disk quotas

BostonU SysMgr root%bostonu.csnet at csnet-relay.arpa
Sat Aug 3 10:17:18 AEST 1985


Having started this flamage I am really quite surprised at the varied
opinions pro and con. I always assumed disk quotas weren't in V6 cuz it was
intended for a clubhouse environment anyhow and slowly as sites got online
with V7 etc they added them locally (Harvard I remember did this.) Finally,
UCB added a good implementation to their distribution.

I assumed SYSV didn't have it either because no one had bothered or no one
had screamed (probably both.) So, I screamed I guess.

Surprise, now people are claiming it's a philosophical issue (that's not a
bug...) I guess it ain't obvious.

Granted, it's worse in a student environment because so many people are
learning to play with powerful software constructs (like creat()? well,
yes.) Ok, people make mistakes (try "tar -f foo ." sometime on a 4.2
sys, maybe others I dunno.)  These days, commercial environments are
also faced with many non-technical users.

Ulimit solves part of the problem but the extension to quotas is not very
difficult, why not just solve the whole problem?  To me Ulimit is a very
non-unix'ish special case until someone thinks of a general tool, which is
a quota system.  Also, people can be lazy (I am carefully avoiding the
'malicious user' issue.)  As far as the sum of the quotas being larger than
the disk, that's probably ok, its a minority you are generally trying to
control, we do it all the time, never had a problem.

Yes, we have really really come down because of user accidents on quota-less
systems, its embarrasing and the reaction usually is 'why didnt it stop me?'

Unfortunately, disk is one of those resources that gets used up very
passively (as opposed to CPU cycles or terminal lines.) It just accretes
over time. On our research system we have no quotas, on our student system
we have fairly liberal quotas (I don't think a student has asked for more
yet, our UNIX sys is just over a year old for students.) The only place we
have had troubles is our research system and the person always apologized
and was usually unaware they had 'that much junk' [I send out mail messages
occasionally to the top several users who often account for 50% or more.]

Nope, no compromise on this one. I think its wrong that SYSV does not offer
a disk quota system and it remains in my list of negative features when
asked for a comparative evaluation from various people. I just don't like
being woken up by a phone call from operations that a disk is full and the
system is screaming (sometimes its not that trivial to fix.)  I have never
had any reaction but outrage over the lack of it from people listening to
the comparison (that is, the people about to buy the system.)  No, I am not
hardnose about control, as I said, we run large systems without quotas as a
matter of, not lack of, choice. If we go to quotas on those systems I will
probably let people change their own quotas.

	-Barry Shein, Boston University

"A toy shouldn't break just because a child plays with it" -- TONKA



More information about the Comp.unix.wizards mailing list