The known problem scenario is Redhat Enterprise Linux shipping only Python 2.6.6, and there are people using Pywikibot on these boxes, and those boxes are locked down.

We've managed to keep Python 2.6.6 supported (the future package makes this easy), but T102461 is a real hurdle.

A good step forward would be to segregate py2.6-specific code into a separate file.

Resolved as T114464, I guess? :)

In T103063#1700385, @jayvdb wrote:

No. That is a release branch. master continues to support 2.6 while we have users who need it. Start an RFC if you want to drop support against the needs of some of our userbase.

A stable release drops support while master, supposedly bleeding edge, carries on? WTF?

It is more feasible to keep 2.6 users supported if they are using master.

We are not going to backport python 2.6 fixes to v2.0, as that would require a panicked response and a release whenever a dependency drops 2.6 support, which isnt worth the effort.

i.e. we believe python 2.6 can't be stable, and is 'best-effort' support only. It will break occasionally for reasons beyond our control.

An interesting read regarding dropping support for Python 2.6. http://www.curiousefficiency.org/posts/2015/04/stop-supporting-python26.html

We could ask our RHEL6 users why they are not able to use the available 'Python Upgrade Paths' described there.

Interesting post @jayvdb . I'll follow up on that next week.

ping @Multichill