*BSD News Article 21112


Return to BSD News archive

Newsgroups: comp.os.386bsd.bugs
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!howland.reston.ans.net!europa.eng.gtefsd.com!uunet!cs.utexas.edu!swrinde!menudo.uh.edu!uuneo!sugar!taronga!bonkers!peter
From: peter@taronga.com (Peter da Silva)
Subject: Re: bug with ufs file creation
Organization: Taronga Park BBS
References: <CD44wx.LHs@rex.uokhsc.edu> <27202e$7u@umd5.umd.edu> <CDD909.603@taronga.com> <277hfr$ka1@umd5.umd.edu>
Message-ID: <CDG6I7.rA@taronga.com>
Date: Thu, 16 Sep 1993 12:53:52 GMT
Lines: 28

In article <277hfr$ka1@umd5.umd.edu>,
Mark Sienkiewicz <mark@roissy.umd.edu> wrote:
>I didn't say you can't do it.  I said nobody does it.

Sorry, looked like you were saying it didn't work. My mistake.

>Of course, I'm
>being overly general, because some programs actually make use of this
>feature, but lots of them do not.

Any program that calls fopen(..., "a") does, unless your stdio is broken.
The shell should on '>>', and anything that opens a log file should. I
certainly use O_APPEND in these circumstances.

>If you grant/deny access based on
>O_APPEND, I expect you will violate the "principle of least astonishment"
>regularly.

Let's see... we could spend the next week going back and forth on this,
but I'm not up to it. I don't agree that this would be at all astonishing,
and I'd like to see an "extend" permission added in the next turn of the
wheel. But it'd really take an actual implementation to see how it looks
in normal use, so a bunch of discussion about hypotheticals isn't going to
do much good.

And, of course, the principle of least astonishment has already been
thoroughly broken. NFS breaks it regularly, cheerfully, and even brags
about it.