httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bill Stoddard" <>
Subject Re: 2.0.24 tagged.
Date Thu, 16 Aug 2001 18:12:08 GMT

> -1 on rolling until mod_mime is fixed.
> I pointed out on a thread earlier that the hash _entities_ are still broken,
> although the hash _table_ works with Greg's patch.
> Please review the 'optimizing' musings.  If this is acceptable to everyone, I'll
> apply the patch [still waiting for Greg's test of that patch on daledeus], if it
> is less than acceptable, Mr. Stoddard can begin reverting the tables/hash patch
> that broke us in the first place.  Unless it's fixed _and_ optimized, this is a
> standing veto of the hash patch for mod_mime.  The overhead of the tables is
> minimal if the hash is only walked a handful of times per request (or subreq.)

I'm a bit behind in my mailing list reading so perhaps this has been discussed but I'll
bring it up again anyway...

Where do we spend cycles in mod_mime today (in httpd-1.3)? In merge or in lookups? Using
the hash table will clearly improve look-up times.  One of the major reasons we do merges
is to handle .htaccess files. If we have .htaccess files, I would venture that we spend
MUCH more time doing file i/o than we would spend in doing hash table merges. In other
words, improving the merge performance -may- not significantly improve overall performance
in the .htaccess file case.


View raw message