Re: [whatwg/url] Should file URLs have opaque hostnames? (#599)

As an update, driven by the exciting increase in interop as part of Interop 2024 (thanks everyone for your efforts!), I think we can resolve #599 (this issue), #618 (preserve `localhost`), and #805 (permit `@` and spaces, though escape them) together.

Some thoughts on the specifics:

- We should continue to parse IPv6, but don't try to normalize "DNS-based" IPv6 to it. [RFC 8089](https://datatracker.ietf.org/doc/html/rfc8089) expects IPv6 and we have it for all URL schemes so far. (I hope @karwa forgot to include `[` and `]` above and there's not `file:` URL specific IPv6 syntax on Windows.
- I would be okay with trying to detect IPv4 and parsing it in the exact same way we do today for special schemes. In fact, now that we are all much more aligned on non-special schemes we could consider doing that across all of them for IPv4. (This might be too ambitious and maybe we should not change this aspect of opaque host parsing at all, be it for `file:` or other schemes. Open to input.)
- In the "file host state", we should percent-encode `@` and space before appending them to _buffer_. We should evaluate if there are other code points that need a similar treatment.
- I'm a little bit concerned about `localhost`. According to https://www.w3.org/Bugs/Public/show_bug.cgi?id=26387 Internet Explorer did strip `localhost`. Perhaps this only happened under limited circumstances? I would love input from @dandclark on this what would be best for Windows. Also looping in @SimonSapin as the original reporter.

I think we can even make changes here before Interop 2024 is frozen (February 2025) as all the `file:` URL tests should be excluded, so thoughts welcome and let's see if we can finally settle `file:` URL hosts.

That still leaves several Windows drive letter issues that I think we should also seek to resolve those, but tackling those independently seems okay. @dandclark if you could generally review all https://github.com/whatwg/url/labels/topic%3A%20file issues from a Microsoft/Windows perspective I'd greatly appreciate it.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/whatwg/url/issues/599#issuecomment-2514616966
You are receiving this because you are subscribed to this thread.

Message ID: <whatwg/url/issues/599/2514616966@github.com>