[Developers] sandbox caching old values, slow expiration?

[Coral Link]

Thanks for the reply Kirrily. Makes sense.  I may incorporate some of this
into my code as a temporary work around.

-Indy

On Dec 29, 2007 6:43 AM, Kirrily Robert <skud at infotrope.net> wrote:

> On Dec 30, 2007 12:51 AM, Coral Link <corallink at gmail.com> wrote:
> > Hi, quite a few times while working on my app I've run into instances
> when
> > I've made updates/changes to objects in the sandbox BUT when I query the
> > just changed object, I get its old values.  I  run my queries in
> > queryeditor, and the values returned are the updated ones. After some
> > magical period, the mqlwrite/mqlread service starts showing the expected
> > values.
> >
> > Is there some sort of caching going on?  What are the timeouts if any,
> to
> > pull the most current values from the repository?  Is there some sort of
> > tickle mechanism to ask mqlread to pull a fresh copy instead of a cached
> on?
>
> Sounds like the same thing I discovered a few weeks back.  The (short)
> thread about it can be found at
> http://lists.freebase.com/pipermail/developers/2007-November/000932.html,
> but in short, "that's not a bug, it's a feature!"
>
> Here's some quoted text from the reply I got:
>
> > Its part of our caching strategy - the idea is that if a client hasn't
> > done a write, then it doesn't need to see the most updated result of a
> > query. so I'm guessing you're writing to www.freebase.com via a browser,
> > but you have another script that is doing the read?
> >
> > There is actually a 2nd cookie called "mwLastWriteTime" that stores a
> > token that tells us how current your data is. If present, it tells us
> > how current your results should be.. if NOT present, it means "I don't
> > care how current my data is" and may give you even older data.
> >
> > If you really, really, need to refresh this cookie to get new data, you
> > should do an empty HTTP POST to /api/service/mqlwrite. I encourage you
> > to do this sparingly - this is not a generally supported mechanism yet,
> > but it should continue to work for the near future.
>
> I wasn't using sandbox when I encountered the problem, but it does
> sound like a similar sort of thing.
>
> K.
>
> --
> Kirrily Robert
> skud at infotrope.net
> http://infotrope.net
> _______________________________________________
> 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/20071229/ad9b797d/attachment.htm