[lug] [OT] (X)emacs mode for xslt?

D. Stimits stimits at attbi.com
Wed Mar 12 13:25:02 MST 2003


rm at fabula.de wrote:

> On Wed, Mar 12, 2003 at 01:15:03PM -0700, D. Stimits wrote:
>
> >...
> >
> >
> >>
> >>Well, my customer switched from a static-render model (a _very_ 
> expensive
> >>"commercial" solution) over to live rendering -- it makes the workflow
> >>so much easier for all involved parties. Rendering of a weeks edition
> >>took several hours (on a Sun Enterprise box) ....
> >>Keeping a single "master version" of an article was one of the goals
> >>for the new system (sometimes the editor in charge _has_ to be shure
> >>that an article got changed everywhere on the website :-/
> >>
> >
> >What would be ideal might be a cache system (not in the sense of browser
> >based cache) where the server renders the static page the first time it
> >is called (to a private file), creating the content once; then if the
> >static page currently in residence on the server is up to date at the
> >moment of a page hit, it would simply cat the static page to the hit
> >request. If the page is altered only once per week, it would seem
> >efficient to not generate it every single hit. Something like a time
> >stamp change or MD5 change on the XML side could be used as a cache
> >regeneration trigger.
>
>
> A good web cache like Squid will do that -- as long as you configure
> your server to send correct modified-since answers ...
> I toyed with such a solution for a while (with a modified squid that
> got recache instructions every time a resource was modified in the
> content management system) but after seeing the performance we get
> with the current system it wasn't worth the time (and extra administrative
> complication that such a setup would generate - more components means
> more possible points of failure). Our limiting factor seems to be the
> number of parallel processes (and resources like file descriptors) --
> i had to patch the kernel to fix that.
>
>   Ralf


Were you running enough hits that you were finding your max file 
descriptor limit?

D. Stimits, stimits AT attbi DOT com




More information about the LUG mailing list