HTTP/2fy request-upload test around streaming

[yoichio]

That's because wpt h1 server doesn't support chunked encoding but h2 supports it natively.
Related thread: whatwg/fetch#966

[foolip]

The wpt-chrome-dev-* CI failures here were likely due to #19360. A workaround was landed in #28238, so I'm closing and reopening to re-trigger the CI checks.

[annevk]

I guess we haven't quite resolved whether to support H/1, but ideally we have some kind of test for that as well.

[yoichio]

I guess we haven't quite resolved whether to support H/1,

Yes but supporting H/2 has been decided. Should we also have a wpt that confirms the feature is not supported over H/1?

[yoichio]

Undid confirming failure tests, which required wptserve/wptserve/server.py change. Will do later-on.

[yoichio]

Ping?

[yutakahirano]

LGTM.

[annevk]

@yoichio I'm not sure I understand your comment above. Has there been a decision to only support H/2 and up? If so, a test or a few tests for non-support for H/1 would be great.

(To be clear, that shouldn't be a blocker to landing this PR.)

[yoichio]

Has there been a decision to only support H/2 and up?

No to "only" but yes to supporting at least H/2.

If so, a test or a few tests for non-support for H/1 would be great.

That's bit confusing. Should we put a wpt test that points an API doesn't work? This is different from how an API raises an exception.

[yoichio]

I'm O.K to just put same tests both in request-upload.any.js and request-upload.h2.any.js

[annevk]

All I'm saying is that once we make a decision on what to do for H/1 we should have test coverage for that decision.

[yoichio]

Sure.

[yoichio]

Ping?

[annevk]

@yoichio you should feel free to merge this. Or do you want another review?

[yoichio]

I don't have a write access to this repository.
Could you give me the right?

[annevk]

(I've asked on W3C IRC #testing for someone to give you access.)