Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!seismo!gatech!ulysses!dgk From: dgk@ulysses.homer.nj.att.com (David Korn[eww]) Newsgroups: comp.unix.wizards Subject: Re: /usr/tmp Message-ID: <2707@ulysses.homer.nj.att.com> Date: Mon, 6-Jul-87 00:46:10 EDT Article-I.D.: ulysses.2707 Posted: Mon Jul 6 00:46:10 1987 Date-Received: Mon, 6-Jul-87 06:46:04 EDT References: <431@murphy.UUCP> <6064@brl-smoke.ARPA> Organization: AT&T Bell Laboratories, Murray Hill Lines: 26 Summary: Another possible way to distinguish /tmp from /usr/tmp In article <6064@brl-smoke.ARPA>, gwyn@brl-smoke.UUCP writes: > In article <431@murphy.UUCP> dave@murphy.UUCP (The Giant Hogweed) writes: > >Does anyone still need /usr/tmp? > With a system that does distributed computing you might want to consider using separate /tmp on each host and a single shared /usr/tmp. For example, the script foo > /usr/tmp/file$$ bar /usr/tmp/file$$ would be guaranteed to work even if foo and bar execute on separate nodes but foo > /tmp/file$$ bar /tmp/file$$ would not since /tmp is allowed to be different on different nodes. Of course you need a shell that generates a unique $$ over the network. David Korn {ihnp4|attunix|seismo}!ulysses!dgk