Gregor Hoffleit <[EMAIL PROTECTED]> writes: > I'm still alive, I'm not lost ;-)
You're not dead, which is the most important ;-) > And that's the problem where I was stuck. > > The dependencies of the current experimental python1.5 packages aren't > good enough to allow an easy upgrade from potato or earlier. AFAICS, I > would have to include empty transitional packages for almost all > python-* packages built from the python source. That's butt ugly IMHO. That is what happened for the Perl transition. I don't think you can avoid this. What you have to do: . provide transitional packages. These packages will guaranty that nothing will break during the transition. . ask people to change their dependencies. You will regularly list packages that have not been updated and warn their maintainers to do so. > > Furthermore, it's my current impression that we have to orphan all > python-* packages in order to get a clean upgrade path! Huh?! > > There are many packages with broken dependencies in potato (i.e. > packages that install stuff in /usr/lib/python1.5, but don't have a > versioned dependency on Python 1.5). Therefore, any future python-base > package either has to be compatible with stuff in /usr/lib/python1.5, > or has to list all those broken packages as conflicts. Either is ugly. Do we have a list of them? > > Therefore I think we have to get rid of all the problem-ridden package > names, i.e. at least python-base and python-tk, perhaps even python-dev. > I don't see any other solution. Sorry, I did not understand this. ... > Python package maintainers should then change their packages to build > python1.5-* and python2.1-* packages (python2.0 if needed), and make > them depend on python1.5-base etc. That would remove the need for > versioned dependencies. Right. > > At a later point, we could implement a kind of registry like the emacsen > have, so that third-party packages could build only one python-* > package, that registers itself with all existing pythons. We have dealt with this a long time ago and we have always said that it was too late because we were approaching the freeze. But we are still not frozen (optional packages) and I think it's time to implement this, whenever is the freeze (it will be usefull whether it is integrated in woody or not). Cheers, -- Jérôme Marant <[EMAIL PROTECTED]> <[EMAIL PROTECTED]>