Current squid issues (ctd)

Internals are still trying to milk performance out of older machines
Trying to buffer data due to small socket buffers
Trying to do limit disk IO due to slower CPUs/busses

Can't take advantage of newer APIs
win32 completion ports
FreeBSD kqueue, Solaris /dev/poll
POSIX AIO

WCCP1 support has been reported to not work in a lot of cases
Log rotations cause squid to temporarily stop accepting requests
Larger caches can pause up to a minute whilst writing logs
Reported to cause WCCP1 to break