[Catalyst] OT: Windows bashing (was: Catalyst for (really) bigapplications)

[Peter Edwards]

>    Windows (n.): 32-bit extensions to a 16-bit graphical shell
>    for an 8-bit operating system originally coded for a 4-bit
>    microprocessor by a 2-bit company that can't stand 1 bit of
>    competition.

What did I receive in the post from M$ this week?

Answer: a small cardboard box containing a stone.
Underneath the stone a note saying "there's nowhere to hide".

Reading the attached leaflet, it transpires that Micro$oft's latest stroke
of marketing genius is to send this to all registered Micro$oft Partners
along with a plea to turn in anyone we know using an unlicensed copy of M$
software, and at the same time to issue a not very well veiled threat saying
they would be visiting vendors in <mailmerge field town> area real soon now
to check we have our licensing correct.

This attitude seems to date back to when Bill Gates got pissed off to find
that hobbyists were duplicating the punched tape spool his first BASIC came
on. I've always wondered if, when he wrote DOS and borrowed the Unix file
path, he changed / to \, that was to make it proprietary? LOL.

Another answer to the original query about mod_perl memory sizes.
Work out mean transaction processing time and maximum mean hits over that
time period to calculate how many Apache child processes you need to serve
the hits without a queue. 10-15 is usually enough for 150 users.
Run an exercising tool against your mod_perl app then see what the in-memory
process sizes get up to (on Linux run "top" and press M).
Real world example, I have a 250,000 line application that caches lookup
data (memory's cheaper than disk accesses) and it's unlikely to get higher
than 40MB per process. I have Apache set to restart a child after 100 runs
just in case but it doesn't seem to leak memory anyway.
A modern server with 1GB+ memory should be more than adequate.

Regards, Peter

"Where do you want to hide today?"