[Koha-patches] [PATCH] cache the session for the duration of the script

[Michael Hafen]

I've finished my investigation.  I compared entries in the sessionlog of
logged out for inactivity to the apache access log.  I didn't see any
premature session timeout's, but that only covers one day.

Preliminary conclusion is that this patch solves this problem.

I don't mind if you want to hold this patch till after release and roll
it into the 3.4 alpha.  I can run the patch locally just fine either
way.  My code isn't up-to-date with 3.2 yet anyway, that's my project
for this summer :)

On Mon, 2010-05-10 at 09:53 -0400, Galen Charlton wrote:
> Hi,
> 
> On Mon, May 10, 2010 at 9:47 AM, Michael Hafen <mdhafen at tech.washk12.org> wrote:
> > I don't have any profile data, it's just a supposition.  I figure it
> > will save however many milliseconds it takes to parse the session data
> > again to form the object.
> 
> My concern with that is that a micro-improvement to performance can
> cause an unexpected breakage in semantics by adding an extra layer of
> caching.  Some of this is a theoretical concern - for example, in a
> mod_perl setup, a package-level cache of the session would break
> per-user session management badly.  Of course, since we don't run Koha
> using full mod_perl, that isn't a direct issue.  But this close to
> release, I'm reluctant to add package-level caching of anything absent
> a significant performance boost.
> 
> Of course, if you track down your premature session timeouts issue,
> that could be a different story.
> 
> Regards,
> 
> Galen

-- 
Michael Hafen
Systems Analyst and Programmer
Washington County School District
Utah, USA

for Koha checkout
http://development.washk12.org/gitweb/
or
git://development.washk12.org/koha