[Bf-committers] Proposal for clarified VFX Reference Platform Support

[Campbell Barton]

On Sat, Jan 22, 2022 at 12:28 AM Dalai Felinto via Bf-committers
<bf-committers at blender.org> wrote:
>
> Hi,
> I believe Blender should stick to the VFX platform.
>
> After all that has been said, I think it may boil down to making a decision
> between immediate known benefits and strategic uncertain long-term benefits.
>
> * On one hand we have tangible benefits for some users that we know of
> (e.g., Python scripters will benefit from Python 3.10).
> * On the other hand sticking to the VFX platform can pay off in the long
> run with making Blender more likely to be integrated in large pipelines.

It *could* but from my perspective with Python - I didn't see any
evidence this was the case for Blender 2.8x / 2.9x which followed the
VFX platform Python version (for 2.8x, 2.9x ... until we ran into
problems, see [0]).

A reminder that using the VFX platform's Python version means (at least
some of the time) Blender's Python version won't be getting bug-fixes
as each release only gets fixes for 18 months [2].
To be fair, running into bugs in Python is fairly rare, so I don't
consider this a huge down side. Nevertheless missing out on fixes +
new features is still a down side.

At some point strategic decisions like this should have tangible
benefits beyond the *possability* of attracting a user base. Maybe it
will be different this time - in that case there should be an
explanation as to why.

> One of the long-term goals for the Blender project, is to welcome more
> contributions by the industry. And I think investing on that vision trumps
> the immediate benefit the latest Python (or other library) brings to
> Blender.

It seems likely to me the benefits of Python sticking to the VFX
platform are being perceived as greater than they actually are (beyond
messaging that "we support the VFX platform").

While there are scenarios with Python ABI compatibility (relating to
the VFX platform) can cause problems, and I'm not saying nobody ever
ran into these issues - this seems more like a corner case which isn't
actually blocking people in the VFX industry using Blender in
practice. If it was, they were not vocal when it was announced we
planned to upgrade to Python 3.9.

Part of my skepticism gets into the details of what the VFX platform
is generally used for, from what I can gather the QT graphical toolkit
and it's Python bindings are a significant factor deciding if Python
can be upgraded for the VFX platform. (PySide [1] sometimes lags in
it's Python support).

Since using QT from Blender is impractical (last I checked at least),
it's not clear if sticking to an older Python has all that much
benefit for VFX users either (as native Python modules typically
aren't a problem).

> To have studios contributing to Blender is a two-way street. And Blender
> sticking to the VFX is the least the Blender project can do on its end.

As far as I can see we tried this and it didn't yield much, if you
propose to try it again - it's reasonable to question what success
would look like - and what would be a reasonable time frame to decide.

> I look forward to see this and other efforts in that direction, such as
> onboarding, code documentation, infrastructure and development practices.

Supporting the VFX platform and other topics such as onboarding,
development practices ... etc seem unrelated.

> To move this forward I'm setting up a call with the other bf-admins next
> week. We will report back afterwards.
>
> Thanks everything for the contributions,
> -Dalai-

[0]: https://bugs.python.org/issue35523
[1]: https://pypi.org/project/PySide
[2]: https://www.python.org/dev/peps/pep-0602/