[Python-Dev] Intent to accept PEP 561 -- Distributing and Packaging Type Information

[Nick Coghlan]

On 23 June 2018 at 01:16, Guido van Rossum <guido at python.org> wrote:
> That sounds like you're supporting PEP 561 as is, right?

Aye, I'm personally fine with it - we do need to do something about
automatically reserving the derived names on PyPI, but I don't think
that's a blocker for the initial PEP acceptance (instead, it will go
the other way: PEP acceptance will drive Warehouse getting updated to
handle the convention already being adopted by the client tools).

> Excuse my
> ignorance, but where are API testing stub interfaces described or used?

They're not - it's just the context for Donald referring to "stubs" as
being a general technical term with other meanings beyond the "type
hinting stub file" one.

As such, there's three parts to explaining why we're not worried about
the terminology clash:

- Ethan searched for projects called "*-stubs" or "*_stubs" and didn't
find any, so the practical impact of any terminology clash will be low
- there isn't an established need to automatically find testing stub
libraries based on an existing project name the way there is for type
hints
- even if such a need did arise in the future, the "py.typed" marker
file and the different file extension for stub files within a package
still gives us an enormous amount of design flexibility

Cheers,
Nick.

-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia