Ensure opaque paths always roundtrip

[annevk]

In fdaa0e5 we tackled a problem whereby removing the fragment or query from a URL with an opaque path through the API would not make the URL roundtrip due to the opaque path being able to end in non-percent-encoded spaces.

However, this failed to address other ways of serializing the URL. As such this is a new approach whereby opaque paths simply cannot end with non-percent-encoded spaces. Enforcing this in the URL parser allows us to completely revert the aforementioned commit, greatly simplifying the API implementation.

Fixes #784.

---

• At least two implementers are interested (and none opposed):
  • Chromium (at least during Interop 2024)
  • Gecko
  • WebKit
• Tests are written and can be reviewed and commented upon at:
  • URL: trailing spaces always get removed from opaque paths web-platform-tests/wpt#51129
• Implementation bugs are filed:
  • Chromium: https://issues.chromium.org/issues/400919461
  • Gecko: https://bugzilla.mozilla.org/show_bug.cgi?id=1951963
  • WebKit: https://bugs.webkit.org/show_bug.cgi?id=289160
  • Deno: Ensure opaque URL paths always roundtrip denoland/deno#28395
  • Node.js: Ensure opaque URL paths always roundtrip nodejs/node#57329
• MDN issue is filed: Since URL standard update: small change to URL API mdn/content#23320 is not fixed yet we can perhaps reuse that.
• The top of this comment includes a clear commit message to use.

(See WHATWG Working Mode: Changes for more details.)

---

Preview | Diff

[annevk]

@karwa @hayatoito @valenting what do you think? We should be able to make test changes again so would be great to hear what the interest in this change is.

[valenting]

I'm supportive of this change. It's a nice simplification, and slightly better than having to strip spaces from the path when clearing the search or hash. I plan to be making these changes in Firefox once WPT tests are available.

[karwa]

(thanks for this, and apologies for the long delay!)

+1. This is a lot more predictable for users.

I hope it is web-compatible. I understand that there are use-cases which unfortunately require that unescaped spaces are preserved (e.g. javascript: URLs), but I can't imagine that trailing spaces are important to any of them.

[annevk]

Unfortunately I think I need a bit more assurance that everyone is still willing to try this after encountering some unexpected data: and javascript: URLs in tests: #784 (comment).

[domenic]

I guess this also obsoletes #785.