System mailboxes corruption

[Steve Platt]

Eric - exactly what sort of rubbish did you find?  If it was really null
characters, sounds like one of the NFS "features", I think that NFS
clients "stat"ing a remote file while it's being written can cause whole
blocks to get zero-ed out ... however:-

We had some trouble a while back when our 3.5 mail server got its root
filesystem full and /pricate/usr/spool/mail is in there of course ...

We got users invoking mail, ignoring a message like "/: filesystem full"
and getting garbage header lines printed, so they "quit" from mail (silly
people!); thus writing that junk back into their mailboxes - the garbage
was often a repeating 4byte pattern (maybe a 32 bit number) with sometimes
a solitary "Status: O" in the middle.  Of course, Mail didn't think much
of that second tiome around - sort of shocked me that Mail didn't handle
the lack of space (for /tmp/files?) better.

By the way, our full / filesystem was due to cron dying in the middle of
my fortnight away - ergo big log files!

Also, the number of sunos utilities that did silly things when / filled
was amazing - I take it real programmers don't check the return value from
write(2)!