[3.8] bpo-27657: Fix urlparse() with numeric paths (GH-661)

[orsenthil]

• bpo-27657: Fix urlparse() with numeric paths

Revert parsing decision from bpo-754016 in favor of the documented
consensus in bpo-16932 of how to treat strings without a // to
designate the netloc.

• bpo-22891: Remove urlsplit() optimization for 'http' prefixed inputs.
  (cherry picked from commit 5a88d50)

Co-authored-by: Tim Graham timograham@gmail.com

https://bugs.python.org/issue27657

[orsenthil]

This is a backport PR. Merging this.

[bedevere-bot]

@orsenthil: Please replace # with GH- in the commit message next time. Thanks!

[Ma27]

This is a backport PR. Merging this.

May I ask why this was backported into Python 3.8.1? I totally understand the motivation behind this change, but isn't it a breaking change in a patch release?

To quote an older comment from the original issue (bpo-27657):

I hate to say it, but this may require a python-dev discussion.

I observed a certain fallout from this change (e.g. mozilla/bleach#503), so is there any chance that this gets reverted in python 3.8.2 or do you have any suggestions how to properly deal with cases like localhost:8080?

[orsenthil]

Hi @Ma27 - I re-read the whole issue again, I see that this message https://bugs.python.org/issue27657#msg289557 and the Versions in the issue set played a role in 'reverting' the patch or making the change in a patch release.

This seems like a mistake when a simple path like localhost:8080 could fail in a patch release of Python. I apologize that this was not considered properly.

Let's continue the discussion on https://bugs.python.org/issue27657 on the next steps. If this needs to be reverted again for the patch release. I will re-read associated issues again and share the possible next steps.