Can't connect to IPv6 Address with Zone ID

From @alex-fu27 psf/requests#5126:

Requests fails on IPv6 URLs that contain a zone identifier when used with the current urllib3 (1.25).

Expected Result Prior to urllib3 v1.25, I could do this:

import requests requests.get(“http://[fe80::145e:58ff:fec3:4698%enp0s20u1]:8080/”) And it worked as intended.

Actual Result Since I upgraded to urllib 1.25 and requests 2.22, this fails with [Errno -2] Name or service not known, because requests is escaping the % character and trying to access [fe80::145e:58ff:fec3:4698%25enp0s20u1] instead.

Minimum requirements

💵 You can get paid to complete this issue! Please read the docs for more information.

• Read https://datatracker.ietf.org/doc/html/rfc4007 and https://datatracker.ietf.org/doc/html/rfc6874 to understand the feature
• Research and document why urllib3<1.25 worked while urllib3>=1.25 doesn’t work in this issue
• Figure out the best percent encoding strategy for Python across all platforms (Linux, macOS, and Windows). You can reach out to maintainers with a script to test this on their own machine if you don’t have access to a certain platform. A valid outcome of this is “we shouldn’t modify the Zone ID separator” but let’s test instead of assume!
• Implement the feature to forward Zone ID from the authority to the socket.create_connection call.
• Add test cases that can run on CI (might have to use mocks of urllib3.util.connection.create_connection?) but should have the ability to run a live test when available (figure out how to detect this situation at least on Linux)

[sethmlarson]

I’m working on resolving this issue and another issue related to IPv6 Zone IDs.

[sethmlarson]

This issue has been closed in https://github.com/urllib3/urllib3/pull/2629 and and #2635. Thanks @delroth! I look forward to seeing your invoice to collect the bounty.