On Mon, Apr 6, 2009 at 11:00 AM, Greg Stein <gstein_at_gmail.com> wrote:
> I think this is an issue that the svn devs need to discuss: what
> Python versions are required, expected, and supported.
>
> I'm going to throw out a strawman here for discussion.
>
>
> End-users
> - Python is not required to use "core" svn
> - Python scripts that are part of our distribution bundle
> - Minimum: 2.4
> - Supported: 2.x
> - Possible for some scripts: 3.0
>
> Packagers
> - Python is not required
>
> Developers
> - Python is required: gen-make, test suite, dist packaging, etc
> - Minimum: 2.4
> - Supported: 2.x
>
>
> There are two primary goals with my selections above:
> - Support end-user choices
> - Reduce maintenance costs
>
> It would be nice to support end-user scripts < 2.4, but that increases
> maintenance costs for us. I believe that 2.4 is a fair balance between
> user choice and our cost. During the 1.6 release process, we ended up
> bumping the minimum user version to 2.4. That happened without
> explicit discussion, so I'm bringing it up for closure on whether that
> is what we're going to shoot for.
>
> Also note that Python 3.0 is *not* part of the Developer toolset. Not
> even supported. I've been reviewing the changes on trunk for the 3.0
> compatibility effort with increasing concern. The changes introduce
> maintenance cost (*), yet with no perceivable benefit to the Developer
> community. If we confine our support to 2.x, then the dev toolset will
> be easier to expand, modify, and refine.
>
> Thoughts?
I am all for someone doing work so that our Python bindings could work
in 3.0, but I do not see any reason why our test suite or developer
tools should work with 3.0. I do not believe we are within 5 years of
a developer desktop not having a Python 2.x interpreter installed and
running our test suite in 3.0 offers zero benefits. It is not like it
makes the tests run 2x faster. At the same time, the code that is
being added makes the existing code harder to maintain and it is
regularly breaking the tests on Windows and other platforms.
The costs of the Python 3.x work seems to dwarf the benefits, which I
see as almost zero. I think all of this Python 3.x compat code should
be removed from trunk and this work should stop. Let's just declare
that the tests must be run with 2.x.
I do not know what the state of our bindings and Python 3.x is. I
could see that someone might want to write their own tools using
Python 3.x and our bindings. If someone wants to make that work,
great. But it needs to come without the costs that the current work
does.
--
Thanks
Mark Phippard
http://markphip.blogspot.com/
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1562356
Received on 2009-04-06 17:10:05 CEST