Chris Angelico rosuav at gmail.com
Sun Aug 23 01:06:44 CEST 2015

On Sun, Aug 23, 2015 at 3:08 AM, Tim Golden <mail at timgolden.me.uk> wrote:
> On 22/08/2015 02:02, Chris Angelico wrote:
>> The security concerns of XP aren't Python's problem, and Python isn't
>> in the business of twisting people's arms to make them upgrade just
>> for the sake of upgrading. However, every new version of Windows
>> introduces new APIs and features, so maintaining support for an older
>> version means ignoring all features added since then; conversely,
>> dropping support for XP means taking advantage of anything that was
>> added in Vista. That's why the change in support.
> Thanks for saying this, Chris. Just to add, from a Python developer
> perspective: any system -- Microsoft or not, open or not, old or new --
> which core Python supports, brings a measure of complexity to the codebase.
> #ifdefs, conditional LoadLibrary calls &c. From the point of view simply of
> the maintenance burden, less code is better. Obviously there is more to
> deciding on platform support than code maintenance ...
> This isn't some kind of political move by the Python dev team to undercut
> Windows users: it's entirely pragmatic. And using the Windows support
> calendar is a common-sense way of giving ourselves a set of cut-off dates.

Precisely. Every time you support multiple versions of some
dependency, you have to test your code on all of them, and in the
common case (new features added in newer versions), you have to target
the oldest and weakest version. When you're writing a Python program
that has to run on CPython back as far as 2.4, there's a lot you can't
do... dropping support for everything pre-2.7 lets you improve your
code significantly. Does dropping support for Python 2.4 consist of
"undercutting RHEL users"? Nope. It's that same pragmatism - I want a
cleaner codebase.

CPython 3.4 will continue to run on Windows XP. If you're still using
an old Windows, you just have to keep using an old Python too.
Eventually Python 3.4 will be out of support, but at that point, it's
no different from the OS anyway. There's nothing stopping you from
using an ancient OS, an ancient CPython, and an ancient Python
application, if that's what it requires... I'm not sure what OS people
are running Python 1.5 on, but if anyone complains that it doesn't
install properly on Windows 10, I rather doubt that python.org will
release a patch :)


More information about the Python-list mailing list