Closed jameshilliard closed 2 years ago
:-1:
-1 from me as well. Setuptools must not be the only option for bootstrapping the ecosystem, or we're stuck in the same situation of one backend being special that we're trying to avoid with our standardisation efforts.
Setuptools must not be the only option for bootstrapping the ecosystem
Multiple options is fine but I don't see why setuptools shouldn't be an option at all for a critical toolchain dependency like this, I mean it's by far the best positioned for pep517 bootstrapping in general since everyone already has useable build+install tooling for it already.
we're stuck in the same situation of one backend being special that we're trying to avoid with our standardisation efforts
So it should be flit
instead? I mean that sounds to me like we would just be swapping one special backend for another.
Anyways distributions don't really want many different build backends in the first place(since that's just more maintenance burden for us but at least it's manageable as long as there aren't circular dependencies) but having multiple different frontends required just for pep517 bootstrapping is a problem as it really makes the integration a good bit more complex.
Closing this for now, so we can try to keep discussion in one place: https://github.com/pypa/build/issues/430
We can reopen it later if necessary, but I don't think that's likely.
Fixes #138.