increasing the python requirement

[Vincent Ladeuil]

>>>>> Toshio Kuratomi <a.badger at gmail.com> writes:

    > On Tue, Jan 04, 2011 at 04:54:29PM -0600, John Arbash Meinel wrote:
    >> -----BEGIN PGP SIGNED MESSAGE-----
    >> Hash: SHA1
    >> 
    >> On 1/4/2011 4:53 PM, Toshio Kuratomi wrote:
    >> ...
    >> 
    >> >> There are no immediate plans to implement a new format. But it is
    >> >> certainly something that we would be unlikely to backport into an old
    >> >> stable release.
    >> >>
    >> > A plugin would be fine (as long as it works ;-).  Is that what you're
    >> > proposing?
    >> > 
    >> > -Toshio
    >> 
    >> I'm saying that if we do introduce a new format, and have discontinued
    >> python2.4 compatibility, it shouldn't be hard to write a plugin for
    >> older bzr to interoperate with newer ones.
    >> 

    > Thanks,
    > -Toshio "who feels bad asking if someone else plans to do work on some piece
    > of free software but knows he lacks the time to meet even his current free
    > software commitments :-("

Hehe, right to the point.

This is the core of the discussion, should we continue to spread our
efforts maintaining compatibility with an increasing number of python
versions or do we start dropping compatibility with the oldest ones. And
if we drop them, who's left in the cold and what can we do.

    > Just one clarification -- are you saying, "it would be easy for us
    > (bzr developers) to create a plugin so we should plan on doing
    > that rather than support the entire codebase on older python" or
    > "it would be easy for some interested party to create a plugin if
    > they choose to"?

Well, we are in the same boat as you as far as resources are concerned
:)

But whatever new format we create, we will have a bunch of associated
tests, which should help lower the maintenance effort on the plugin
side (modulo test backports, yada yada, at least the specs will be
clearly defined).

    > I ask because well... I'll be able to move the servers I care about to RHEL6
    > (and python-2.6) so I probably won't scratch that itch.

And if enough users, reasonably, take the same approach, then this
discussion may be enough...

    > But I know that I'll get bug reports asking why the bzr on RHEL5
    > doesn't work with launchpad and I hate closing bugs as CANTFIX.

+1

        Vincent