On Wed, Oct 01, 2003 at 02:28:52PM +0200, Matthias Urlichs wrote: > Hi, Colin Watson wrote: > > > Incidentally, I think that python2.3 should definitely depend on python, > > even if unversioned. This is what python2.2 in testing does, and it > > would avoid the "I installed python2.3 but my programs that use > > /usr/bin/python still don't work!" FAQ. > > So include a "Recommends: python" control line. As per policy, section > 7.2, that would be strong enough for this situation. ("The Recommends > field should list packages that would be found together with this one in > all but unusual installations.")
Yeah, the "Recomends" would be much better than a "Depends". There are several problems with having python2.3 with "Depends: python (>2.3)". The first problem is it is not true; there is no reason why python2.3 shouldn't work with python (2.1), or even without any python package at all. This is what is causing unnecisary conflicts between mixed testing/unstable systems. The second problem is is when we get python (2.4), a new python2.3 package will need to be released just to fix the dependencies. The Python Policy was designed so that no pythonX.Y(-foo) packages would need to be updated when python (X.Y+1) is released. Having the python policy more visible would probably do more to answer FAQ's than having busted dependencies on packages :-) -- ---------------------------------------------------------------- Donovan Baarda http://minkirri.apana.org.au/~abo/ ----------------------------------------------------------------