PEP 776: Address some of pfmoore's comments #4310
Conversation
This attempts to address most of pfmoore's comments here: https://discuss.python.org/t/pep-776-emscripten-support/84996/11 * Clarify that we are speaking of _platform_ tags. * List some examples of full tags that are Pyodide-compatible * Include code to generate compatible tag list * Reference PR to add this code to pypa/packaging and specify that it will be merged. * Add section about dependency specifier markers * Add section on trove classifier * Add note that adding platform tags should have no backwards compatibility implications * Add link to Pyodide's documentation on building, testing, and ci for package maintainers
hoodmane
changed the title
|
|
||
| We recommend that package indexes accept any wheel whose platform tag matches | ||
| ``pyodide_[0-9]+_[0-9]+_wasm32``. We recommend that package indexes continue not | ||
| accepting wheels that match ``emscripten_[0-9]+_[0-9]+_[0-9]+_wasm32``. |
There was a problem hiding this comment.
It would be good to explain why here. Specifically, what is the reason for having a platform tag that's explicitly not for publishing? We have the "linux" platform tag that essentially means "only usable on the machine it's built on", but that's really just because it took a while to work out how to do Linux compatibility. This is different - we're standardising a tag that people can't publish, and it's not clear what the point is.
Ah, I just noticed, in a part of the text you didn't change, "intended to be the same as the
relationship between manylinux<version> and linux". IMO, this is a bad precedent to follow. There's really very little good reason for the linux tag to exist nowadays, and similarly I don't think there's much point in a set of emscripten_[0-9]+_[0-9]+_[0-9]+_wasm32 tags.
If we want an "unpublishable, usable only on the machine it was built on" tag, I'd prefer we standardised on a single generic name ("local", for example, or maybe just repurpose "linux", although the name is unfortunate in that case) rather than having multiple ones.
This attempts to address most of @pfmoore's comments here: https://discuss.python.org/t/pep-776-emscripten-support/84996/11
📚 Documentation preview 📚: https://pep-previews--4310.org.readthedocs.build/