"Extract a MIME type" algorithm should pick the first entry?

[tyoshino]

extract header list values algorithm returns a list. extract a MIME type and sites calling extract a MIME type look using the result as a byte sequence.

Should extract a MIME type pick the first entry of the result of extract header list values if it's not null?

[annevk]

Yeah, although we should probably have tests for multiple Content-Type headers. I haven't researched that much at all.

[tyoshino]

I see. Considering the following, extract header list values would return failure when there are multiple entries. So, the fix is not what I said in the OP but pick the single entry while asserting it's a single element list if it's not either null or failure.

• https://tools.ietf.org/html/rfc7230#section-3.2.2
• https://tools.ietf.org/html/rfc7231#section-3.1.1.5
• https://tools.ietf.org/html/rfc7231#section-3.1.1.1

[annevk]

Is that what browsers implement? What if you navigate to a URL with two Content-Type headers. Do you get a network error? We should probably have a bunch of web-platform-tests. The HTTP WG generally does a bad job of making sure their requirements can actually be implemented.

[tyoshino]

Chrome's fetch() is currently broken (Blink's ExtractMIMEType is using combined value http://crbug.com/713634), but looks Chrome has been not erroring for multiple Content-Types but using the last Content-Type header before the break.

Navigating to a resource with multiple Content-Type also works. Haven't checked which one is used for navigation.

[annevk]

Another angle to test would be a single Content-Type header with a comma-separated value. They ought to be treated identically, but browsers...

Relates to whatwg/mimesniff#30.

[annevk]

whatwg/mimesniff#30 (comment) has my findings on this. Input there much appreciated, even though we'll likely define it here.

[annevk]

For anyone subscribed to this issue, please note that #831 is pretty close towards solving this. Tests are linked from there as well.