That's the wrong bit of spec. I think the spec is unclear here, but the relevant part is the validate capabilites algorithm. In this case the question is "is foo:unmatched the name of an extension capability"? I think the spec is unclear about whether any capability including the : matches that condition. My understanding is that the intent is in fact that only implementation-recognised extension capabilites are matched there. So I think the test is correct to the extent that it's trying to test something meaningful, but there ought to be a spec clarification to be sure.

Note that this test is also in conflict with https://github.com/web-platform-tests/wpt/blob/master/webdriver/tests/new_session/support/create.py#L38, where "test:extension" is considered to be a valid capability. If we decide to interpret the spec as intending to reject all unrecognized extension capabilities, then "test:extension" should be moved to invalid_data section.

To your last point, we can’t reject unknown extension capabilities because they may not all be meant for the endpoint node. For example, there may exist extension capabilities in the new session request that are meant for intermediary nodes, so the design of the capabilities matching is meant for them to be passed through uninterrupted. This is why we test that test:extension is not being rejected.

If test:extension is allowed, then shouldn't foo:unmatchable also be allowed?

I also do not understand the reasoning behind test:extension being allowed if foo:unmatchable must be rejected. An explanation would be really appreciated!

Updates on this? Would like to not forget about this PR

This issue was discussed during TPAC (https://www.w3.org/2019/09/19-webdriver-minutes.html#item11), and the resolution was to always accept unknown extension capabilities. So I believe this PR should be merged. Could someone with write access to the repo handle it? Thanks.

@JohnChen0 is going to send a PR to clarify the spec.

@JohnChen0 In light of w3c/webdriver#1454 (review), perhaps this should be changed to reject unknown extension capabilities instead of beng deleted?

@JohnChen0 In light of w3c/webdriver#1454 (review), perhaps this should be changed to reject unknown extension capabilities instead of beng deleted?

@andreastt Please see my comment at w3c/webdriver#1454 (comment), and the TPAC minutes. My understanding is unknown extension capabilities should be accepted instead of rejected.

WebDriver spec has been updated for clarification (w3c/webdriver#1454). Can this PR be merged now? Thanks.

Retriggered tests. We can merge once they pass.

Ahh we need a rebase.

Done, and thanks for the patience!

Thanks!