On Fri, Oct 20, 2017, at 01:18 PM, Donald Stufft wrote:
I guess we shouldn’t have done PEP 517 or PEP 518 because, by your logic here, since it won’t be supported by existing tooling, there won’t be any incentive for people to use it ever.
I see this as having a similar purpose to those PEPs: reducing dependence on setuptools. The difference is that for building packages, pip explicitly uses setuptools, so the practical way forward was to define an alternative to achieve the same ends. For this, the existing mechanism does not directly rely on setuptools, so it's sufficient to document it so that other tools can confidently produce and consume it.
I also get annoyed at times by arguments that it's not worth improving something because it will be a long time before the change is useful. But I don't think that's what Nick is saying here.