Re: SPF - exim4 + debian.org
On Mon, Jul 26, 2004 at 09:06:37AM -0400, Stephen Frost wrote:
> > Why should anyone assume there's a connection between packages installed
> > and usernames used? Indeed, if more than few packages did this, it
> > would utterly unworkable, because package names are often very short and
> > quite likely to intersect with usernames.
> Well, gee, how about because, in general, the username is the same as
> the package name, and it makes it nice and clear what username is
> associated w/ what package? Looking at one of my systems I see:
> daemon, bin, sys, sync, games, man, lp, mail, news, uucp, proxy,
> majordom, postgres, www-data, backup, msql, operator, list, irc, gnats,
> mysql, sshd, identd, smmsp
I guess in some cases such user names are historically so entrenched that
they must be retained, but I think a policy that helps to avoid future
conflicts would be a good thing.
Anyway, names like "deb-foo" or "_foo" are also pretty easy to associate
with package name "foo".
Actually I think alternative names could be even _clearer_ than the
current-style unadorned names -- e.g., seeing a username like "sys-foo" or
"pkg-foo" immediately gives you a hint that there's something special about
> I'm very happy to note that *none* of them are 'deb-<blah>' or
> 'Debian-<blah>', and you know what? I've never had any problems w/
> namespace pollution. Have any of our users actually filed bugs about
> problems because of these names?
Are you serious? Of course most home users don't have too many problems
because they've only got a few usernames themselves. The problem (as was
mentioned) is large installations that have a huge base of existing usernames
which must be shared between disparate system types, and who can't easily
change real usernames.
> It seems to me that a problem is being invented so that it can be
> 'fixed'. The problem doesn't exist and 'fixing' it just makes things
> ugly for no reason.
No, there's certainly a problem. The _extent_ of the problem is certainly
open to discussion, and it's certainly possible that debian could decide that
pretty system user names are very important, and "what the hell, screw those
large institutional users".
Fast, small, soon; pick any 2.