On 28.10.2018 05:15, Andrew Godwin wrote:
> Right. This is why I think I'm unsure quite how to approach replacing it.
Why would you want to replace it, if what you have in mind is a
different standard for a different audience and use case ?
WSGI is not going to go away as a standard since it is useful
and works well in the context of non-async web applications.
The word "replace" there was maybe not quite right - I think we need an asynchronous equivalent of WSGI for the same reasons we have WSGI (inter-operability between servers, frameworks, and applications). The use case is quite similar, to be honest.
ASGI is designed to be able to encapsulate WSGI applications if desired for backwards compatibility, but I wouldn't expect it to "replace" it per se for a long time, if ever.
You'll win more fans for ASGI by not going confrontational,
since that just sidetracks discussions into non-productive
terrains.
Oh totally. I'm not here to argue, just to work out how to best get to a Python-wide standard we can all benefit from in the same way as WSGI.
> Essentially to put it on the same platform as things like WSGI and
> DBAPI2 - to have a directly accepted standard that forms part of the
> language core. Obviously this is not required for things to function and
> people to develop against it, but it wasn't required for WSGI either, so
> in some ways the reason I think it should be a PEP is pretty much purely
> because WSGI is.
It's a good idea to turn this into an informational PEP similar
to the DB API. These standards are developed outside the usual
Python development process, but provide good guidance for the
Python community at large.
I agree - I don't think it makes sense as either of the other options, and both PEP 3333 and PEP 249 are Informational. They're the two most similar things I can think of in the PEP ecosystem.
Andrew