[Python-modules-team] Django 1.9

[Brian May]

Raphael Hertzog <hertzog at debian.org> writes:

> I think we should discuss it indeed. Right now I'm tempted to package only
> LTS releases in Debian unstable and use "experimental+bikeshed" for the other non-LTS
> releases.

Yes, this is a point. It does make sense to have only LTS releases on
Debian stable. Upstream will support it for us until at least April 2018
- assuming we can upload new versions to stable or security updates. How
does this date fit in with the release schedule for stretch?

To be completely contrary, here are some arguments against however:

* Django upstream doesn't always ensure upgrades between LTS versions
  are painless. Consider upgrading Django 1.4 to Django 1.8 - I suspect
  upgrading to Django 1.6 first would have been better and then running
  the south database migrations. Yes, hopefully this situation won't
  happen again...

* Similarly, if we keep Django 1.8 in Debian, Debian maintainers will
  become more complacent about ensuring compatibility with future Django
  versions. The new version in experimental is likely to get
  forgotten. Instead of packages fixed for 1.9, 1.10, 1.11 we will
  suddenly jump from 1.8 to 1.11 (assuming this will be the next LTS,
  according to the documentation it will be) in unstable and have to fix
  all version increments at once. As a result we could get more
  resistance to jumping to the latest LTS version, especially if it gets
  released not long before the Debian freeze.

  This could be reduced by automatic / semi-automatic testing packages
  against Django in experimental, however tests don't always catch
  everything.

* Are there any compelling improvements in 1.9 that we should try to get
  in the next stable? Trying to lookup the release notes, however not
  able to access the website right now.


> And we should just fix Python 3.5 support in 1.8.x. Brian, can you file
> bugs for the issues we have?

So far I don't know of any problems.
-- 
Brian May <bam at debian.org>