On Fri, 5 Mar 2010 18:21:21 +1300
Sess <[email protected]> wrote:
> O.K.; let's put it like this. I'm a post-graduate information
> retrieval researcher. The open-source C++ Wikipedia indexer (ANT) we
> leverage for ranked retrieval over all roughly three million articles
> of the English Wikipedia corpus consumes about 10ms per average query
> consisting of 2.3 words.
> 
> 10ms. Three million articles. Now, Inquisitio isn't indexing
> article-length text. It also isn't indexing three million exheres.
> (Not on my machine, anyway!) Can't we do better? And if not, why not?
> It's slowness really is a bit of a pain.

Because we don't pregenerate an index. We don't even pregenerate a
metadata cache yet.

If you've got metadata pregenerated, inquisitio takes under a minute to
run, and it takes you much longer than that to read its output, visit
package homepages to see if they're what you want and so on.

We *could* generate an index, but then you'd have to keep it up to date
after every sync, which is sloooooooow. Or we could generate it
centrally, but then we'd run into the same problems that eix has: what
we have centrally isn't the same as what's on your local box.

-- 
Ciaran McCreesh

Attachment: signature.asc
Description: PGP signature

_______________________________________________
Exherbo-dev mailing list
[email protected]
http://lists.exherbo.org/mailman/listinfo/exherbo-dev

Reply via email to