Operation Questions

Just a couple of quick questions about Squid's behaviour:

1) Say I have 128M of RAM, I believe the recommended setup is to
configure Squid to use about 1/3 of this for it's primary memory pool -
does it's memory usage grow and grow beyond physical RAM and then into
swap (if needed) or does it just limit itself somehow? What would be
the ideal memory config parameters if using 128MB?

2) If I have the choice of configuring several disk partitions as either
a single multi-device as in Linux's /dev/md or leave them single, which
would give me better performance (if any difference)? If I configured
several partitions using RAID-0 under MD, would this be better than
telling squid to use several disk partitions?

3) As the cache becomes full, does Squid automatically begin to purge
the oldest data kept, or what stops it from overflowing?

4) If I ever wanted to - what's the best way to purge the whole cache or
clean it up; is there a special Squid function or is it just done by
nuking the cache dir?

-- 
Tim Carlier                            Tim.Carlier@OZsupermalls.com.au