ed for their suggestions. Any more welcome :-) > > > > Cheers, > > > > Phil > > > > -- > > Philip Kendall > > http://www.shadowmagic.org.uk/ > > _______________________________________________ > > Developers mailing list > > Developers at freebase.com > > http://lists.freebase.com/mailman/listinfo/developers > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090914/8cd2e248/attachment.htm From tyler at metaweb.com Mon Sep 14 18:12:07 2009 From: tyler at metaweb.com (Tyler Pirtle) Date: Mon, 14 Sep 2009 11:12:07 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <4AAA7ADD.3000102@metaweb.com> Message-ID: <4AAE8777.6020207@metaweb.com> Found a bug for ya: http://schemaviz.freebaseapps.com/?domain=%2Fmusic Unhandled exception: TypeError: Cannot read property "attributes" from undefined >>> Philip Kendall wrote: >>> >>>> Some of you may have already seen Kirrily's tweet, but if not... a >>>> little >>>> app while people may be interested in: >>>> >>>> http://schemaviz.freebaseapps.com/ >>>> >>>> It should give you a visualisation of any domain you're interested >>>> in, >>>> showing how the types in that domain are linked by properties. >>>> >>>> * Nodes represent types, edges represent properties. >>>> * Reverse properties are shown in parentheses after the master >>>> property >>>> name. >>>> * Grey edges represent hidden properties >>>> * Blue nodes are types outside the domain being viewed, but which are >>>> linked to from the domain (the edge will be dashed if the master >>>> property is from the type outside the domain to that inside the >>>> domain) >>>> * Red nodes represent undocumented types >>>> * All nodes are clickable to view that type >>>> >>>> Please let me know of any bugs / feature requests. Thanks should >>>> obviously go to everyone at Metaweb for this, in particular the Acre >>>> developers and Alexander (for his Genealogy app which was the >>>> inspiration for this). >>>> >> >> > > > ------------------------------------------------------------------------ > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > From iainsproat at gmail.com Mon Sep 14 18:22:24 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Mon, 14 Sep 2009 22:22:24 +0400 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <4AAE7D44.10607@metaweb.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE7D44.10607@metaweb.com> Message-ID: On Mon, Sep 14, 2009 at 9:28 PM, Stefano Mazzocchi wrote: > We're thinking of linking/embedding schemaviz from the schema explorer > domain views, how's that? That works. >wo new entry points in the dev version: > >* /dot: returns the graphviz source for the graph[1] >* /json: returns a JSON object with two parameters: > + imageurl: URL for the picture > + imagemap: HTML imagemap to make the image clickable Awesome, thanks Phil Iain From stefano at metaweb.com Mon Sep 14 18:25:58 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 11:25:58 -0700 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <20090914012234.GC20986@spaceship.com> References: <20090913012455.GB20876@spaceship.com> <4AAD8465.7050805@metaweb.com> <20090914012234.GC20986@spaceship.com> Message-ID: <4AAE8AB6.9010205@metaweb.com> Kurt Bollacker wrote: > On Sun, Sep 13, 2009 at 04:46:45PM -0700, Stefano Mazzocchi wrote: >> NOTE: these are my personal opinions, not necessarily what Metaweb >> thinks or will do. >>> 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios >>> like this for which tools like this were built, so usage is >>> (hopefully) straightforward. >> OAuth (aka allowing 3rd-party-apps to act on behalf of a freebase user) >> support is there already and not just for Acre. > > That's great! I can find some OAuth docs for acre at: > > http://www.freebase.com/docs/acre/oauth > > that make reference to OAuth URLs for freebase.com, but where are the > docs of using OAuth to access Freebase? My case is probably more > difficult than the standard use case for OAuth because I need my > middle tier to do the Freebase writes (just like Acre does for acre > apps). uh? this is *precisely* the usecase for OAuth, unless I'm missing something. > I'll have to proxy all of the calls for the "OAuth dance" > through my middle tier in order to grab the cookie. This defies the purpose of OAuth entirely: if you have their cookies, your app can act as them, which basically means they have to trust your app entirely. To avoid this trust chain is the only reason why OAuth was created. > I wonder if there is a better way? > >> OpenID (aka allowing users to log into freebase using credentials from >> another one of their accounts) is *not*, but we are interested in >> providing the service, at least with a selected number of identity >> providers. This is because, like you, we would like to minimize the >> amount of effort a user has to go thru to be able to contribute to freebase. > > That's good. Are there any official Metaweb plans for such? https://bugs.freebase.com/browse/CLI-9031 Note, this is basically to enable OpenID support in freebase.com so that people can log into freebase with their OpenID accounts instead of having to sign up and contribute using the client. Niall just took a deeper look at the openid+oauth hybrid protocol and found out this: http://step2.googlecode.com/svn/spec/openid_oauth_extension/latest/openid_oauth_extension.html#anchor3 which basically says: In its current form, it addresses the use case where the OpenID Provider and OAuth Service Provider are the same service. which means that it's completely useless for this scenario as your app would be the oauth consumer, freebase the oauth service and, say, google the openid provider. This scenarios is a fully federated authentication/authorization system with 4 actors user -> app -> identity provider -> service provider I, for one, would be delighted to have a system like this in place, if only because if done well it would increase Freebase's appeal as a service platform considerably. Unfortunately, I don't know of any lightweight protocol (or combination of protocols) that enables this... although, admittedly, I haven't researched it that closely. >>> 3. (BAD) Freebase provides an API for account creation. I create an >>> account in Freebase for each of my users and write on their behalf. >>> It's still their account, but I set it up for them. The big >>> disadvantage is that the user does not have as much control and >>> awareness of their Freebase account. >> Oh boy, this is even worse than #2: spammers could be able to route >> around #2 by creating a myriad of fake accounts. Sure, we could wrap the >> service around captcha and the like, but again, the effort required to >> implement #1 is probably the same as implementing a good #3 and a lot >> easier to justify. > > The API for account creation could be made so that only a few trusted > users could use it. A trusted "account partner" could log in and > create an account, which would then be marked as having been created > by the partner. My guess is that the number of partners would be > small enough for some time, making them manageable for some time. This is true, yet the burden of account creation on the user remains, reducing your app ability to simplify the path to contribution. > That said, this is not my first choice for a solution. I don't want > to be in the account creation business. Neither does Freebase, honestly. I think we see it as a necessary evil at this point. >>> 4. (UGLY) I create a pool of Freebase accounts for my app to use. When >>> a user needs to write, I grab an account from the pool and use the >>> FB attribution mechanism to mark each write to correspond to the >>> user in my app. When I hit the daily write limit for that account, >>> I switch to another account in the pool. This approach is terrible >>> in so many ways I can't count-- yet it's the only one I can >>> implement today. >> Agreed is horrible and again, your pooling contributions together so >> that if one of your users infringes somebody's copyright and they ask us >> to take it down, all your app's contribution in that pool might go down >> with it. > > The attribution mechanism could still be used to pinpoint the > particular abusing user, it's just that the queries to filter by > "actual user" would be more complex. I have personally investigated > this and know for a fact that such filtering could be done. > > Are there other objections to this approach, especially Shawn's > refinement of it? Its main fault is that /user/* accounts would no > longer correspond to nice "units of provenance". Its only advantage > is that it can be implemented today. it's true it would be possible to add additional layers of provenance, but, frankly, it feels like a hack. At the same time, I agree that it's the simplest thing that could possibly work at this time. >>> I'm hoping to start testing my app with real users in 2-3 months. I'm >>> really hoping someone has a better and/or more workable proposal than >>> any of the ones I mentioned here. I'm all ears. >> AFAIK, there is an hybrid OpenID + OAuth protocol draft in the wild >> >> http://www.techcrunch.com/2009/01/29/openid-oauth-two-great-tastes-that-taste-great-together/ >> http://googledataapis.blogspot.com/2009/01/bringing-openid-and-oauth-together.html >> >> but it doesn't seem like it's getting that much traction in the real >> world (nor in the various expert groups, really) >> >> http://wiki.openid.net/OpenID-and-OAuth-Hybrid-Extension >> >> although its usability claims are pretty impressive >> >> http://www.readwriteweb.com/archives/comcast_property_sees_92_success_rate_openid.php > > This is the glorious future. I hope it will become the glorious > present. Unfortunately, as I mention above, it doesn't seem to help us as it's a hybrid protocol for only a very special usecase (the one where the authorizaiton provider is also the authentication provider, which is unfortunately not our case) >> In short: it would be *awesome* if a user could contribute to freebase >> with an identity that was hard to mechanically falsify (doesn't have to >> be legally binding, just unique enough) and one that the user already >> had (and hopefully was already signed into) instead of having to create >> a new one specifically for Freebase. >> >> How to achieve that, well, that's the big question.... one thing is for >> sure: if we supported just Google, Yahoo and Facebook accounts, that >> would cover pretty much 99% of our potential contributors population so >> we might just do that instead of waiting for a shiny protocol to get >> finalized. > > Thanks for the suggestions. I'm still searching for something that > can be implemented in a timely fashion and is a natural fit to > Freebase. I intend to shepherd this along till I find something. Does anybody else have suggestions on how Kurt's requirements could be achieved? -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From iainsproat at gmail.com Mon Sep 14 18:34:44 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Mon, 14 Sep 2009 22:34:44 +0400 Subject: [Developers] accessing Acre from Git? In-Reply-To: <72446BDD-C59A-4CC6-BB2F-DF673A961812@metaweb.com> References: <4AAE7CA8.9070409@metaweb.com> <72446BDD-C59A-4CC6-BB2F-DF673A961812@metaweb.com> Message-ID: Reilly - correct assumption, thanks for the quick answer. Iain On Mon, Sep 14, 2009 at 9:40 PM, Reilly Hayes wrote: > > Iain, > > Unless I am wrong, the answer to your question is no. ?Acre maintains > version history in the graph (OTG) and document store (CDB). > > > Stefano, > > git has a well defined remote repository interface. ?It allows users to > clone the repository, get incremental changes, post new changes, etc. Given > that Acre does not implement a full version control system, I doubt we could > implement a git interface. > > -r > > > On Sep 14, 2009, at 10:26 AM, Stefano Mazzocchi wrote: > >> Iain Sproat wrote: >>> >>> Out of interest, does Acre have a Git interface? >> >> What do you mean by 'git interface'? >> >> -- >> Stefano Mazzocchi ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Application Catalyst >> Metaweb Technologies, Inc. ? ? ? ? ? ? ? ? ? ? ?stefano at metaweb.com >> ------------------------------------------------------------------- >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > From stefano at metaweb.com Mon Sep 14 19:05:08 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 12:05:08 -0700 Subject: [Developers] accessing Acre from Git? In-Reply-To: References: <4AAE7CA8.9070409@metaweb.com> <72446BDD-C59A-4CC6-BB2F-DF673A961812@metaweb.com> Message-ID: <4AAE93E4.9000808@metaweb.com> Iain Sproat wrote: > Reilly - correct assumption, thanks for the quick answer. The reason why I asked is that GIT (unlike SVN) does not define an official protocol of interaction over the wire (there are several and they all pretty much require some sort of shell access) At one point, we did try to implement WebDAV on top of Acre apps to allow people to 'mount' them as DAV folders, but given previous experience with DAV clients compatibility hell, we decided to follow another route: Alex wrote a python app that allows you to 'fetch' an acre app and make a copy on disk and then to 'upload' it back. This has many advantages: 1) it allows you to edit Acre apps with your text editor/IDE of choice, in case you don't like the browser-based appeditor (which is totally understandable as advanced text editing is not precisely browser's best feature) 2) it allows you to do version control on your end with whatever version control tool you like.. this has the advantage, for example, of allowing branches and more complicated version control apps lifecycles without having us to replicate all that complexity inside Acre's guts. We're currently trying to polish it up for release and hopefully we'll be able to do that fairly soon. > Iain > > On Mon, Sep 14, 2009 at 9:40 PM, Reilly Hayes wrote: >> Iain, >> >> Unless I am wrong, the answer to your question is no. Acre maintains >> version history in the graph (OTG) and document store (CDB). >> >> >> Stefano, >> >> git has a well defined remote repository interface. It allows users to >> clone the repository, get incremental changes, post new changes, etc. Given >> that Acre does not implement a full version control system, I doubt we could >> implement a git interface. >> >> -r >> >> >> On Sep 14, 2009, at 10:26 AM, Stefano Mazzocchi wrote: >> >>> Iain Sproat wrote: >>>> Out of interest, does Acre have a Git interface? >>> What do you mean by 'git interface'? >>> >>> -- >>> Stefano Mazzocchi Application Catalyst >>> Metaweb Technologies, Inc. stefano at metaweb.com >>> ------------------------------------------------------------------- >>> >>> _______________________________________________ >>> Developers mailing list >>> Developers at freebase.com >>> http://lists.freebase.com/mailman/listinfo/developers >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers >> >> > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From stefano at metaweb.com Mon Sep 14 19:29:58 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 12:29:58 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090914121651.GU13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> Message-ID: <4AAE99B6.8070709@metaweb.com> Philip Kendall wrote: > On Fri, Sep 11, 2009 at 01:42:29PM +0100, Philip Kendall wrote: >> Some of you may have already seen Kirrily's tweet, but if not... a little >> app while people may be interested in: >> >> http://schemaviz.freebaseapps.com/ > > Mildly updated version now published: > > * Node shapes now indicate the display style of the type (ellipse: > standard, rectangle: CVT, triangle: enumeration) > * New option to show included types via dotted edges (this deliberately > excludes /common/topic) > * New option to ignore "common types" (those in /type, eg integer, text, > etc). This can make the graph significantly simpler and (I think) > easier to understand in some cases (try American Football) > * One bugfix: if a domain contains more than 100 types (I'm looking at > you, /location), then the app would fail as it only fetched half the > data. > There's still an issue with /location in that the graphviz server > times out if given the full graph, but it does work if you turn *off* > "show included types" and turn *on" "ignore common types". > > Thanks to Jeff, Iain and Ed for their suggestions. Any more welcome :-) Phil, I'm trying to integrate schemaviz into the freebase schema explorer and I'm thinking that it would be way cool to show directly a thumbnail for people to click on, for that I would need the schemaviz to be able to able to answer something like http://schemaviz.freebaseapps.com/image?domain=/whatever and redirect to the right image, that way I could embed it as a clickable thumbnail in the domain page of the schema explorer. Thoughts? -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From kurt at spaceship.com Mon Sep 14 21:27:33 2009 From: kurt at spaceship.com (Kurt Bollacker) Date: Mon, 14 Sep 2009 14:27:33 -0700 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <4AAE8AB6.9010205@metaweb.com> References: <20090913012455.GB20876@spaceship.com> <4AAD8465.7050805@metaweb.com> <20090914012234.GC20986@spaceship.com> <4AAE8AB6.9010205@metaweb.com> Message-ID: <20090914212733.GG20986@spaceship.com> On Mon, Sep 14, 2009 at 11:25:58AM -0700, Stefano Mazzocchi wrote: > Kurt Bollacker wrote: > > On Sun, Sep 13, 2009 at 04:46:45PM -0700, Stefano Mazzocchi wrote: > >> NOTE: these are my personal opinions, not necessarily what Metaweb > >> thinks or will do. > >>> 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios > >>> like this for which tools like this were built, so usage is > >>> (hopefully) straightforward. > >> OAuth (aka allowing 3rd-party-apps to act on behalf of a freebase user) > >> support is there already and not just for Acre. > > > > That's great! I can find some OAuth docs for acre at: > > > > http://www.freebase.com/docs/acre/oauth > > > > that make reference to OAuth URLs for freebase.com, but where are the > > docs of using OAuth to access Freebase? My case is probably more > > difficult than the standard use case for OAuth because I need my > > middle tier to do the Freebase writes (just like Acre does for acre > > apps). > > uh? this is *precisely* the usecase for OAuth, unless I'm missing something. > > > I'll have to proxy all of the calls for the "OAuth dance" > > through my middle tier in order to grab the cookie. > > This defies the purpose of OAuth entirely: if you have their cookies, > your app can act as them, which basically means they have to trust your > app entirely. To avoid this trust chain is the only reason why OAuth was > created. I won't claim I know all of the details about oauth, having only played with one provider. My comment was just due to the reading I had done that indicates that theory and reality of oauth implementations are often not in sync. My understanding is that the Oauth service provider can allow multiple levels of permissiveness and that at the less permissive levels, you have to use hacks like mentioned to get the desired functionality. I guess I should have been more clear. > Niall just took a deeper look at the openid+oauth hybrid protocol and > found out this: > > http://step2.googlecode.com/svn/spec/openid_oauth_extension/latest/openid_oauth_extension.html#anchor3 > > which basically says: > > In its current form, it addresses the use case where the OpenID Provider > and OAuth Service Provider are the same service. > > which means that it's completely useless for this scenario as your app > would be the oauth consumer, freebase the oauth service and, say, google > the openid provider. *sigh* > This scenarios is a fully federated authentication/authorization system > with 4 actors > > user -> app -> identity provider -> service provider > > I, for one, would be delighted to have a system like this in place, if > only because if done well it would increase Freebase's appeal as a > service platform considerably. > > Unfortunately, I don't know of any lightweight protocol (or combination > of protocols) that enables this... although, admittedly, I haven't > researched it that closely. I don't think there is a good general solution. Luckily, I only need one specific to Freebase access, so I am still hopeful. Kurt :-) From stefano at metaweb.com Mon Sep 14 22:09:20 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 15:09:20 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <4AAE99B6.8070709@metaweb.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> Message-ID: <4AAEBF10.9000202@metaweb.com> Stefano Mazzocchi wrote: > Philip Kendall wrote: >> On Fri, Sep 11, 2009 at 01:42:29PM +0100, Philip Kendall wrote: >>> Some of you may have already seen Kirrily's tweet, but if not... a little >>> app while people may be interested in: >>> >>> http://schemaviz.freebaseapps.com/ >> Mildly updated version now published: >> >> * Node shapes now indicate the display style of the type (ellipse: >> standard, rectangle: CVT, triangle: enumeration) >> * New option to show included types via dotted edges (this deliberately >> excludes /common/topic) >> * New option to ignore "common types" (those in /type, eg integer, text, >> etc). This can make the graph significantly simpler and (I think) >> easier to understand in some cases (try American Football) >> * One bugfix: if a domain contains more than 100 types (I'm looking at >> you, /location), then the app would fail as it only fetched half the >> data. >> There's still an issue with /location in that the graphviz server >> times out if given the full graph, but it does work if you turn *off* >> "show included types" and turn *on" "ignore common types". >> >> Thanks to Jeff, Iain and Ed for their suggestions. Any more welcome :-) > > Phil, > > I'm trying to integrate schemaviz into the freebase schema explorer and > I'm thinking that it would be way cool to show directly a thumbnail for > people to click on, for that I would need the schemaviz to be able to > able to answer something like > > http://schemaviz.freebaseapps.com/image?domain=/whatever > > and redirect to the right image, that way I could embed it as a > clickable thumbnail in the domain page of the schema explorer. > > Thoughts? Nevermind, found something close enough, here we go: http://schemas.freebaseapps.com/domain?id=/food -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From stefano at metaweb.com Mon Sep 14 22:21:36 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 15:21:36 -0700 Subject: [Developers] Freebase accounts and Webapp writes In-Reply-To: <20090914212733.GG20986@spaceship.com> References: <20090913012455.GB20876@spaceship.com> <4AAD8465.7050805@metaweb.com> <20090914012234.GC20986@spaceship.com> <4AAE8AB6.9010205@metaweb.com> <20090914212733.GG20986@spaceship.com> Message-ID: <4AAEC1F0.2010104@metaweb.com> Kurt Bollacker wrote: > On Mon, Sep 14, 2009 at 11:25:58AM -0700, Stefano Mazzocchi wrote: >> Kurt Bollacker wrote: >>> On Sun, Sep 13, 2009 at 04:46:45PM -0700, Stefano Mazzocchi wrote: >>>> NOTE: these are my personal opinions, not necessarily what Metaweb >>>> thinks or will do. >>>>> 1. (BEST) Freebase adds support for OpenID and OAuth. It's scenarios >>>>> like this for which tools like this were built, so usage is >>>>> (hopefully) straightforward. >>>> OAuth (aka allowing 3rd-party-apps to act on behalf of a freebase user) >>>> support is there already and not just for Acre. >>> That's great! I can find some OAuth docs for acre at: >>> >>> http://www.freebase.com/docs/acre/oauth >>> >>> that make reference to OAuth URLs for freebase.com, but where are the >>> docs of using OAuth to access Freebase? My case is probably more >>> difficult than the standard use case for OAuth because I need my >>> middle tier to do the Freebase writes (just like Acre does for acre >>> apps). >> uh? this is *precisely* the usecase for OAuth, unless I'm missing something. >> >>> I'll have to proxy all of the calls for the "OAuth dance" >>> through my middle tier in order to grab the cookie. >> This defies the purpose of OAuth entirely: if you have their cookies, >> your app can act as them, which basically means they have to trust your >> app entirely. To avoid this trust chain is the only reason why OAuth was >> created. > > I won't claim I know all of the details about oauth, having only > played with one provider. My comment was just due to the reading I > had done that indicates that theory and reality of oauth > implementations are often not in sync. My understanding is that the > Oauth service provider can allow multiple levels of permissiveness and > that at the less permissive levels, you have to use hacks like > mentioned to get the desired functionality. I guess I should have > been more clear. This frankly depends on the OAuth provide and what it does. Freebase is currently an OAuth provider and can use oauth *or* freebase.com cookies to authorize special services like 'mqlwrite' and 'get_user_info'. Since all Acre apps are all de-facto 3rd party apps since they are served from a different virtual host than freebase.com, I'm pretty confident in the real-life usefulness of our authorization granularity of our web services... but in case you find yourself in need for something we don't provide, you know we're all ears. >> Niall just took a deeper look at the openid+oauth hybrid protocol and >> found out this: >> >> http://step2.googlecode.com/svn/spec/openid_oauth_extension/latest/openid_oauth_extension.html#anchor3 >> >> which basically says: >> >> In its current form, it addresses the use case where the OpenID Provider >> and OAuth Service Provider are the same service. >> >> which means that it's completely useless for this scenario as your app >> would be the oauth consumer, freebase the oauth service and, say, google >> the openid provider. > > *sigh* Yeah, this is unfortunate. Luckily I have several friends in the OAuth and/or OpenID expert groups, so I'll ask around if they know anything more about this. >> This scenarios is a fully federated authentication/authorization system >> with 4 actors >> >> user -> app -> identity provider -> service provider >> >> I, for one, would be delighted to have a system like this in place, if >> only because if done well it would increase Freebase's appeal as a >> service platform considerably. >> >> Unfortunately, I don't know of any lightweight protocol (or combination >> of protocols) that enables this... although, admittedly, I haven't >> researched it that closely. > > I don't think there is a good general solution. Luckily, I only need > one specific to Freebase access, so I am still hopeful. Many of the critical Freebase APIs are OAuth-enabled *right now* and have been so for a while (since all acre apps heavily depend on them)... Acre is nothing special here, it just knows what to do and does it for you. We don't have public-facing documentation on how to do enable your app and request a consumer token/secret pair to enable your app. And this is something I plan to fix very soon, unfortunately there are some issues, mostly to do with how we know you own the domain you are asking an oauth token for: some OAuth implementations (say, Google's) are really paranoid about that while others (say, Yahoo!'s) are really not. We have to decide where we want to stand on that problem. -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From rfh at metaweb.com Mon Sep 14 22:24:50 2009 From: rfh at metaweb.com (Reilly Hayes) Date: Mon, 14 Sep 2009 15:24:50 -0700 Subject: [Developers] accessing Acre from Git? In-Reply-To: <4AAE93E4.9000808@metaweb.com> References: <4AAE7CA8.9070409@metaweb.com> <72446BDD-C59A-4CC6-BB2F-DF673A961812@metaweb.com> <4AAE93E4.9000808@metaweb.com> Message-ID: > > The reason why I asked is that GIT (unlike SVN) does not define an > official protocol of interaction over the wire (there are several and > they all pretty much require some sort of shell access) How does the git pack protocol not fulfill this requirement? Github has built a business around git and the pack protocol. -r -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090914/2f0c3583/attachment.htm -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 2434 bytes Desc: not available Url : http://lists.freebase.com/pipermail/developers/attachments/20090914/2f0c3583/attachment.bin From stefano at metaweb.com Mon Sep 14 23:07:32 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Mon, 14 Sep 2009 16:07:32 -0700 Subject: [Developers] accessing Acre from Git? In-Reply-To: References: <4AAE7CA8.9070409@metaweb.com> <72446BDD-C59A-4CC6-BB2F-DF673A961812@metaweb.com> <4AAE93E4.9000808@metaweb.com> Message-ID: <4AAECCB4.4000607@metaweb.com> Reilly Hayes wrote: >> >> The reason why I asked is that GIT (unlike SVN) does not define an >> official protocol of interaction over the wire (there are several and >> they all pretty much require some sort of shell access) > > How does the git pack protocol not fulfill this requirement? Github has > built a business around git and the pack protocol. You're right, we could have done that.... but you're also right saying that it's going to be pretty hard to implement something that will be completely round-trippable in the GIT world, unless, of course, our own internal versioning system was idempotent with GIT's and that's hard. -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From tfmorris at gmail.com Mon Sep 14 23:22:31 2009 From: tfmorris at gmail.com (Tom Morris) Date: Mon, 14 Sep 2009 19:22:31 -0400 Subject: [Developers] Flickr scavenging app? Message-ID: There was an app started at the last Hack Day which was designed, as I remember, to take a list of Freebase topics which had no images and query Flickr for candidate photos which were licensed appropriately for uploading to Freebase and then have the user pick the best match to be uploaded. Did it ever get finished? Anyone remember what it's called? I tried a bunch of different search terms on freebaseapps.com without success. Was it a non-Acre app? Any pointers appreciated... Tom From philip-freebase at shadowmagic.org.uk Tue Sep 15 07:08:55 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Tue, 15 Sep 2009 08:08:55 +0100 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <4AAEBF10.9000202@metaweb.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> Message-ID: <20090915070855.GA13273@sphinx.mythic-beasts.com> On Mon, Sep 14, 2009 at 03:09:20PM -0700, Stefano Mazzocchi wrote: > > Nevermind, found something close enough, here we go: > > http://schemas.freebaseapps.com/domain?id=/food Thanks! (and thanks Kirrily for the blog post) However... you may want to back this off for now, or at the very least wrap it in a try/catch block - some domains (including at least /time and /music) are erroring out for reasons I don't understand, and unfortunately I won't have time to look at this today and that's now taking down the schema viewer as well. [ If anyone wants to look at it, it's something to do with a reverse property being found when the master hasn't been found. I apologise for the lack of comments in the source. ] Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From philip-freebase at shadowmagic.org.uk Tue Sep 15 17:39:21 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Tue, 15 Sep 2009 18:39:21 +0100 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090915070855.GA13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> <20090915070855.GA13273@sphinx.mythic-beasts.com> Message-ID: <20090915173921.GB13273@sphinx.mythic-beasts.com> On Tue, Sep 15, 2009 at 08:08:55AM +0100, Philip Kendall wrote: > > However... you may want to back this off for now, or at the very least > wrap it in a try/catch block - some domains (including at least /time > and /music) are erroring out for reasons I don't understand, and > unfortunately I won't have time to look at this today and that's now > taking down the schema viewer as well. /time is now fixed in the dev version (a problem with /event/event also being /time/event). /music I can see what's causing the problem, but I don't understand what's going on. A small stanza from the query result: { "id": "/music/concert/concert_tour", "expected_type": "/music/concert_tour", "master_property": "/user/philg/default_domain/concert_tour/concerts" } which I find slightly strange as I would always expect the master property to be in the /music/concert_tour namespace. Is this a valid assumption? Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From bryan.cheung at metaweb.com Tue Sep 15 18:03:37 2009 From: bryan.cheung at metaweb.com (Bryan Cheung) Date: Tue, 15 Sep 2009 11:03:37 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090915173921.GB13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> <20090915070855.GA13273@sphinx.mythic-beasts.com> <20090915173921.GB13273@sphinx.mythic-beasts.com> Message-ID: <295E8DFF-BA6D-47ED-B61E-EC84B2C56F62@metaweb.com> > { > "id": "/music/concert/concert_tour", > "expected_type": "/music/concert_tour", > "master_property": "/user/philg/default_domain/concert_tour/concerts" > } > > Looks like Phil, one of the admins in /music created the reverse into one of his domains. I'll write up a jira to get this corrected. This may also have been overlooked in this complicated promotion: DA-564 (https://bugs.freebase.com/browse/DA-564) > which I find slightly strange as I would always expect the master > property to be in the /music/concert_tour namespace. Is this a valid > assumption? In general, commons schema should not have master/reverses or expected types to non-common schema. On Sep 15, 2009, at 10:39 AM, Philip Kendall wrote: > On Tue, Sep 15, 2009 at 08:08:55AM +0100, Philip Kendall wrote: >> >> However... you may want to back this off for now, or at the very >> least >> wrap it in a try/catch block - some domains (including at least /time >> and /music) are erroring out for reasons I don't understand, and >> unfortunately I won't have time to look at this today and that's now >> taking down the schema viewer as well. > > /time is now fixed in the dev version (a problem with /event/event > also > being /time/event). > > /music I can see what's causing the problem, but I don't understand > what's going on. A small stanza from the query result: > > { > "id": "/music/concert/concert_tour", > "expected_type": "/music/concert_tour", > "master_property": "/user/philg/default_domain/concert_tour/concerts" > } > > which I find slightly strange as I would always expect the master > property to be in the /music/concert_tour namespace. Is this a valid > assumption? > > Cheers, > > Phil > > -- > Philip Kendall > http://www.shadowmagic.org.uk/ > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From stefano at metaweb.com Tue Sep 15 18:59:15 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Tue, 15 Sep 2009 11:59:15 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090915070855.GA13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> <20090915070855.GA13273@sphinx.mythic-beasts.com> Message-ID: <4AAFE403.9080907@metaweb.com> Philip Kendall wrote: > On Mon, Sep 14, 2009 at 03:09:20PM -0700, Stefano Mazzocchi wrote: >> Nevermind, found something close enough, here we go: >> >> http://schemas.freebaseapps.com/domain?id=/food > > Thanks! (and thanks Kirrily for the blog post) > > However... you may want to back this off for now, or at the very least > wrap it in a try/catch block - some domains (including at least /time > and /music) are erroring out for reasons I don't understand, and > unfortunately I won't have time to look at this today and that's now > taking down the schema viewer as well. Will do. > [ If anyone wants to look at it, it's something to do with a reverse > property being found when the master hasn't been found. I apologise > for the lack of comments in the source. ] If you grant me access to the app, I can try to fix it. -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From rictic at gmail.com Wed Sep 16 03:45:26 2009 From: rictic at gmail.com (Peter Burns) Date: Tue, 15 Sep 2009 20:45:26 -0700 Subject: [Developers] Flickr scavenging app? In-Reply-To: References: Message-ID: <29DF4ADA-3152-4185-95EB-E322F42B185F@gmail.com> It has stuttered a bit at the moment. It's in Acre at http://acre.freebase.com/#app=/user/lethain/isfdb-titles-odesk-c&file=index The query done to flickr could use some improvement, and currently there's nothing to take decisions and write them back into Freebase. It will be a few months before I'll be free to work on it again. School has started up for me and my time is suddenly quite dear. If someone is interested in hacking on it some I'd be happy to answer any questions or anything. It's an Acre app, so forking encouraged :) -Peter On Sep 14, 2009, at 4:22 PM, Tom Morris wrote: > There was an app started at the last Hack Day which was designed, as I > remember, to take a list of Freebase topics which had no images and > query Flickr for candidate photos which were licensed appropriately > for uploading to Freebase and then have the user pick the best match > to be uploaded. > > Did it ever get finished? Anyone remember what it's called? I tried > a bunch of different search terms on freebaseapps.com without success. > Was it a non-Acre app? > > Any pointers appreciated... > > Tom > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 2415 bytes Desc: not available Url : http://lists.freebase.com/pipermail/developers/attachments/20090915/3474e104/attachment.bin From robert at metaweb.com Wed Sep 16 04:31:37 2009 From: robert at metaweb.com (Robert Cook) Date: Tue, 15 Sep 2009 21:31:37 -0700 Subject: [Developers] Flickr scavenging app? In-Reply-To: <29DF4ADA-3152-4185-95EB-E322F42B185F@gmail.com> References: <29DF4ADA-3152-4185-95EB-E322F42B185F@gmail.com> Message-ID: <1661CE70-7A5E-4B0B-AD44-D6179505D64B@metaweb.com> This ACRE app, written by Dae uses the Yahoo image search API to find candidates: http://odesk.freebaseapps.com/ You can add topic ids in the first box, and then /common/topic in the second. You'll then be able to edit the image property. Select the "Yahoo image search" tab and you'll get a gallery you can choose from. The idea behind this app, by the way, is to set up a queue of topics you can add data to in sequence, so you could put in a list of topic keys to edit. Note that images will be added as "Fair use" and served as a small size from the Freebase API even if the selected image has been released under a permissive license. R On Sep 15, 2009, at 8:45 PM, Peter Burns wrote: > It has stuttered a bit at the moment. It's in Acre at http://acre.freebase.com/#app > =/user/lethain/isfdb-titles-odesk-c&file=index > > The query done to flickr could use some improvement, and currently > there's nothing to take decisions and write them back into > Freebase. It will be a few months before I'll be free to work on it > again. School has started up for me and my time is suddenly quite > dear. If someone is interested in hacking on it some I'd be happy > to answer any questions or anything. It's an Acre app, so forking > encouraged :) > > -Peter > > On Sep 14, 2009, at 4:22 PM, Tom Morris wrote: > >> There was an app started at the last Hack Day which was designed, >> as I >> remember, to take a list of Freebase topics which had no images and >> query Flickr for candidate photos which were licensed appropriately >> for uploading to Freebase and then have the user pick the best match >> to be uploaded. >> >> Did it ever get finished? Anyone remember what it's called? I tried >> a bunch of different search terms on freebaseapps.com without >> success. >> Was it a non-Acre app? >> >> Any pointers appreciated... >> >> Tom >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From philip-freebase at shadowmagic.org.uk Wed Sep 16 14:40:37 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Wed, 16 Sep 2009 15:40:37 +0100 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <295E8DFF-BA6D-47ED-B61E-EC84B2C56F62@metaweb.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> <20090915070855.GA13273@sphinx.mythic-beasts.com> <20090915173921.GB13273@sphinx.mythic-beasts.com> <295E8DFF-BA6D-47ED-B61E-EC84B2C56F62@metaweb.com> Message-ID: <20090916144036.GD13273@sphinx.mythic-beasts.com> On Tue, Sep 15, 2009 at 11:03:37AM -0700, Bryan Cheung wrote: > > { > > "id": "/music/concert/concert_tour", > > "expected_type": "/music/concert_tour", > > "master_property": "/user/philg/default_domain/concert_tour/concerts" > > } > > > > > Looks like Phil, one of the admins in /music created the reverse into > one of his domains. Understood, but in that case I would have thought the expected_type would be /user/philg/default_domain/concert_tour, rather than /music/concert_tour (and the app would have coped with this case). It's the discrepancy between the expected_type and the schema of the master_property that is breaking the app. Putting this somewhat differently; in this query { "id": "/foo/bar/wibble/reverse_property_name", "type": "/type/property", "expected_type": null, "master_property": { "schema": null } } is it a valid assumption that expected_type and master_property.schema should be equal? For now, I've made the app log a warning in this case and carry on, so the /music domain can now be displayed. [ If you haven't found it, the same issue applies to /user/philg/default_domain/concert_tour/tour_manager ] Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From vishal at metaweb.com Wed Sep 16 22:30:38 2009 From: vishal at metaweb.com (Vishal Talwar) Date: Wed, 16 Sep 2009 15:30:38 -0700 (PDT) Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <98211837.92701253140058008.JavaMail.root@zimbra01.corp.sjc1.metaweb.com> Message-ID: <1662025027.92751253140238095.JavaMail.root@zimbra01.corp.sjc1.metaweb.com> Hi Philip, I think you have just described a constraint that says: for all results in { "id": null, "type": "/type/property", "r:expected_type": null, "r:schema": null, "master_property": { "id": null, "type": "/type/property", "m:expected_type": null, "m:schema": null } }, m:expected_type = r:schema & r:expected_type = m:schema I will add a gardening check for this. Vishal ----- Original Message ----- From: "Philip Kendall" To: developers at freebase.com Sent: Wednesday, September 16, 2009 7:40:37 AM GMT -08:00 US/Canada Pacific Subject: Re: [Developers] [Data-modeling] Schema Visualisation app On Tue, Sep 15, 2009 at 11:03:37AM -0700, Bryan Cheung wrote: > > { > > "id": "/music/concert/concert_tour", > > "expected_type": "/music/concert_tour", > > "master_property": "/user/philg/default_domain/concert_tour/concerts" > > } > > > > > Looks like Phil, one of the admins in /music created the reverse into > one of his domains. Understood, but in that case I would have thought the expected_type would be /user/philg/default_domain/concert_tour, rather than /music/concert_tour (and the app would have coped with this case). It's the discrepancy between the expected_type and the schema of the master_property that is breaking the app. Putting this somewhat differently; in this query { "id": "/foo/bar/wibble/reverse_property_name", "type": "/type/property", "expected_type": null, "master_property": { "schema": null } } is it a valid assumption that expected_type and master_property.schema should be equal? For now, I've made the app log a warning in this case and carry on, so the /music domain can now be displayed. [ If you haven't found it, the same issue applies to /user/philg/default_domain/concert_tour/tour_manager ] Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ _______________________________________________ Developers mailing list Developers at freebase.com http://lists.freebase.com/mailman/listinfo/developers From philip-freebase at shadowmagic.org.uk Thu Sep 17 10:29:32 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Thu, 17 Sep 2009 11:29:32 +0100 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <1662025027.92751253140238095.JavaMail.root@zimbra01.corp.sjc1.metaweb.com> References: <98211837.92701253140058008.JavaMail.root@zimbra01.corp.sjc1.metaweb.com> <1662025027.92751253140238095.JavaMail.root@zimbra01.corp.sjc1.metaweb.com> Message-ID: <20090917102932.GE13273@sphinx.mythic-beasts.com> On Wed, Sep 16, 2009 at 03:30:38PM -0700, Vishal Talwar wrote: > > I think you have just described a constraint that says: > > for all results in > { > "id": null, > "type": "/type/property", > "r:expected_type": null, > "r:schema": null, > "master_property": { > "id": null, > "type": "/type/property", > "m:expected_type": null, > "m:schema": null > } > }, > > m:expected_type = r:schema > & r:expected_type = m:schema Definitely (I think I realised this sometime last night as well :-) ). I think I find it slightly more intuitive as { "id": null, "type": "/type/property", "m:expected_type": null, "m:schema": null, "reverse_property": { "id": null, "type": "/type/property", "r:expected_type": null, "r:schema": null } } but they're fairly obviously the same thing :-) Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From philip-freebase at shadowmagic.org.uk Thu Sep 17 12:16:09 2009 From: philip-freebase at shadowmagic.org.uk (Philip Kendall) Date: Thu, 17 Sep 2009 13:16:09 +0100 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <295E8DFF-BA6D-47ED-B61E-EC84B2C56F62@metaweb.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> <20090915070855.GA13273@sphinx.mythic-beasts.com> <20090915173921.GB13273@sphinx.mythic-beasts.com> <295E8DFF-BA6D-47ED-B61E-EC84B2C56F62@metaweb.com> Message-ID: <20090917121609.GF13273@sphinx.mythic-beasts.com> On Tue, Sep 15, 2009 at 11:03:37AM -0700, Bryan Cheung wrote: > > In general, commons schema should not have master/reverses or expected > types to non-common schema. * /astronomy/astronomical_observatory/exoplanets_discovered -> /base/exoplanetology/exoplanet * /astronomy/astronomical_survey_project_organization/telescope -> /base/exoplanetology/telescope * /astronomy/star/planet_s -> /base/exoplanetology/exoplanet * /boats/ship_designer/boats_designed -> /user/jamie/default_domain/boat * /boats/warship_armament/armament -> /user/jamie/default_domain/naval_armament * /conferences/conference/venue -> /base/politicalconventions/convention_venue * /cricket/cricket_bowler_stats/best_bowling_in_a_match -> /user/aseem/default_domain/integer_ratio * /cricket/cricket_bowler_stats/best_bowling_in_an_inning -> /user/aseem/default_domain/integer_ratio * /cricket/cricket_bowler_stats/best_record -> /user/aseem/default_domain/integer_ratio * /film/film_company/film_ranches_owned -> /base/filmextras/film_ranch * /military/military_combatant/victories -> /user/skud/boats/naval_engagement * /spaceflight/satellite/launch_vehicle -> /user/patrick/satellite/rocket That list includes the following hidden properties, which I'm guessing aren't so important: * /conferences/conference_series/geographical_scope -> /base/events/geographical_area_for_recurring_events * /location/citytown/city_limits -> /user/jonathanwlowe/location/city_limits * /military/military_person/served_aboard -> /user/patrick/default_domain/warship_crew_tenure Cheers, Phil -- Philip Kendall http://www.shadowmagic.org.uk/ From bryan.cheung at metaweb.com Thu Sep 17 13:59:35 2009 From: bryan.cheung at metaweb.com (Bryan Cheung) Date: Thu, 17 Sep 2009 06:59:35 -0700 Subject: [Developers] [Data-modeling] Schema Visualisation app In-Reply-To: <20090917121609.GF13273@sphinx.mythic-beasts.com> References: <20090911124229.GK13273@sphinx.mythic-beasts.com> <20090914121651.GU13273@sphinx.mythic-beasts.com> <4AAE99B6.8070709@metaweb.com> <4AAEBF10.9000202@metaweb.com> <20090915070855.GA13273@sphinx.mythic-beasts.com> <20090915173921.GB13273@sphinx.mythic-beasts.com> <295E8DFF-BA6D-47ED-B61E-EC84B2C56F62@metaweb.com> <20090917121609.GF13273@sphinx.mythic-beasts.com> Message-ID: Thanks Phil. I'll use this as a starting point. I've opened DA-935 to generally evaluated all ects master/reverses of commons schema. On Sep 17, 2009, at 5:16 AM, Philip Kendall wrote: > On Tue, Sep 15, 2009 at 11:03:37AM -0700, Bryan Cheung wrote: >> >> In general, commons schema should not have master/reverses or >> expected >> types to non-common schema. > > > > * /astronomy/astronomical_observatory/exoplanets_discovered -> /base/ > exoplanetology/exoplanet > * /astronomy/astronomical_survey_project_organization/telescope -> / > base/exoplanetology/telescope > * /astronomy/star/planet_s -> /base/exoplanetology/exoplanet > * /boats/ship_designer/boats_designed -> /user/jamie/default_domain/ > boat > * /boats/warship_armament/armament -> /user/jamie/default_domain/ > naval_armament > * /conferences/conference/venue -> /base/politicalconventions/ > convention_venue > * /cricket/cricket_bowler_stats/best_bowling_in_a_match -> /user/ > aseem/default_domain/integer_ratio > * /cricket/cricket_bowler_stats/best_bowling_in_an_inning -> /user/ > aseem/default_domain/integer_ratio > * /cricket/cricket_bowler_stats/best_record -> /user/aseem/ > default_domain/integer_ratio > * /film/film_company/film_ranches_owned -> /base/filmextras/film_ranch > * /military/military_combatant/victories -> /user/skud/boats/ > naval_engagement > * /spaceflight/satellite/launch_vehicle -> /user/patrick/satellite/ > rocket > > That list includes the following hidden properties, which I'm guessing > aren't so important: > > * /conferences/conference_series/geographical_scope -> /base/events/ > geographical_area_for_recurring_events > * /location/citytown/city_limits -> /user/jonathanwlowe/location/ > city_limits > * /military/military_person/served_aboard -> /user/patrick/ > default_domain/warship_crew_tenure > > Cheers, > > Phil > > -- > Philip Kendall > http://www.shadowmagic.org.uk/ > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From spencerkelly86 at gmail.com Thu Sep 17 17:18:01 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Thu, 17 Sep 2009 13:18:01 -0400 Subject: [Developers] disappearing URI properties Message-ID: is there a reason why uri properties don't show in the client, even in edit mode? i've been using short-text so i can see what i'm doing.... (eg. http://www.freebase.com/edit/topic/guid/9202a8c04000641f8000000008ce6486) i've recently convinced elmcity (a m$ research project) to migrate to freebase, and it'd be nice if this curation could be done through the client without confusion. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090917/85c7b535/attachment.htm From jeff at metaweb.com Thu Sep 17 17:44:54 2009 From: jeff at metaweb.com (Jeff Prucher) Date: Thu, 17 Sep 2009 10:44:54 -0700 Subject: [Developers] disappearing URI properties In-Reply-To: References: Message-ID: Properties with an expected type of /type/uri or /common/webpage appear in the Weblinks box in the right sidebar. The property can be selected in the "type" column. If you want your URIs to appear in the body of the topic page, you need to set /freebase/property_hints/dont_display_in_weblinks to "true" (via the query editor) for that property. Jeff ________________________________ From: developers-bounces at freebase.com [mailto:developers-bounces at freebase.com] On Behalf Of Spencer Kelly Sent: Thursday, September 17, 2009 10:18 AM To: For discussions about MQL,Freebase API and apps built on Freebase Subject: [Developers] disappearing URI properties is there a reason why uri properties don't show in the client, even in edit mode? i've been using short-text so i can see what i'm doing.... (eg. http://www.freebase.com/edit/topic/guid/9202a8c04000641f8000000008ce6486) i've recently convinced elmcity (a m$ research project) to migrate to freebase, and it'd be nice if this curation could be done through the client without confusion. From gordon at metaweb.com Thu Sep 17 17:46:47 2009 From: gordon at metaweb.com (Gordon Mackenzie) Date: Thu, 17 Sep 2009 10:46:47 -0700 Subject: [Developers] disappearing URI properties In-Reply-To: References: Message-ID: <84083014-E0A1-4631-ABAB-D805442F212A@metaweb.com> Would the Popstra's Infidelity property have an example to suit you? http://www.freebase.com/edit/topic/en/meg_ryan Scroll on down to the Postra Celebrity type, there is an editable/ viewable URI property contained there. ~ Gordon <<< gordon at metaweb.com >>> On Sep 17, 2009, at 10:18 AM, Spencer Kelly wrote: > http://www.freebase.com/edit/topic/guid/9202a8c04000641f8000000008ce6486 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090917/f6c30c0e/attachment.htm From tfmorris at gmail.com Thu Sep 17 17:57:42 2009 From: tfmorris at gmail.com (Tom Morris) Date: Thu, 17 Sep 2009 13:57:42 -0400 Subject: [Developers] Flickr scavenging app? In-Reply-To: <1661CE70-7A5E-4B0B-AD44-D6179505D64B@metaweb.com> References: <29DF4ADA-3152-4185-95EB-E322F42B185F@gmail.com> <1661CE70-7A5E-4B0B-AD44-D6179505D64B@metaweb.com> Message-ID: Thanks for the pointers. I definitely never would have guessed those app names! Tom On Wed, Sep 16, 2009 at 12:31 AM, Robert Cook wrote: > This ACRE app, written by Dae uses the Yahoo image search API to find > candidates: > > http://odesk.freebaseapps.com/ > > You can add topic ids in the first box, and then /common/topic in the > second. ?You'll then be able to edit the image property. ?Select the > "Yahoo image search" tab and you'll get a gallery you can choose from. > > The idea behind this app, by the way, is to set up a queue of topics > you can add data to in sequence, so you could put in a list of topic > keys to edit. > > Note that images will be added as "Fair use" and served as a small > size from the Freebase API even if the selected image has been > released under a permissive license. > > R > > On Sep 15, 2009, at 8:45 PM, Peter Burns wrote: > >> It has stuttered a bit at the moment. ?It's in Acre at http://acre.freebase.com/#app >> =/user/lethain/isfdb-titles-odesk-c&file=index >> >> The query done to flickr could use some improvement, and currently >> there's nothing to take decisions and write them back into >> Freebase. ?It will be a few months before I'll be free to work on it >> again. ?School has started up for me and my time is suddenly quite >> dear. ?If someone is interested in hacking on it some I'd be happy >> to answer any questions or anything. ?It's an Acre app, so forking >> encouraged :) >> >> -Peter >> >> On Sep 14, 2009, at 4:22 PM, Tom Morris wrote: >> >>> There was an app started at the last Hack Day which was designed, >>> as I >>> remember, to take a list of Freebase topics which had no images and >>> query Flickr for candidate photos which were licensed appropriately >>> for uploading to Freebase and then have the user pick the best match >>> to be uploaded. >>> >>> Did it ever get finished? ?Anyone remember what it's called? ?I tried >>> a bunch of different search terms on freebaseapps.com without >>> success. >>> Was it a non-Acre app? >>> >>> Any pointers appreciated... >>> >>> Tom >>> _______________________________________________ >>> Developers mailing list >>> Developers at freebase.com >>> http://lists.freebase.com/mailman/listinfo/developers >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > From kirrily at metaweb.com Thu Sep 17 19:55:52 2009 From: kirrily at metaweb.com (Kirrily Robert) Date: Thu, 17 Sep 2009 12:55:52 -0700 Subject: [Developers] Freebase wiki Message-ID: <61451E93-ED0E-43A2-B707-871F5C59615F@metaweb.com> Hey everyone, We've set up a Mediawiki installation for community-contributed docs, recipes, FAQs, and more. You can check it out at http://wiki.freebase.com/ We'll need volunteers to trawl through the dev and d-m list archives and find interesting content to wikify. Anyone up for it? K. -- Kirrily Robert Freebase Community Director kirrily at metaweb.com http://freebase.com/ From kirrily at metaweb.com Thu Sep 17 20:10:14 2009 From: kirrily at metaweb.com (Kirrily Robert) Date: Thu, 17 Sep 2009 13:10:14 -0700 Subject: [Developers] [Data-modeling] Freebase wiki In-Reply-To: <0252333A-16F5-4C2C-B87B-A6F64095D3DC@fractalanalytics.com> References: <61451E93-ED0E-43A2-B707-871F5C59615F@metaweb.com> <0252333A-16F5-4C2C-B87B-A6F64095D3DC@fractalanalytics.com> Message-ID: <88A6E667-1B8C-4B80-B98B-73A4E673A72B@metaweb.com> Well, you could start with the archives either on http://lists.freebase.com/ or http://freebase.markmail.org/ (whichever you find nicer to work with) and then work back through time, looking for interesting posts that provide useful information, answer frequently asked questions, etc. Then go to the wiki and find (or make) a page on that topic, and paste the email in, linking to the original mailing list post. We can edit it up as we go forward. I've done an example at http://wiki.freebase.com/wiki/ISBNs K. On Sep 17, 2009, at 12:58 PM, Nirmal wrote: > Sure. What do I do? > > -myPhone > > On 18-Sep-2009, at 1:25 AM, Kirrily Robert > wrote: > >> Hey everyone, >> >> We've set up a Mediawiki installation for community-contributed docs, >> recipes, FAQs, and more. You can check it out at http://wiki.freebase.com/ >> >> We'll need volunteers to trawl through the dev and d-m list archives >> and find interesting content to wikify. Anyone up for it? >> >> K. >> >> -- >> Kirrily Robert >> Freebase Community Director >> kirrily at metaweb.com >> http://freebase.com/ >> >> >> >> >> _______________________________________________ >> Data-modeling mailing list >> Data-modeling at freebase.com >> http://lists.freebase.com/mailman/listinfo/data-modeling > _______________________________________________ > Data-modeling mailing list > Data-modeling at freebase.com > http://lists.freebase.com/mailman/listinfo/data-modeling -- Kirrily Robert Freebase Community Director kirrily at metaweb.com http://freebase.com/ From alecf at metaweb.com Wed Sep 23 22:08:22 2009 From: alecf at metaweb.com (Alec Flett) Date: Wed, 23 Sep 2009 15:08:22 -0700 Subject: [Developers] Flash and the Freebase API Message-ID: Flash developers: Due to the odd way the Flash API manages domain security, we are changing the way that Flash can access the APIs at freebase.com. We have deployed a read-only version of the freebase service, freebaseread.com to address the problem. Flash developers need to start making freebase API calls through http:// freebaseread.com rather than http://api.freebase.com On or after October 15th, we will remove the file 'crossdomain.xml' from api.freebase.com, effectively disabling Flash access to the APIs through that host. This ONLY affects flash applications. After this date, Flash applications will only be able to access Freebase APIs through freebaseread.com This file is now available at http://freebaseread.com/crossdomain.xml, which means you can update your flash applications to use the new domain today. This does mean that Flash applications cannot WRITE or UPLOAD to the freebase service using this new domain, and this is part of the security problem we are trying to fix. If this is a problem, please let us know - just drop me an e-mail and I'll be happy to discuss it. Alec From stefano at metaweb.com Wed Sep 23 22:54:36 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Wed, 23 Sep 2009 15:54:36 -0700 Subject: [Developers] Introducing ids.freebaseapps.com Message-ID: <4ABAA72C.4070600@metaweb.com> One of the interesting pieces of the Freebase dataset that not many people know about is the fact that when we harvest or link data from various sources, we retain their original identifiers and also the uri_templates that we can use to remap those local identifiers back to worldwide web identifiers. Since this is a very useful feature, I've spent a little time to make this information easier to use and create an Acre app that is capable of taking either a freebase id and return all the equivalent URIs for that concept or get a URI and try to find all the equivalent URIs. NOTE: The service is powered by both Freebase and SameAs (http://www.sameAs.org) and therefore is capable of surfacing all known equivalent identifiers for a given topic or URI starting from freebase and then via its mappings with dbpedia with the rest of the existing 'linking open data' effort. It also offers a JSONP service so that you can access it from your own web pages if you need to do so. Find it at: http://ids.freebaseapps.com/ -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From spencerkelly86 at gmail.com Thu Sep 24 01:15:06 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Wed, 23 Sep 2009 21:15:06 -0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: <4ABAA72C.4070600@metaweb.com> References: <4ABAA72C.4070600@metaweb.com> Message-ID: this is great stefano! i think it'd be neat to have a firefox extension that blinks when you are browsing linked data. it doesnt look like the uri search pops off sameas, bug? this is way cool. cheers -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090923/ffe385b1/attachment.htm From stefano at metaweb.com Thu Sep 24 02:18:45 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Wed, 23 Sep 2009 19:18:45 -0700 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> Message-ID: <4ABAD705.8070706@metaweb.com> Spencer Kelly wrote: > this is great stefano! > i think it'd be neat to have a firefox extension that blinks when you > are browsing linked data. this could be easily done, although some people would not want to enable it because it would basically relay all their browsing history to us (although they do it for phishing protection with google so it might not be that big of a problem in practice) A more effective (and portable way) would be a bookmarklet that takes the current URL, asks this service about it and then popup up a list of alternative locations on the web that talk about the same thing that this page is about. You would have to click it to activate it, but it would be even simpler to write and it would work on all browsers. > it doesnt look like the uri search pops off sameas, bug? not sure I understand what you mean, can you elaborate? > this is way cool. thanks! -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From narphorium at gmail.com Thu Sep 24 03:22:37 2009 From: narphorium at gmail.com (Shawn Simister) Date: Wed, 23 Sep 2009 23:22:37 -0400 Subject: [Developers] MQL result count deprecated? In-Reply-To: <4ABAD705.8070706@metaweb.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAD705.8070706@metaweb.com> Message-ID: <4ABAE5FD.4050509@gmail.com> I just noticed that one of my Acre applications isn't working because its expecting a "count" value that used to be returned by default (I think) in the query envelope of any MQL read. Was this feature formally deprecated or is this just a temporary glitch? Shawn From narphorium at gmail.com Thu Sep 24 03:51:15 2009 From: narphorium at gmail.com (Shawn Simister) Date: Wed, 23 Sep 2009 23:51:15 -0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> Message-ID: <4ABAECB3.1030100@gmail.com> What you're describing kind of reminds me of the Operator plugin and how it tells you when there's RDFa data on the page that you're viewing. As Stefano pointed out you can make a really simple version of this by creating a bookmark and pointing it at: javascript:window.open("http://ids.freebaseapps.com/?uri="+location.href); However, this won't tell you beforehand whether a given URL is "linked" and it doesn't seem to work for simple weblinks like http://www.freebase.com Shawn Spencer Kelly wrote: > this is great stefano! > i think it'd be neat to have a firefox extension that blinks when you > are browsing linked data. > it doesnt look like the uri search pops off sameas, bug? > this is way cool. > cheers > ------------------------------------------------------------------------ > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090923/1e6e7cca/attachment.htm From iainsproat at gmail.com Thu Sep 24 07:04:43 2009 From: iainsproat at gmail.com (Iain Sproat) Date: Thu, 24 Sep 2009 11:04:43 +0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: <4ABAECB3.1030100@gmail.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> Message-ID: Stefano, That's a great app, especially as it can be called from other apps to provide a service. I'd really like to see more apps which provide small API's so can be mashed up. (get a bit more of the DRY principle going amongst Acre apps) Coincidentally, I've been working on something very similar for URI matching only.... http://linker.freebaseapps.com (doesn't return any SameAs or other keys by default though). I'll share my thoughts on a few issues I'd run into: #1 - the list of URI templates is very small, so a lot of URIs don't work (yet). #2 - any app requiring data for a topic has to make 2 calls; the first to get the id from http://ids.freebaseapps.com; and another to the freebase api to get data from the topic with the corresponding id. This can have a lot of overhead. I worked around this by providing an 'extend' parameter that accepts users' custom MQL. #3 - external apps have to call the service for every single URI. If you're trying to match links on a page, this can be a few dozen or hundred calls. A 'queries' envelope is useful for optimizing external apps. #4 - users have to be aware that URIs are case sensitive, as are Freebase keys On Thu, Sep 24, 2009 at 7:51 AM, Shawn Simister wrote: > However, this won't tell you beforehand whether a given URL is "linked" and > it doesn't seem to work for simple weblinks like http://www.freebase.com Might this be solved with additional URI template of 'http://{key}', and having the key of www.freebase.com inserted into the Freebase topic? There's also an issue of whether an URI IS_A representation of a topic (such as a freebase, dbpedia, or wikipedia uri); or is a page with a SUBJECT of the topic (a newspaper article, or a blog post) which means that not all URIs are suitable. Iain From jason at metaweb.com Thu Sep 24 07:23:40 2009 From: jason at metaweb.com (Jason Douglas) Date: Thu, 24 Sep 2009 00:23:40 -0700 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> Message-ID: On Sep 24, 2009, at 12:04 AM, Iain Sproat wrote: > > #2 - any app requiring data for a topic has to make 2 calls; the first > to get the id from http://ids.freebaseapps.com; and another to the > freebase api to get data from the topic with the corresponding id. > This can have a lot of overhead. I worked around this by providing an > 'extend' parameter that accepts users' custom MQL. That's an ideal scenario for a MQL extension. We plan to promote that capability from labs to /api/service/mqlread pretty soon (not default- on though... will be triggered with an envelope parameter). Maybe we can all convince Stefano to have an extension in place by that launch? :-) -jason From stefano at metaweb.com Thu Sep 24 14:21:29 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Thu, 24 Sep 2009 07:21:29 -0700 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: <4ABAECB3.1030100@gmail.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> Message-ID: <4ABB8069.7080905@metaweb.com> Shawn Simister wrote: > What you're describing kind of reminds me of the Operator plugin > and how it tells > you when there's RDFa data on the page that you're viewing. the difference is that that check can be done locally without telling anybody else about where you are.... I doubt people would download a 200Gb extension to do the same in this context ;-) > As Stefano pointed out you can make a really simple version of this by > creating a bookmark and pointing it at: > > javascript:window.open("http://ids.freebaseapps.com/?uri="+location.href); I was thinking of something more sophisticated that would work in context and not make you leave the current page but yes, that's basically the idea. > However, this won't tell you beforehand whether a given URL is "linked" > and it doesn't seem to work for simple weblinks like http://www.freebase.com Well, www.freebase.com is not, per se, linked (and we don't treat http:// as the namespace for web links... although we could) And what do you mean by "linked"? "site that participates in the linking open data movement"? or else? -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From stefano at metaweb.com Thu Sep 24 14:34:58 2009 From: stefano at metaweb.com (Stefano Mazzocchi) Date: Thu, 24 Sep 2009 07:34:58 -0700 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> Message-ID: <4ABB8392.4080808@metaweb.com> Iain Sproat wrote: > Stefano, > > That's a great app, especially as it can be called from other apps to > provide a service. I'd really like to see more apps which provide > small API's so can be mashed up. (get a bit more of the DRY principle > going amongst Acre apps) Right, I like that too. Many pieces loosely joined... and let our community of practitioners show us what they use and what they don't. > Coincidentally, I've been working on something very similar for URI > matching only.... > http://linker.freebaseapps.com (doesn't return any SameAs or other > keys by default though). Ah cool. > I'll share my thoughts on a few issues I'd run into: > > #1 - the list of URI templates is very small, so a lot of URIs don't work (yet). Yes... but the first goal of my app was to show the concept and the potential benefits. I hope that peer pressure or volunteer work will do the rest. For example, from a legal point of view, crawling a web site and storing all their identifiers into freebase for 'link back' purposes is probably considered 'fair use' so you can imagine that writing a bunch of crawlers + reconciliation + data games that do this kind of things. (note: we're working really hard to make the new RABJ service available to external folks so that people can fork data games and/or build their own with the same power as ours.. that's actually my first priority these days) > #2 - any app requiring data for a topic has to make 2 calls; the first > to get the id from http://ids.freebaseapps.com; and another to the > freebase api to get data from the topic with the corresponding id. > This can have a lot of overhead. I worked around this by providing an > 'extend' parameter that accepts users' custom MQL. Sure, but I thought about crossing that bridge when I get there. > #3 - external apps have to call the service for every single URI. If > you're trying to match links on a page, this can be a few dozen or > hundred calls. A 'queries' envelope is useful for optimizing external > apps. Good idea.. but again, I'll implement it when somebody complains explicitly about it. > #4 - users have to be aware that URIs are case sensitive, as are Freebase keys Good point. > On Thu, Sep 24, 2009 at 7:51 AM, Shawn Simister wrote: >> However, this won't tell you beforehand whether a given URL is "linked" and >> it doesn't seem to work for simple weblinks like http://www.freebase.com > > Might this be solved with additional URI template of 'http://{key}', > and having the key of www.freebase.com inserted into the Freebase > topic? > > There's also an issue of whether an URI IS_A representation of a topic > (such as a freebase, dbpedia, or wikipedia uri); or is a page with a > SUBJECT of the topic (a newspaper article, or a blog post) which means > that not all URIs are suitable. Yes, this is more tricky than it seems at first. http://www.freebase.com/view/en/blah can be thought to mean many things: 1) a page on freebase.com 2) a page that has information about '/en/blah' 3) a wed id for '/en/blah' One suggestion would be to have a 'web domain' namespace and basically link '/en/freebase' with -> "freebase.com" and have a uri_template of "http://{key}/" and then you can make assertions about the web site that contains certain identifiers. Not sure how useful this would be though. -- Stefano Mazzocchi Application Catalyst Metaweb Technologies, Inc. stefano at metaweb.com ------------------------------------------------------------------- From brendan at metaweb.com Thu Sep 24 16:27:22 2009 From: brendan at metaweb.com (brendan) Date: Thu, 24 Sep 2009 09:27:22 -0700 Subject: [Developers] MQL result count deprecated? In-Reply-To: <4ABAE5FD.4050509@gmail.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAD705.8070706@metaweb.com> <4ABAE5FD.4050509@gmail.com> Message-ID: <75A1DED7-D392-400D-A763-F3D5D34E7C33@metaweb.com> count was not deprecated. Sounds like a bug. Could you provide the query, or, even better, open a bug with all the details. thanks! brendan On Sep 23, 2009, at 8:22 PM, Shawn Simister wrote: > I just noticed that one of my Acre applications isn't working because > its expecting a "count" value that used to be returned by default (I > think) in the query envelope of any MQL read. Was this feature > formally > deprecated or is this just a temporary glitch? > > Shawn > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090924/813cbf65/attachment.htm From narphorium at gmail.com Thu Sep 24 16:46:44 2009 From: narphorium at gmail.com (Shawn Simister) Date: Thu, 24 Sep 2009 12:46:44 -0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: <4ABB8069.7080905@metaweb.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8069.7080905@metaweb.com> Message-ID: <4ABBA274.1020006@gmail.com> Stefano Mazzocchi wrote: >> , this won't tell you beforehand whether a given URL is "linked" >> and it doesn't seem to work for simple weblinks like http://www.freebase.com >> > > Well, www.freebase.com is not, per se, linked (and we don't treat > http:// as the namespace for web links... although we could) And what do > you mean by "linked"? "site that participates in the linking open data > movement"? or else? > > Yeah, not really sure of the right terminology to use here. For example, if you look up /en/metaweb it will show that it links to http://www.metaweb.com/ . Not as part of a URI template but just as a web link in Freebase. It would be nice then be able to go to http://www.metaweb.com/ and click on a bookmarklet to see all the equivalent URIs for /en/metaweb. Shawn -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090924/5b6c96c5/attachment.htm From narphorium at gmail.com Thu Sep 24 16:51:06 2009 From: narphorium at gmail.com (Shawn Simister) Date: Thu, 24 Sep 2009 12:51:06 -0400 Subject: [Developers] MQL result count deprecated? In-Reply-To: <75A1DED7-D392-400D-A763-F3D5D34E7C33@metaweb.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAD705.8070706@metaweb.com> <4ABAE5FD.4050509@gmail.com> <75A1DED7-D392-400D-A763-F3D5D34E7C33@metaweb.com> Message-ID: <4ABBA37A.8000205@gmail.com> From what I remember, it worked with any query. So a simple query like this: [{ "id": null, "name": null, "type": "/astronomy/asteroid", "limit" : 100 }] Would return a query envelope that looked something like this: { "code": "/api/status/ok", "result": [ ... ], "count" : 4567, "status": "200 OK", "transaction_id": "cache;cache01.p01.sjc1:8101;2009-09-24T04:38:25Z;0014" } Where "count" was the number of results that the query produced regardless of the limit. At least thats what I remember. I wrote the code almost a year ago and now that the feature has disappeared I'm just basing this off of what values my code seems to expect. I've talked to John Giannandrea and Alec Flett about this and neither of them remember this being a feature so it could well be that I'm just losing my mind. Either way, I think I've found a work-around for my application now. Shawn brendan wrote: > count was not deprecated. Sounds like a bug. Could you provide the > query, or, even better, open a bug with > all the details. thanks! > > brendan > On Sep 23, 2009, at 8:22 PM, Shawn Simister wrote: > >> I just noticed that one of my Acre applications isn't working because >> its expecting a "count" value that used to be returned by default (I >> think) in the query envelope of any MQL read. Was this feature formally >> deprecated or is this just a temporary glitch? >> >> Shawn >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers > > ------------------------------------------------------------------------ > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090924/848ed129/attachment.htm From spencerkelly86 at gmail.com Thu Sep 24 18:28:30 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Thu, 24 Sep 2009 19:28:30 +0100 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: <4ABB8392.4080808@metaweb.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> Message-ID: > One suggestion would be to have a 'web domain' namespace and basically > link '/en/freebase' with -> "freebase.com" and have a uri_template of > "http://{key}/" and then you can make assertions about the web site that > contains certain identifiers. > > +1. this is a modelling problem. I've never been confident with how we model websites. and the task of connecting a billion domains to a billion topics seems like something we'd have accomplished already. /common/topic/webpage can't do it (too vague) and the 'website' type is broken, abandoned ( look at this discussion !) a type with a uri property should do it shouldn't it? how bout this? on this note, i dont think non-admins have write access to uri_template. could this be changed? -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090924/067c9eac/attachment.htm From tfmorris at gmail.com Sun Sep 27 20:00:22 2009 From: tfmorris at gmail.com (Tom Morris) Date: Sun, 27 Sep 2009 16:00:22 -0400 Subject: [Developers] Copy semantics of acre.freebase.extend_query ? Message-ID: What are the copy semantics of acre.freebase.extend_query? They don't appear to be documented. Will either of the following return a deep clone (a la jQuery.extend) of a query q? acre.freebase.extend_query({},q) acre.freebase.extend_query(q,{}) What's the easiest way in Acre to take a base query and create two independent copies based on it which do *not* share with each other? >From my limited testing, it appears that the straightforward case ends up creating objects which contain references to each other so that modifying one, modifies both, which is not what I want in this case. Tom From jason at metaweb.com Sun Sep 27 22:06:16 2009 From: jason at metaweb.com (Jason Douglas) Date: Sun, 27 Sep 2009 15:06:16 -0700 Subject: [Developers] Copy semantics of acre.freebase.extend_query ? In-Reply-To: References: Message-ID: <723871F9-A9D4-4E83-B5E9-27AB24DD0245@metaweb.com> The second argument is deep copied onto the first. The second argument can also take advantage of the path syntax used by the MQL sort command (dot-notation, including skipping over arrays). Intermediate nodes are also automatically created when using the path notation. I personally think of the second argument as basically a dictionary of patches of the form path : patch. So something like: acre.freebase.extend_query([{"id" : null}], {"name" : null, "/film/ film/performance.actor" : { "name" : null }); should result in a query that looks like: [{ "id" : null, "name" : null, "/film/film/perfomance" : { "actor" : { "name" : null } } }] -jason On Sep 27, 2009, at 1:00 PM, Tom Morris wrote: > What are the copy semantics of acre.freebase.extend_query? They don't > appear to be documented. > > Will either of the following return a deep clone (a la jQuery.extend) > of a query q? > > acre.freebase.extend_query({},q) > acre.freebase.extend_query(q,{}) > > What's the easiest way in Acre to take a base query and create two > independent copies based on it which do *not* share with each other? >> From my limited testing, it appears that the straightforward case >> ends > up creating objects which contain references to each other so that > modifying one, modifies both, which is not what I want in this case. > > Tom > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers From jason at metaweb.com Sun Sep 27 22:41:16 2009 From: jason at metaweb.com (Jason Douglas) Date: Sun, 27 Sep 2009 15:41:16 -0700 Subject: [Developers] Copy semantics of acre.freebase.extend_query ? In-Reply-To: References: Message-ID: Oops. Didn't really answer your question... On Sep 27, 2009, at 1:00 PM, Tom Morris wrote: > What are the copy semantics of acre.freebase.extend_query? They don't > appear to be documented. > > Will either of the following return a deep clone (a la jQuery.extend) > of a query q? > > acre.freebase.extend_query({},q) > acre.freebase.extend_query(q,{}) The first argument is the query object and it is also what is returned by the function. So that first syntax should actually create a new query object from q, whereas the second would just return q. > > What's the easiest way in Acre to take a base query and create two > independent copies based on it which do *not* share with each other? Your first syntax may just be the simplest way. I'll defer to Alex to confirm that though. >> From my limited testing, it appears that the straightforward case >> ends > up creating objects which contain references to each other so that > modifying one, modifies both, which is not what I want in this case. I believe that's true. var q = {"id" : null}; var query1 = acre.freebase.extend_query(q, patch1); var query2 = acre.freebase.extend_query(q, patch2); Would result in query1 === query2. I suppose you could reverse it, as you say: var q = {"id" : null}; var query1 = acre.freebase.extend_query(patch1, q); var query2 = acre.freebase.extend_query(patch2, q); .... to get two independent objects. I'll defer to Alex to confirm that though. -jason From alexbl at metaweb.com Sun Sep 27 22:58:48 2009 From: alexbl at metaweb.com (Alexander Botero-Lowry) Date: Sun, 27 Sep 2009 15:58:48 -0700 Subject: [Developers] Copy semantics of acre.freebase.extend_query ? In-Reply-To: References: Message-ID: <16136.1254092328@temperantia> I'm seeing some subtle hints in this email that I'm supposed to respond to it... ;) > > Will either of the following return a deep clone (a la jQuery.extend) > > of a query q? > > > > acre.freebase.extend_query({},q) > > acre.freebase.extend_query(q,{}) > > The first argument is the query object and it is also what is returned > by the function. So that first syntax should actually create a new > query object from q, whereas the second would just return q. > Yeah.. sort of. It won't be deep though. Which may work for you, but may not. So if you have something like var q = {'name':{'value':null, 'lang':null}} and use the [{},q] application of extend_query, you'll be sharing the same object hanging off the name key, so if you modify one, you modify the other. > Your first syntax may just be the simplest way. I'll defer to Alex to > confirm that though. > Recursive copy function: function deep_copy(obj) { if (obj == null) return obj; var nobj = {}; for (var a in obj) { if (typeof obj[a] === 'object') nobj[a] = deep_copy(obj[a]) else nobj[a] = obj[a]; } return nobj; } Mind you that's untested, but it should do the trick. Alex From tfmorris at gmail.com Sun Sep 27 23:43:23 2009 From: tfmorris at gmail.com (Tom Morris) Date: Sun, 27 Sep 2009 19:43:23 -0400 Subject: [Developers] Copy semantics of acre.freebase.extend_query ? In-Reply-To: <16136.1254092328@temperantia> References: <16136.1254092328@temperantia> Message-ID: Thanks for the replies. Alex's description matches the behavior that I'm seeing. It would be nice to have a version of the extend method which does do a deep copy. In the mean time, highlighting the shared semantics in the docs would help avoid people shooting themselves in the foot. In my case, it was obvious right away that there was a problem, but I can certainly imagine circumstance which would be much less obvious to the programmer. Tom From spencerkelly86 at gmail.com Mon Sep 28 19:55:09 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Mon, 28 Sep 2009 15:55:09 -0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> Message-ID: if this type is a go, i'll begin filling it. just a basic 'is a' for connecting topics+domains.... like dbpprop:website . http://www.freebase.com/edit/topic/en/bruce_cockburn?domain=%2Fbase%2Fwebsites#/base/websites/website afaik, a 'what's its website' query isnt possible without this type. On Thu, Sep 24, 2009 at 2:28 PM, Spencer Kelly wrote: > > One suggestion would be to have a 'web domain' namespace and basically >> link '/en/freebase' with -> "freebase.com" and have a uri_template of >> "http://{key}/" and then you can make assertions about the web site that >> contains certain identifiers. >> >> > +1. this is a modelling problem. I've never been confident with how we > model websites. and the task of connecting a billion domains to a billion > topics seems like something we'd have accomplished already. > /common/topic/webpage can't do it (too vague) > and the 'website' type is broken, abandoned ( look at this discussion > !) > > a type with a uri property should do it shouldn't it? how bout this? > > on this note, i dont think non-admins have write access to uri_template. > could this be changed? > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090928/f1490053/attachment.htm From jeff at metaweb.com Mon Sep 28 20:56:18 2009 From: jeff at metaweb.com (Jeff Prucher) Date: Mon, 28 Sep 2009 13:56:18 -0700 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> Message-ID: I have reservations about typing legal entities (people, companies, organizations, etc.) as "Official Websites" -- these things are not websites (official or otherwise), they have websites. Would adding an "official website" property to /common/topic suffice? (This would make it one of the default drop-down choices in the type column of the "Weblinks" box.) That would allow for semantic queries about "what is this topic's website" without requiring a new cotype. Common seems like a sensible place to have this, since pretty much anything (including books and movies) can have an official website. Jeff _____ From: developers-bounces at freebase.com [mailto:developers-bounces at freebase.com] On Behalf Of Spencer Kelly Sent: Thursday, September 24, 2009 11:29 AM To: For discussions about MQL,Freebase API and apps built on Freebase Subject: Re: [Developers] Introducing ids.freebaseapps.com One suggestion would be to have a 'web domain' namespace and basically link '/en/freebase' with -> "freebase.com" and have a uri_template of "http://{key}/" and then you can make assertions about the web site that contains certain identifiers. +1. this is a modelling problem. I've never been confident with how we model websites. and the task of connecting a billion domains to a billion topics seems like something we'd have accomplished already. /common/topic/webpage can't do it (too vague) and the 'website' type is broken, abandoned ( look at this discussion!) a type with a uri property should do it shouldn't it? how bout this ? on this note, i dont think non-admins have write access to uri_template. could this be changed? -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090928/7985035a/attachment.htm From jeff at metaweb.com Tue Sep 29 04:23:22 2009 From: jeff at metaweb.com (Jeff Prucher) Date: Mon, 28 Sep 2009 21:23:22 -0700 Subject: [Developers] Transit system length schema change Message-ID: <8BCF3C80B39B4E5AAF20362A7642F871@amd> I posted this some time ago, and it met with no objections, but the task was never executed. Since it's been some time, I'm resubmitting it to the list in case anyone has created new code that uses this property: The type Transit System has a property System Length (/metropolitan_transit/transit_system/system_length) which expects a unique floating point number. Tfmorris has suggested that this property should really be a CVT with values for date and length, and should therefore additionally be non-unique. I think this makes good sense; would this interfere with anyone's code? Or does anyone have further comments on this? Jeff From jeff at metaweb.com Tue Sep 29 04:24:39 2009 From: jeff at metaweb.com (Jeff Prucher) Date: Mon, 28 Sep 2009 21:24:39 -0700 Subject: [Developers] Date-mediating Website/Website Owner types Message-ID: <2F8FC0D014F9492DB6C104E78D442F3A@amd> I posted this some time ago, and it met with no objections, but the task was never executed. Since it's been some time, I'm resubmitting it to the list in case anyone has created new code that uses these properties: There's been a request to add a way to store "From" and "To" dates for Websites and Website Owners. This will involve breaking the link between the properties /internet/website/owner and /internet/website_owner/websites_owned. Will this break anyone's code? Jeff From spencerkelly86 at gmail.com Tue Sep 29 17:52:37 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Tue, 29 Sep 2009 13:52:37 -0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> Message-ID: +1 fanwebsites and other forms of websites can be handled off-common/topic. cheers;) On Mon, Sep 28, 2009 at 4:56 PM, Jeff Prucher wrote: > I have reservations about typing legal entities (people, companies, > organizations, etc.) as "Official Websites" -- these things are not websites > (official or otherwise), they have websites. Would adding an "official > website" property to /common/topic suffice? (This would make it one of the > default drop-down choices in the type column of the "Weblinks" box.) That > would allow for semantic queries about "what is this topic's website" > without requiring a new cotype. Common seems like a sensible place to have > this, since pretty much anything (including books and movies) can have an > official website. > > Jeff > > ------------------------------ > *From:* developers-bounces at freebase.com [mailto: > developers-bounces at freebase.com] *On Behalf Of *Spencer Kelly > *Sent:* Thursday, September 24, 2009 11:29 AM > *To:* For discussions about MQL,Freebase API and apps built on Freebase > *Subject:* Re: [Developers] Introducing ids.freebaseapps.com > > > One suggestion would be to have a 'web domain' namespace and basically >> link '/en/freebase' with -> "freebase.com" and have a uri_template of >> "http://{key}/" and then you can make assertions about the web site that >> contains certain identifiers. >> >> > +1. this is a modelling problem. I've never been confident with how we > model websites. and the task of connecting a billion domains to a billion > topics seems like something we'd have accomplished already. > /common/topic/webpage can't do it (too vague) > and the 'website' type is broken, abandoned ( look at this discussion > !) > > a type with a uri property should do it shouldn't it? how bout this? > > on this note, i dont think non-admins have write access to uri_template. > could this be changed? > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090929/dc53a40c/attachment-0001.htm From spatial.db at gmail.com Tue Sep 29 17:56:42 2009 From: spatial.db at gmail.com (Ed Laurent) Date: Tue, 29 Sep 2009 13:56:42 -0400 Subject: [Developers] Introducing ids.freebaseapps.com In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> Message-ID: +1 -Ed On Tue, Sep 29, 2009 at 1:52 PM, Spencer Kelly wrote: > +1 > fanwebsites and other forms of websites can be handled off-common/topic. > cheers;) > > On Mon, Sep 28, 2009 at 4:56 PM, Jeff Prucher wrote: > >> I have reservations about typing legal entities (people, companies, >> organizations, etc.) as "Official Websites" -- these things are not websites >> (official or otherwise), they have websites. Would adding an "official >> website" property to /common/topic suffice? (This would make it one of the >> default drop-down choices in the type column of the "Weblinks" box.) That >> would allow for semantic queries about "what is this topic's website" >> without requiring a new cotype. Common seems like a sensible place to have >> this, since pretty much anything (including books and movies) can have an >> official website. >> >> Jeff >> >> ------------------------------ >> *From:* developers-bounces at freebase.com [mailto: >> developers-bounces at freebase.com] *On Behalf Of *Spencer Kelly >> *Sent:* Thursday, September 24, 2009 11:29 AM >> *To:* For discussions about MQL,Freebase API and apps built on Freebase >> *Subject:* Re: [Developers] Introducing ids.freebaseapps.com >> >> >> One suggestion would be to have a 'web domain' namespace and basically >>> link '/en/freebase' with -> "freebase.com" and have a uri_template of >>> "http://{key}/" and then you can make assertions about the web site that >>> contains certain identifiers. >>> >>> >> +1. this is a modelling problem. I've never been confident with how we >> model websites. and the task of connecting a billion domains to a billion >> topics seems like something we'd have accomplished already. >> /common/topic/webpage can't do it (too vague) >> and the 'website' type is broken, abandoned ( look at this discussion >> !) >> >> a type with a uri property should do it shouldn't it? how bout this? >> >> on this note, i dont think non-admins have write access to uri_template. >> could this be changed? >> >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers >> >> > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090929/3e3931fe/attachment.htm From jeff at metaweb.com Tue Sep 29 21:08:20 2009 From: jeff at metaweb.com (Jeff Prucher) Date: Tue, 29 Sep 2009 14:08:20 -0700 Subject: [Developers] New "official website" property on /common/topic In-Reply-To: References: <4ABAA72C.4070600@metaweb.com><4ABAECB3.1030100@gmail.com><4ABB8392.4080808@metaweb.com> Message-ID: I'm retitleing this thread so that people who have been ignoring the "ids" thread might notice it. Since we don't like to make changes to /common/topic too cavalierly, even non-breaking changes, I want to make sure anyone else who might be interested sees this. To recap: I'm proposing adding a new property (expected type of /common/webpage) for official websites. This would also users to make a query-able assertion about the relationship of the official website to a topic. (More details downthread, below.) Jeff _____ From: developers-bounces at freebase.com [mailto:developers-bounces at freebase.com] On Behalf Of Ed Laurent Sent: Tuesday, September 29, 2009 10:57 AM To: For discussions about MQL,Freebase API and apps built on Freebase Subject: Re: [Developers] Introducing ids.freebaseapps.com +1 -Ed On Tue, Sep 29, 2009 at 1:52 PM, Spencer Kelly wrote: +1 fanwebsites and other forms of websites can be handled off-common/topic. cheers;) On Mon, Sep 28, 2009 at 4:56 PM, Jeff Prucher wrote: I have reservations about typing legal entities (people, companies, organizations, etc.) as "Official Websites" -- these things are not websites (official or otherwise), they have websites. Would adding an "official website" property to /common/topic suffice? (This would make it one of the default drop-down choices in the type column of the "Weblinks" box.) That would allow for semantic queries about "what is this topic's website" without requiring a new cotype. Common seems like a sensible place to have this, since pretty much anything (including books and movies) can have an official website. Jeff _____ From: developers-bounces at freebase.com [mailto:developers-bounces at freebase.com] On Behalf Of Spencer Kelly Sent: Thursday, September 24, 2009 11:29 AM To: For discussions about MQL,Freebase API and apps built on Freebase Subject: Re: [Developers] Introducing ids.freebaseapps.com One suggestion would be to have a 'web domain' namespace and basically link '/en/freebase' with -> "freebase.com" and have a uri_template of "http://{key}/" and then you can make assertions about the web site that contains certain identifiers. +1. this is a modelling problem. I've never been confident with how we model websites. and the task of connecting a billion domains to a billion topics seems like something we'd have accomplished already. /common/topic/webpage can't do it (too vague) and the 'website' type is broken, abandoned ( look at this discussion !) a type with a uri property should do it shouldn't it? how bout this ? on this note, i dont think non-admins have write access to uri_template. could this be changed? _______________________________________________ Developers mailing list Developers at freebase.com http://lists.freebase.com/mailman/listinfo/developers _______________________________________________ Developers mailing list Developers at freebase.com http://lists.freebase.com/mailman/listinfo/developers -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090929/150176dd/attachment.htm From narphorium at gmail.com Tue Sep 29 22:14:24 2009 From: narphorium at gmail.com (Shawn Simister) Date: Tue, 29 Sep 2009 18:14:24 -0400 Subject: [Developers] New "official website" property on /common/topic In-Reply-To: References: <4ABAA72C.4070600@metaweb.com><4ABAECB3.1030100@gmail.com><4ABB8392.4080808@metaweb.com> Message-ID: <4AC286C0.3070109@gmail.com> Isn't /common/topic a little too general? Most of the topics that would have an official topic would likely be companies or organizations. Some people also have official websites although most people these days just have a personal blog (which I assume is different?). Furthermore I think there are types like location where the official website may be ambiguous. Would you use the tourism department's website for that region or the local government's website? Don't get me wrong, I think this is a really good idea overall. I just want to make sure that it doesn't become the target of future edit wars. Shawn Jeff Prucher wrote: > I'm retitleing this thread so that people who have been ignoring the > "ids" thread might notice it. Since we don't like to make changes to > /common/topic too cavalierly, even non-breaking changes, I want to > make sure anyone else who might be interested sees this. To recap: I'm > proposing adding a new property (expected type of /common/webpage) for > official websites. This would also users to make a query-able > assertion about the relationship of the official website to a topic. > (More details downthread, below.) > Jeff > > ------------------------------------------------------------------------ > *From:* developers-bounces at freebase.com > [mailto:developers-bounces at freebase.com] *On Behalf Of *Ed Laurent > *Sent:* Tuesday, September 29, 2009 10:57 AM > *To:* For discussions about MQL,Freebase API and apps built on > Freebase > *Subject:* Re: [Developers] Introducing ids.freebaseapps.com > > +1 > > -Ed > > On Tue, Sep 29, 2009 at 1:52 PM, Spencer Kelly > > wrote: > > +1 > fanwebsites and other forms of websites can be handled > off-common/topic. > cheers;) > > On Mon, Sep 28, 2009 at 4:56 PM, Jeff Prucher > > wrote: > > I have reservations about typing legal entities (people, > companies, organizations, etc.) as "Official Websites" -- > these things are not websites (official or otherwise), > they have websites. Would adding an "official website" > property to /common/topic suffice? (This would make it one > of the default drop-down choices in the type column of the > "Weblinks" box.) That would allow for semantic queries > about "what is this topic's website" without requiring a > new cotype. Common seems like a sensible place to have > this, since pretty much anything (including books and > movies) can have an official website. > > Jeff > > ------------------------------------------------------------------------ > *From:* developers-bounces at freebase.com > > [mailto:developers-bounces at freebase.com > ] *On Behalf > Of *Spencer Kelly > *Sent:* Thursday, September 24, 2009 11:29 AM > *To:* For discussions about MQL,Freebase API and apps > built on Freebase > *Subject:* Re: [Developers] Introducing > ids.freebaseapps.com > > > One suggestion would be to have a 'web domain' > namespace and basically > link '/en/freebase' with -> "freebase.com > " and have a uri_template of > "http://{key}/" and then you can make assertions > about the web site that > contains certain identifiers. > > > +1. this is a modelling problem. I've never been > confident with how we model websites. and the task of > connecting a billion domains to a billion topics seems > like something we'd have accomplished already. > /common/topic/webpage can't do it (too vague) > and the 'website' type is broken, abandoned ( look at > this discussion > !) > > a type with a uri property should do it shouldn't it? > how bout this > > ? > > on this note, i dont think non-admins have write > access to uri_template. could this be changed? > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > > ------------------------------------------------------------------------ > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090929/ef72aa10/attachment-0001.htm From jack.alves at gmail.com Wed Sep 30 16:22:54 2009 From: jack.alves at gmail.com (Jack Alves) Date: Wed, 30 Sep 2009 09:22:54 -0700 Subject: [Developers] New "official website" property on /common/topic In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> Message-ID: <554723b0909300922w220d236axbb67fc2cb9a83054@mail.gmail.com> Intended usage of /common/topic/webpage property has always been fuzzy. My understanding is that the existing /common/topic/webpage property is intended for "official" websites. I think it would be great to have a generic property for webpage relationships to topics. With the proposed change, it sounds like the existing topic/webpage could be used for any related webpage. I think it would be more useful to keep the existing webpage property for official websites and create a more versatile mediator property for generic webpage relationships. Since the web is so loosely structured, additional information about the relatedness would be useful. For instance, the link may be to a news article or blog post that "describes", "reviews" or "mentions" the topic. The link may be to a page that represents the topic in a positive, negative, or neutral way. The link may be strongly or weakly related. This may be over anticipating usage but the point is to consider how a change would be helpful beyond a specific case. Changes to common/topic are rare. This is an opportunity for people to suggest practical usage for a new property. At some point, someone will build a bookmarking tool to connect topics to related websites. This is a great time to consider what that tool might need to advance tagging functionality while keeping the barrier to contributions low. Okay, okay, I understand the ultimate objective is to create new types with properties for different kinds of relationships. I just think it would be useful to lower the barrier for creating new kinds of relationships if the appropriate types do not exist. In any case, after a change is considered it would be nice to have a description on the webpage properties in the topic schema. On Tue, Sep 29, 2009 at 2:08 PM, Jeff Prucher wrote: > I'm retitleing this thread so that people who have been ignoring the > "ids" thread might notice it. Since we don't like to make changes to > /common/topic too cavalierly, even non-breaking changes, I want to make > sure anyone else who might be interested sees this. To recap: I'm > proposing adding a new property (expected type of /common/webpage) for > official websites. This would also users to make a query-able assertion > about the relationship of the official website to a topic. (More details > downthread, below.) > Jeff > > ------------------------------ > *From:* developers-bounces at freebase.com [mailto: > developers-bounces at freebase.com] *On Behalf Of *Ed Laurent > *Sent:* Tuesday, September 29, 2009 10:57 AM > *To:* For discussions about MQL,Freebase API and apps built on Freebase > *Subject:* Re: [Developers] Introducing ids.freebaseapps.com > > +1 > > -Ed > > On Tue, Sep 29, 2009 at 1:52 PM, Spencer Kelly wrote: > >> +1 >> fanwebsites and other forms of websites can be handled off-common/topic. >> cheers;) >> >> On Mon, Sep 28, 2009 at 4:56 PM, Jeff Prucher wrote: >> >>> I have reservations about typing legal entities (people, companies, >>> organizations, etc.) as "Official Websites" -- these things are not websites >>> (official or otherwise), they have websites. Would adding an "official >>> website" property to /common/topic suffice? (This would make it one of the >>> default drop-down choices in the type column of the "Weblinks" box.) That >>> would allow for semantic queries about "what is this topic's website" >>> without requiring a new cotype. Common seems like a sensible place to have >>> this, since pretty much anything (including books and movies) can have an >>> official website. >>> >>> Jeff >>> >>> ------------------------------ >>> *From:* developers-bounces at freebase.com [mailto: >>> developers-bounces at freebase.com] *On Behalf Of *Spencer Kelly >>> *Sent:* Thursday, September 24, 2009 11:29 AM >>> *To:* For discussions about MQL,Freebase API and apps built on Freebase >>> *Subject:* Re: [Developers] Introducing ids.freebaseapps.com >>> >>> >>> One suggestion would be to have a 'web domain' namespace and basically >>>> link '/en/freebase' with -> "freebase.com" and have a uri_template of >>>> "http://{key}/" and then you can make assertions about the web site >>>> that >>>> contains certain identifiers. >>>> >>>> >>> +1. this is a modelling problem. I've never been confident with how we >>> model websites. and the task of connecting a billion domains to a billion >>> topics seems like something we'd have accomplished already. >>> /common/topic/webpage can't do it (too vague) >>> and the 'website' type is broken, abandoned ( look at this discussion >>> !) >>> >>> a type with a uri property should do it shouldn't it? how bout this? >>> >>> on this note, i dont think non-admins have write access to uri_template. >>> could this be changed? >>> >>> >>> _______________________________________________ >>> Developers mailing list >>> Developers at freebase.com >>> http://lists.freebase.com/mailman/listinfo/developers >>> >>> >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers >> >> > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090930/ef98b209/attachment.htm From spatial.db at gmail.com Wed Sep 30 17:22:03 2009 From: spatial.db at gmail.com (Ed Laurent) Date: Wed, 30 Sep 2009 13:22:03 -0400 Subject: [Developers] New "official website" property on /common/topic In-Reply-To: <554723b0909300922w220d236axbb67fc2cb9a83054@mail.gmail.com> References: <4ABAA72C.4070600@metaweb.com> <4ABAECB3.1030100@gmail.com> <4ABB8392.4080808@metaweb.com> <554723b0909300922w220d236axbb67fc2cb9a83054@mail.gmail.com> Message-ID: I'm retracting my +1 in agreement with Jack and also because I didn't notice that this initial suggestion was to modify common/topic, not expand on website. For what it's worth, we've toyed around with differentiating types of websites on database and ornithological society: http://www.freebase.com/type/schema/base/database/database http://www.freebase.com/type/schema/base/conservationaction/ornithology There are so many types of websites and reasons to link them to topics that is makes sense to add a mediator property to some form of website CVT or to expand on the current weblink CVT. -Ed On Wed, Sep 30, 2009 at 12:22 PM, Jack Alves wrote: > Intended usage of /common/topic/webpage property has always been fuzzy. My > understanding is that the existing /common/topic/webpage property is > intended for "official" websites. I think it would be great to have a > generic property for webpage relationships to topics. With the proposed > change, it sounds like the existing topic/webpage could be used for any > related webpage. > > I think it would be more useful to keep the existing webpage property for > official websites and create a more versatile mediator property for generic > webpage relationships. Since the web is so loosely structured, additional > information about the relatedness would be useful. For instance, the link > may be to a news article or blog post that "describes", "reviews" or > "mentions" the topic. The link may be to a page that represents the topic in > a positive, negative, or neutral way. The link may be strongly or weakly > related. > > This may be over anticipating usage but the point is to consider how a > change would be helpful beyond a specific case. Changes to common/topic are > rare. This is an opportunity for people to suggest practical usage for a new > property. At some point, someone will build a bookmarking tool to connect > topics to related websites. This is a great time to consider what that tool > might need to advance tagging functionality while keeping the barrier to > contributions low. Okay, okay, I understand the ultimate objective is to > create new types with properties for different kinds of relationships. I > just think it would be useful to lower the barrier for creating new kinds of > relationships if the appropriate types do not exist. > > In any case, after a change is considered it would be nice to have a > description on the webpage properties in the topic schema. > > > > > On Tue, Sep 29, 2009 at 2:08 PM, Jeff Prucher wrote: > >> I'm retitleing this thread so that people who have been ignoring the >> "ids" thread might notice it. Since we don't like to make changes to >> /common/topic too cavalierly, even non-breaking changes, I want to make >> sure anyone else who might be interested sees this. To recap: I'm >> proposing adding a new property (expected type of /common/webpage) for >> official websites. This would also users to make a query-able assertion >> about the relationship of the official website to a topic. (More details >> downthread, below.) >> Jeff >> >> ------------------------------ >> *From:* developers-bounces at freebase.com [mailto: >> developers-bounces at freebase.com] *On Behalf Of *Ed Laurent >> *Sent:* Tuesday, September 29, 2009 10:57 AM >> *To:* For discussions about MQL,Freebase API and apps built on Freebase >> *Subject:* Re: [Developers] Introducing ids.freebaseapps.com >> >> +1 >> >> -Ed >> >> On Tue, Sep 29, 2009 at 1:52 PM, Spencer Kelly wrote: >> >>> +1 >>> fanwebsites and other forms of websites can be handled off-common/topic. >>> cheers;) >>> >>> On Mon, Sep 28, 2009 at 4:56 PM, Jeff Prucher wrote: >>> >>>> I have reservations about typing legal entities (people, companies, >>>> organizations, etc.) as "Official Websites" -- these things are not websites >>>> (official or otherwise), they have websites. Would adding an "official >>>> website" property to /common/topic suffice? (This would make it one of the >>>> default drop-down choices in the type column of the "Weblinks" box.) That >>>> would allow for semantic queries about "what is this topic's website" >>>> without requiring a new cotype. Common seems like a sensible place to have >>>> this, since pretty much anything (including books and movies) can have an >>>> official website. >>>> >>>> Jeff >>>> >>>> ------------------------------ >>>> *From:* developers-bounces at freebase.com [mailto: >>>> developers-bounces at freebase.com] *On Behalf Of *Spencer Kelly >>>> *Sent:* Thursday, September 24, 2009 11:29 AM >>>> *To:* For discussions about MQL,Freebase API and apps built on Freebase >>>> *Subject:* Re: [Developers] Introducing ids.freebaseapps.com >>>> >>>> >>>> One suggestion would be to have a 'web domain' namespace and basically >>>>> link '/en/freebase' with -> "freebase.com" and have a uri_template of >>>>> "http://{key}/" and then you can make assertions about the web site >>>>> that >>>>> contains certain identifiers. >>>>> >>>>> >>>> +1. this is a modelling problem. I've never been confident with how we >>>> model websites. and the task of connecting a billion domains to a billion >>>> topics seems like something we'd have accomplished already. >>>> /common/topic/webpage can't do it (too vague) >>>> and the 'website' type is broken, abandoned ( look at this discussion >>>> !) >>>> >>>> a type with a uri property should do it shouldn't it? how bout this? >>>> >>>> on this note, i dont think non-admins have write access to uri_template. >>>> could this be changed? >>>> >>>> >>>> _______________________________________________ >>>> Developers mailing list >>>> Developers at freebase.com >>>> http://lists.freebase.com/mailman/listinfo/developers >>>> >>>> >>> >>> _______________________________________________ >>> Developers mailing list >>> Developers at freebase.com >>> http://lists.freebase.com/mailman/listinfo/developers >>> >>> >> >> _______________________________________________ >> Developers mailing list >> Developers at freebase.com >> http://lists.freebase.com/mailman/listinfo/developers >> >> > > _______________________________________________ > Developers mailing list > Developers at freebase.com > http://lists.freebase.com/mailman/listinfo/developers > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090930/8e5c72e5/attachment-0001.htm From spencerkelly86 at gmail.com Wed Sep 30 20:38:13 2009 From: spencerkelly86 at gmail.com (Spencer Kelly) Date: Wed, 30 Sep 2009 16:38:13 -0400 Subject: [Developers] New "official website" property on /common/topic In-Reply-To: References: <4ABAA72C.4070600@metaweb.com> <4ABB8392.4080808@metaweb.com> <554723b0909300922w220d236axbb67fc2cb9a83054@mail.gmail.com> Message-ID: i support the proposed change. by 'official', one i think, means 'operates the domain name', and something that represents identity (deserves a 'owl:sameas'). whether or not anything can be called official on the internet is one question, but ALL other ontologies I have found use for such a qualifier- foaf uses for people, musicbrainz uses and on and on. i say a domain name is a necessary condition of being an official website. blogs and things don't count - so far, these are all being modeled differently, which is probably best. check out twitter user, myspace user , event feed operator, mailing list subject etc maybe we need 'tourism website', and 'government website' somewhere.. On Wed, Sep 30, 2009 at 12:22 PM, Jack Alves wrote: > >> For instance, the link may be to a news article or blog post that >> "describes", "reviews" or "mentions" the topic. The link may be to a page >> that represents the topic in a positive, negative, or neutral way. The link >> may be strongly or weakly related. > > i've had a go at this here at expressions of value there are alot of ways topics relate to uri's, but 'official site' is a valid idea with strong enough boundaries to have success across all topics. dont let the fringe cases frazzle you. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.freebase.com/pipermail/developers/attachments/20090930/e43eb295/attachment.htm