Sort out MIME type tests

[annevk]

Tests:

• MIME type parameter parsing tests web-platform-tests/wpt#7764 (navigation; charset handling, Blob/File/Request/Response)
• Adjust XMLHttpRequest Content-Type handling web-platform-tests/wpt#8422 (XMLHttpRequest Content-Type)
• Add more XMLHttpRequest overrideMimeType() tests web-platform-tests/wpt#8449 (XMLHttpRequest overrideMimeType())
• data: URL tests web-platform-tests/wpt#6890 (data: URLs)

Proposed processing model:

• Revamp MIME type section #36

TODO:

• Review Content-Type parsing (MIME type parsing) #30 and Do separators need to be preserved when parsing? #39 to come up with a more exhaustive set of endpoints and work out a testing story for those endpoints.
• Maybe file more implementation bugs.
• Close issues that were mostly raised as a tracker for tests and are covered by the above.
• Evaluate test results in all browsers as a means to: potentially change the processing model and tests and add more tests

[annevk]

Here's a start with an overview of the endpoints. Review and additions appreciated:

Endpoint	Parsing	Serializing	"Supports"	"String"
media.canPlayType()/MediaSource.isTypeSupported()/MediaRecorder.isTypeSupported()	✅	❌	✅	DOMString
XMLHttpRequest's Content-Type's charset manipulation	✅	✅	❌	ByteString
XMLHttpRequest's overrideMimeType()	✅	✅	❌	DOMString
Navigation (hard to test)	✅	❌	✅	ByteString
data: URLs	✅	❌/✅	❌/✅	ByteString
Blob's type	?	?	❌	?
Response and Request	✅ (not implemented)	✅ (not implemented)	❌	ByteString
DataTransfer	?	?	✅	DOMString

[annevk]

Thoughts:

• canPlayType() has adequate parsing tests. Those in charge of the isTypeSupported() methods should borrow from there.
• XMLHttpRequest's Content-Type's charset manipulation has some tests. It's worth noting that we only need to serialize the parsed value here if the given value can be successfully parsed and has a charset parameter.
• XMLHttpRequest's overrideMimeType() could use more tests. The problem here is that the serialization depends on Blob's type which is a mess (and doesn't always include parameters).
• Navigation: most of the encoding tests used to tease out parameter parsing are navigation tests of sorts.
• data: URLs: I created lots of tests for these.
• Blob's type: it would really help to get some agreement on what to do here.
• Response and Request: depend on Blob's type again...
• DataTransfer: fairly limited MIME type exposure.

My takeaway is that we need agreement on Blob's type to make lots more progress, but we can make some progress around XMLHttpRequest charset manipulation.

[annevk]

web-platform-tests/wpt#8422 has the XMLHttpRequest Content-Type tests and web-platform-tests/wpt#6890 has them for data: URLs.

Until Blob is sorted that'll have to do, unless someone has further suggestions.

[annevk]

This is basically done, but leaving it open for now as it's a rather nice collection of data I might have to reference a few more times before all the patches have landed.

[snuggs]

WOW @annevk this is really great work. I have been perusing mimesniff as of late and definitely a space i'm interested in. I do realize you are keeping this open for "novelty"'s sake 😄. Is there anywhere here I can either help out? At minimum can contribute documentation.

I'll start understanding with the WPT tests in the description. I think I just answered my own question. 👍

[annevk]

The only bits I still need to finish are with respect to XMLHttpRequest. Implementations still have to align too unfortunately.

[snuggs]

@annevk been wondering this but didn't know how/where to ask. preload seems to send */* for everything that's not style,font,etc.sinceas=document` is broken ATM in chrome. I am versed at the "why?" this is happening as the browser only needs to know if resource is an image or not if not mistaken.

Wanted to know where do I bring up I think as=document should send the same mimetype as using the address bar (or a frame for that matter)?

For example Chrome sends text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8.

I didn't know if this was a discussion for W3C/preload, mime sniff, HTML, WPT. Still getting the hang of wrangling kittens in open source (with your help of course).

Thanks in advance and I can delete comment after you respond if you like.

[annevk]

That's part of Fetch if you're talking about the Accept request header.

[snuggs]

@annevk copy that. Thanks a ton!

[snuggs]

@annevk was going to submit an issue as didn't know where to ask questions. As you taught me, check existing issues/prs first before creating another. Even found the source in chrome. :-) Looks like there is a discussion going on to follow. Thanks again.

whatwg/fetch#274

[annevk]

All the tests discussed here have landed.