Oof, thanks for this info. That may be worthwhile to implement. Perhaps there's also a package that implements this already, too. I'll have to look into this. thanks!

I might submit a pull request to implement RFC 7303. At a glance, it looks like this would require changes to feedparser/encodings.py and docs/character-encoding.rst. Would you prefer if I just replaced the old behavior or added an option to switch between them? Scenarios that could start having problems as a result of a wholesale change (or an option that defaulted to the new behavior) seem very unlikely but not strictly impossible to me.

It seems like RFC 7303 may be specific to content transferred over HTTP. Is that accurate?

I have started to transition the codebase to the requests package, and would prefer to avoid doing additional detection on top of what that package provides (although local files and binary content fed directly into feedparser would still need some detection).

Is it still necessary to implement RFC 7303?

It's specific to content with a MIME Content-Type header. feedparser does let users specify this header themselves in the response_headers argument to parse. I don't know if you care about preserving that feature or if there's a way to do it with the requests package.

Does requests handle format-specific charset detection like for XML? Something somewhere in the pipeline should implement RFC 7303. I agree that feedparser isn't the best place for it, since none of it is specific to Atom or RSS.