Move webrtc/legacy/munge-dont.html out of that directory

[gsnedders]

Per the readme, that directory is mostly concerned with things
specified in
https://w3c.github.io/webrtc-pc/#legacy-interface-extensions, whereas
this test is really just asserting that historical behaviour isn't
still true, which seems like what we normally call "historical" (see
also: /webrtc/historical.html).

[alvestrand]

Do we have precedent for tests that test specified behavior, but for historical reasons we don't expect them to be respected any time soon?
We could easily add more of those - perhaps there should be webrtc/historical/munge-dont-dont.html instead?

[gsnedders]

Do we have precedent for tests that test specified behavior, but for historical reasons we don't expect them to be respected any time soon?

All over the place — plenty of WGs require spec changes come with accompanying test changes, thus a removal of this nature would inherently have a test with it.

If the WG chooses to make a spec change, knowing that implementations don't expect to actually act on it any time soon (or ever!), that is something for the WG to decide how to resolve, in line with the working mode of the SDO the spec is being developed under (and in the W3C case, "how do you meet the requirement of adequate implementation experience if implementations aren't aligning with the spec").

We could easily add more of those - perhaps there should be webrtc/historical/munge-dont-dont.html instead?

That would test what? That the munging does happen? That would be a test out-of-scope of WPT, because that's non-standard behaviour, and would directly contradict the test in munge-dont.html that tests the actual implementation requirements of the spec. We cannot have tests that a fully compliant implementation would fail, excluding tentative and optional tests, and it doesn't meet the bar for either tentative tests (because nobody is intending for that to become specified behaviour) or optional tests (because the spec doesn't make this behaviour OPTIONAL [RFC2119]).

(#49559 removes /webrtc/legacy/onaddstream.https.html on that basis — it is very clear the WG do not wish to specify it, and thus it cannot be either tentative or optional either.)

I have no objection to these tests existing if the spec makes this behaviour optional, similar to what is specified in the Legacy Interface Extensions section today, but it seems like the WG has very much decided to go down the path of requiring implementations to do things they seem unlikely to do.

[youennf]

I would tend to move the test as done in this PR and add tentative to the file name.
I am also fine with adding a historical subfolder.

[gsnedders]

I would tend to move the test as done in this PR and add tentative to the file name.

But it's not tentative; per the documented definition:

Indicates that a test makes assertions not yet required by any specification, or in contradiction to some specification.

The test's assertions are required by a specification and are not in contradiction to any.

We could have a tentative test that munging is allowed, but as far as I'm aware there's no current proposal to change the WebRTC spec to allow it?