[Data-modeling] "Mentions" and "Mentioned In"

Kirrily Robert kirrily at metaweb.com
Sat Feb 14 00:16:08 UTC 2009


Spencer, you have some really interesting points here which I think  
are worth considering.  You're right that types are the least open and  
wiki-like things on Freebase.  Historically this came about because we  
found that, quite frankly, most people don't get it and can make an  
enormous mess which, apart from breaking apps and saved views, can be  
extremely difficult to clean up -- Scott alluded to this.

I'm the sort of person who understands data reasonably well (or at  
least I like to tell myself I am), and I expect everyone on this list  
is likewise.  We don't have too much trouble imagining, for instance,  
many-to-one relationships or class inheritance.  But we're a tiny  
minority.  So the issue is to find ways to allow people like us to  
collaborate, while not opening ourselves up to enormous messes by  
people who don't get it (or even DoS attacks by those who get it and  
don't like us.)

In the open source software world, the equivalent problem exists: they  
want everyone to be able to see the code and edit it, but not  
everyone's edits should be folded into the main branch.  The solution  
there is forking, and I think that's what we're starting to see on  
Freebase with property delegation and the like.  I wouldn't be  
surprised, given what's happened with Acre and other things lately, if  
one day we even had a "Clone" button on types or even whole bases,  
allowing people to fork and edit at will.

That'd be cool, but the problem is how to fold those forks back into  
the main branch.  We need experienced data modelers to do this, and  
we'd love to have tools to make it more effective.  Currently people  
suggest properties, there's some discussion, then the admins implement  
it.  Wouldn't it be cool if you could add properties provisionally,  
and if they are accepted, it's just a button-click to make them  
"official"?

All these things are things we're considering and which get brought up  
from time to time, and I think the discussions about it are getting  
more frequent lately.  Please keep giving us your feedback and ideas  
about how we could make it work!

K.


-- 
Kirrily Robert
Freebase Community Director
kirrily at metaweb.com
http://freebase.com/






More information about the Data-modeling mailing list