Fwd: Some inconsistencies in the search algorithm

[Nicolae Florin Petrovici]

Hi Terry,

Sorry to bother you but I think that after the update you made yesterday some things are broken.

Search isn't working right. 

For instance I have the object: aebfc4a8-c9c4-44da-ae1a-c4b86ec64b8b

It has the following: 

nfpetrovici/tags/TagVector copyrightlaw100falkvinge67pirateparty57performances56majorparts56

When I issue a query: 

nfpetrovici/tags/TagVector matches "*law*"

I get: 

97a08bc4-15e2-4953-adcb-972b5fdc90dd 

AND

35311060-aae4-4a64-94e7-62c644548670

But I do not get the aebfc4a8-c9c4-44da-ae1a-c4b86ec64b8b object.

Why is that? 

I can give you another example:

nfpetrovici/tags/TagVector matches "*kernel*"

It gets me the following object ids:

247b75e4-6959-4e39-a62f-50571490906d

e1c06b8e-46b3-4f9f-b14d-6dbe1525301e

But I do not get the 851e17c5-6731-4d39-8203-6ff271f33644 object id which has the following in the TagVector: kernelversions95longtermkernels93kernelrelease82longtermkerneltrees71differentactivekernel60

We think it is from yesterday's update as this worked earlier. 

Please also tell us more on what you are planning to do, as our entire backend is based on FluidInfo and we want to release this site to the public around March this year.

We expect to have many users and want to have the FluidInfo backend as consistent and fast as possible.

Thanks,

Nicolae

--
Thank you,
Nicolae Florin Petrovici
+40722697347

[Terry Jones]

Hi Nicolae

Thanks for the mail. I'm sorry about the problems. We noticed the
inconsistency this morning too. Several of the guys had a Skype about it
and identified an incorrect setting that was causing some tag values to not
be indexed. We're now running a full re-index of all tag values (there are
about 150 million). That's unfortunately quite slow, we don't expect it to
be finished for a few hours. We only need to do that very occasionally, so
this is quite exceptional - not normal behavior! As the re-indexing runs,
data will gradually become visible. We'll send an update when it's done.

BTW, for real-time updates on what's going on, the IRC channel (#fluidinfo)
is a good place to be.

> We think it is from yesterday's update as this worked earlier. Please
> also tell us more on what you are planning to do, as our entire backend
> is based on FluidInfo and we want to release this site to the public
> around March this year. We expect to have many users and want to have
> the FluidInfo backend as consistent and fast as possible.

Understood! We're also completely dependent on Fluidinfo being up, stable,
fast, etc., so we take all this seriously. The new http://fluidinfo.com
interface is just a regular Fluidinfo client, using the API like any other
application. Thanks for using Fluidinfo and please keep on mailing the list
with experiences / comments (good & bad) and we'll do whatever we can as
fast as we can to support you.

Terry