Is declaring correct deps/conflicts for versions not in stable really no longer needed?

["Nikita V. Youshchenko" <yoush@cs.msu.su>]

> > > > >  * Finally drop the bogus libapache-mod-ssl dependency from the
> > > > > apache1.3 php4 module, as glibc (>= 2.3.2.ds1-17) has fixed the
> > > > > dlopen refcount bug that we were hacking around (closes: #205553,
> > > > > #230956, #271000)
> > > >
> > > > However, libapache-mod-php4 cdoes not depend on libc6 (>=
> > > > 2.3.2.ds1-17), it just depends on libc6 (>= 2.3.2.ds1-4). Isn't
> > > > that a  bug?
> > >
> > > Perhaps, but this lack of precision will have no effect on upgrades
> > > from  woody (glibc -17 -- or rather, 18 -- is targetted for sarge);
> > > it will only have an effect for partial upgrades from earlier
> > > versions of testing/unstable.
> >
> > Quite a few people do use testing today, and may be affected by this.
>
> Perhaps so, but if we went out of our way to depend/conflict against
> every broken package that's been in testing/unstable, package
> relationships would be ridiculousy unwieldly.  The very reason
> testing/unstable exists is to weed out the bugs the best we can before a
> stable release.

I believe you are wrong.
If package requires a version of another package greater than X, it should 
declare so in it's dependences, even if you are sure that next stable 
release will satisfy this automatically.
Current situation is version of libapache-mod-php4 currently in testing 
does not work correctly with version of libc6 currently in testing.

I'm forwarding this to -devel.