On Sun, Jun 18, 2000 at 07:21:22PM +0300, Jarkko Hietaniemi wrote:
> > - install a squid-accelerator in front of the webserver. The squid can
> > run on the same box and is very easy to configure and manage. (It is
> > very easy to misconfigure too, so better let me have a look on the
> > configuration some time:)
>
> I'm a little bit confused of what value would a squid cache be in this
> case. If someone plows through all the files in a Perl distribution,
> those files will most likely not be cached (it's not often someone
> asks for lib/cacheout.pl, for example), some process will still have
> to open up, gunzip and tar xf, the honking big .tar.gz, right?
> And because Apache helpfully parallelizes things by having multiple
> processes/threads, several of these big unpackings are being unleashed
> at the same time.
>
> Either I am unaware of some deep squid magic here or someone will
> still be able to bring the server to its knees with a stupid robot.
I would agree that squid may not be of use against someone plowing through
the site, but it may be of use anyway.
The content that does get generated does not change that often, so caching
the result could help a lot. It would also mean that the cache code I have
for the pods could also be removed.
Graham.
