On Thursday 05 January 2006 11:25, you wrote: > On Thu, 5 Jan 2006, cobaco (aka Bart Cornelis) wrote: > > What Side affects? > As a user, I would become very upset if installing a package would > alter my $HOME/.profile. Whatever I do in my startup scripts > is not a business of any package, it's my business as a user.
So you are seriously telling me that you believe that a user of user-es,
user-euro-es, or user-de, all packages that are effictively mini-cdd's
whose explicit goal it is to change the environment over to spansish or
german would get upset that the environment is changed to actually support
spanish or german?
For desktop-profiles and sysprofile this argument is even more bogus as the
profile.d snippets those packages add don't change the environment by
themselves, any environment changes resulting from running those snippets
are consequence of parsing bits of configuration under admin control (i.e.
when the admin hasn't changed the default configuration nothing will be
changed).
> Allowing /etc/profile.d would be the equivalent of allowing packages
> to modify the user's startup files. It would break the principle
> of least surprise, and therefore a bad thing.
Are you seriously positing that the principle of least surprise would have a
user of user-es whose package description starts with:
This package allows administrators to set the default language
(es_ES) and encoding (ISO-8859-1) used by a Debian GNU/Linux System
(and its applications) easier.
be surprised when the environment is actually changed that automatically to
do exactly that when installing the package? If so I find that a rather
insulting view of our users reading comprehension (not even intelligence).
In the case of desktop-profiles, which is a management framework that lets
the admin set up configuration sets do you really believe that the
principle of least surprise would be to not have those admin-specified
settings apply in the corner-case of logging in through 'ssh -X'? If so the
user that filed #344030 definately doesn't agree with you on that (nor do
I)
> Moreover, dpkg usually asks about changes in configuration files in /etc.
> Having a profile.d would be the equivalent of dpkg being allowed to
> change /etc/profile without the user being prompted about the change.
> The principle of least surprise would be broken again.
nope, all it does is clearly identify the parties responsible for each bit
of configuration, and making it possible to do that configuration in the
first place without messing with other packages conffiles.
Given that this exact mechanism is used by plenty (core) packages such as:
- the Xserver (/etc/X11/Xsession.d)
- lograte (/etc/logrotate.d)
- sysv-rc (/etc/rc[0-6].d)
- cron (/etc/cron.d)
- discover (/etc/discover.d)
- apache (/etc/apache/conf.d/)
-...
-> the position that this a profile.d directory would break the principle of
least suprise is obviously not realistic.
Especially since the earlier bug rapports on this issue mention that Redhat,
Mandrake, Suse and Gentoo do have a profile.d (that's pretty much every
major distro that's not Debian), so at least for users that switch (and
anecdotal evidence suggest that covers most Debian users) this would
definately not violate the 'principle of least surprise'
> So no, I do not believe in legitimate uses of profile.d.
please counter the specific examples I've given above if you still feel that
way
--
Cheers, cobaco (aka Bart Cornelis)
1. Encrypted mail preferred (GPG KeyID: 0x86624ABB)
2. Plain-text mail recommended since I move html and double
format mails to a low priority folder (they're mainly spam)
pgpOTmY2btkha.pgp
Description: PGP signature
