[pabszero]

With ipv6_address=True, email("user@[2001:db8::1]", ...) accepts a bare IPv6 literal, but user@[IPv6:2001:db8::1] is rejected. The tagged form is defined by RFC 5321 section 4.1.3.

Recognize the tag case-insensitively and validate its payload as IPv6. Preserve opt-in flag behavior, legacy address literals and local-part validation. Regressions cover compressed, full and IPv4-embedded addresses, invalid payloads (including zone IDs), repeated brackets and compatibility.

Validation on Windows:

• New regression selection against the original implementation: 4 failed, 6 passed; all pass with the fix.
• Full suite: 1,181 passed on both CPython 3.12 and 3.13.
• Configured tox lint, type and SAST environments passed; check-only formatting passed.

Other operating systems and Python 3.11/3.14 were not run locally. AI assistance (OpenAI Codex) was used for investigation, implementation, review and testing.

[nandgator]

Thanks, I've reviewed this and would like to merge it. The tagged literal matches the IPv6-address-literal rule in RFC 5321, the tag is case-insensitive as ABNF strings are, and zone IDs, CIDR suffixes and IPv4 payloads are rejected.

It now conflicts with master in tests/test_email.py, because I added regression tests at the same spot. Both sides only add tests, so the resolution is to keep both. Could you rebase onto master and force-push? I'll approve CI and merge once it's green.

[pabszero]

Resolved the conflict in ce9157c, keeping your IDNA lookalike tests and all IPv6 literal regressions. I used GitHub’s web conflict resolver, so this is a merge from master rather than a force-pushed rebase. The complete resulting tree matches the locally tested rebased tree.

On Windows CPython 3.12 and 3.13, the updated full suite passes: 1,584 tests per version, including all 45 email cases. Tox lint, type and SAST checks and the check-only formatter also pass. Hosted workflows are awaiting maintainer approval.