Date: Mon, 26 Feb 2001 14:59:43 -0800 (PST) From: Matt Dillon <dillon@earth.backplane.com> To: Kris Kennaway <kris@obsecurity.org> Cc: "Andrey A. Chernov" <ache@nagual.pp.ru>, "Jacques A. Vidrine" <n@nectar.com>, arch@FreeBSD.ORG Subject: Re: rand.c patch for review (was: Re: cvs commit: ports/astro/xglobe/files patch-random) Message-ID: <200102262259.f1QMxhP46130@earth.backplane.com> References: <20010225132152.A39554@mollari.cthul.hu> <20010226005004.B59772@nagual.pp.ru> <20010225135429.A47615@mollari.cthul.hu> <20010226020827.A61007@nagual.pp.ru> <20010225151519.A63582@mollari.cthul.hu> <20010226022902.A61216@nagual.pp.ru> <20010225173445.A37510@spawn.nectar.com> <20010226024456.A61566@nagual.pp.ru> <20010226174852.B435@ringworld.oblivion.bg> <20010226191008.A72434@nagual.pp.ru> <20010226142211.A20768@mollari.cthul.hu>
next in thread | previous in thread | raw e-mail | index | archive | help
:> random() (which is NOT original code). Practice says that nobody seriously
:> want old behaviour. Moreover, standards (like SUSv2 f.e.) don't guarantee
:> that the formula must be the same on different machines or even through
:> different libc versions. It must be the same for single application run
:> only, so in theory we can change formula for each application run :-)
:>=20
:> So, I plan to skip this step as superfuous.
You can't change the formula for each application run. One of the
biggest reasons one specifies a seed is to ensure that the same
random sequence will be produced across multiple program runs.
Otherwise debugging can very difficult to impossible. Calling
srandomdev() for random/srandom is a sufficient statement of intent
to not want a reproducable sequence. calling srandom() is a sufficient
statement of intent that you want a reproducable sequence. For
rand() you don't have a choice. You only have srand() to call (and
get a fixed seed of 1 if you don't). The specs imply that if you will
get the same random sequence given the same seed. While we can change
the algorithm between releases without creating too much dissention,
we sure as hell can't change the algorithm for every process.
The way I see it, we can only protect programmers against their own
stupidity to a certain degree. We should not be wasting resources
trying to make rand() perfect. It just needs to be able to do its
jobs fairly well within the bounds of the API.
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200102262259.f1QMxhP46130>