[Python-Dev] Python and Windows 2000

["Martin v. Löwis"]

Neil Hodgson wrote:
> Martin v. Löwis:
> 
>> See http://bugs.python.org/issue6926
>>
>> The SDK currently hides symbolic constants from us that people are
>> asking for.
> 
>    Setting the version to 0x501 (XP) doesn't actively try to stop
> running on version 0x500 (2K), it just reveals the symbols and APIs
> from 0x501. Including a call to an 0x501-specific API will then fail
> at load.
> 
>     IPPROTO_IPV6 (the cause of issue 6926) isn't a new symbol that
> started working in Windows XP - it was present in older SDKs without a
> version guard so was visible when compiling for any version of
> Windows.

I'm not proposing to actively break Windows 2000 support. I propose to
stop supporting it, which means that we will not strive for Python
running on Windows 2000, and that bug reports that report issues
specific with Windows 2000 will be closed as "won't fix".

I would not want to bump the SDK version while still aiming at Windows
2000 support - as that gets rarely tested, this API version is actually
a mechanism to simplify ongoing support for Windows 2000.

>    I don't really have a strong need for Windows 2000 although I keep
> an instance for checking compatibility of my code and I do still get
> queries from people using old versions of Windows, including 9x. There
> is the question of whether to force failure on Windows 2000 or just
> remove it from the list of known-working platforms while still
> allowing it to run.

I would propose the latter; see above. However, patches that do actively
break it would then be acceptable.

In any case, we would need a warning in the installer that Windows 2000
supports is going away.

Regards,
Martin