On 10/6/2012 16:42, David Holland wrote:
On Thu, Oct 04, 2012 at 09:11:58AM +0200, Thomas Klausner wrote: > Does anyone know of a reason why [python26] should stay? Yes. It is still being used, and it's neither broken nor unsupported upstream.
"It is still being used" needs clarification.As of right now, only a single package requires it, lang/py-psyco, which the upstream maintainer says is dead, unmaintained, and users should migrate to py-py. As such, lang/py-psyco should be removed as soon as convenient. When that happens, _NOTHING_ in pkgsrc requires it, thus its use is limited to outside pkgsrc.
However, based on follow-on discussions that this response did not take into account, the python2.6 package itself is not the issue. That can stay forever as far as I am personally concerned. I am pointedly questioning why we build py- packages with 2.6 when they present no benefit _AT ALL_ over building with 2.7. Eventually the same question will be put forth to python 3.1 and 3.2 packages.
I think we need to review the multiversion policy. Answers such as "it only burns up bulk builder resources" is not a well-thought out defense for keeping them. Having 5 to 6 versions of the same py- package because of the language's self-inflicted lameness seems a bit much to me.
I'm suggesting to pick ONE 2.x python (2.7 obviously) and ONE 3.x python, build packages with those two, and suppress the other versions. I'd like to know what mk fragment controls that.