Plugins?

[webknjaz]

Looking into the project more, it seems that some Sphinx sites might be heavily used in some ecosystems, but not the others. This project pulls in a huge number of references that might not be needed in some projects. So I'm thinking that it'd be reasonable to have a concept of groups of intersphinx sited. So the end-users might use subsets of references by group. Like 'python-generic', 'science', 'ml', 'web', 'data-science' etc.

This might work for a number of things. But what if there's an ecosystem that is fairly isolated, and it wouldn't make sense for it to be listed in this project's hardcoded lists?

I think, that it would be interesting to see this implement a plugin system that would allow external installables to augment or replace the list of intersphinx entries with their own. I can see some ecosystems maintaining their own lists of intersphinx entries maintained as a PyPI project and used across their ecosystems. Building on top of the idea #55, they'd have this project in their extensions list but would pip install intersphinx-registry-custom-ecosystem in the same env, and would get a bunch of entries into their Sphinx envs automatically.

This plugin mechanism wouldn't need a lot of fancy stuff. It could rely on pluggy (the lib that tox and pytest use) but it could also rely on the bare-bones importlib.metadata. Such projects might even get away with shipping a single JSON or TOML file + a few lines of metadata in pyproject.toml with no Python. Or, they could have hooks with some custom logic in Python too.

WDYT?