On Tue, May 8, 2018, at 5:44 PM, Brett Cannon wrote:

If it is an Emacs thing then I would vote "no" since that's very editor-specific and I suspect trying to support every plaintext file format is never-ending.

Currently the metadata spec defines the permitted mimetypes. Maybe we should make this more open ended, and leave the set of useful values up to what projects (primarily PyPI) choose to implement?

The specification could still say that any format should be a) UTF-8 text - so no README.pdf or README.doc, and b) sufficiently 'plain' that it's readable if tools display it unrendered. The second condition is a bit vague - I might find something readable that you don't - but it's meant to exclude things like HTML and Latex.

I don't see any big advantage in specifying a list of allowed content types. Even with the list we have, tools can't rely on any features of the content except that it's UTF-8. The only real difference is whether people who want to use something else file an issue on Warehouse or start a thread here. :-)

Thomas