Fetch streaming upload

[yutakahirano]

Request for position on fetch streaming upload

• WebKittens who can provide input: @youennf

Information about the spec

• Spec Title: Fetch streaming upload
• Spec URL: https://fetch.spec.whatwg.org/
• GitHub repository: https://github.com/whatwg/fetch/
• Issue Tracker (if not the repository's issue tracker): Streaming upload whatwg/fetch#1438
• Explainer (if not README.md in the repository): https://github.com/yutakahirano/fetch-with-streams/blob/master/streaming-upload.md

Design reviews and vendor positions

• TAG Design Review: Fetch streaming upload w3ctag/design-reviews#754
• Mozilla standards-positions issue: fetch streaming upload mozilla/standards-positions#663

Bugs tracking this feature

• WebKit Bugzilla:
• Radar:

Anything else we need to know

Streaming upload allows web developers to attach a ReadableStream to a Request. The web developer can pass the Request to fetch() to upload the ReadableStream.

[jakearchibald]

I'd like developers have a way to synchronously feature detect this feature. My current recommendation is:

const supportsRequestStreams = (() => {
  let duplexAccessed = false;

  const hasContentType = new Request('', {
    body: new ReadableStream(),
    method: 'POST',
    get duplex() {
      duplexAccessed = true;
      return 'half';
    },
  }).headers.has('Content-Type');

  return duplexAccessed && !hasContentType;
})();

This tests that:

• duplex is implemented as part of RequestInit
• body can be a ReadableStream. Otherwise, it's converted to a string and gets a text/plain content type.

Safari has already implemented streams as bodies of Request, but it rejects if this is passed to fetch().

Could Safari avoid implementing duplex until streams as request bodies are fully implemented (including the fetch parts)? Otherwise the feature detect will get messy and asynchronous.

duplex is only currently useful with this feature.

[annevk]

@jakearchibald now that Yutaka left Chrome, is there someone else picking this up? I suppose it has shipped already, but it seems that duplex should probably also be added to the Request object as per whatwg/fetch#1486.

I tend to agree with you with regards to feature detection, does WPT verify that by chance?

[jakearchibald]

I'll pick this up, and whatwg/fetch#1486

What's the right way to write a WPT for this? Something like:

try {
  await fetchWithStreamBody();
  return;
} catch (err) {
  assert(!supportsRequestStreams);
}

[annevk]

Yeah, that looks correct. I suppose we'll only need this test while engines are still implementing this so maybe add a comment in case someone wonders what's up in the future.

[jakearchibald]

PR up: web-platform-tests/wpt#36048

[annevk]

Colleagues and I have looked at this and are supportive. If there is no further feedback from the WebKit community by October 10 I suggest we label this as "position: support".

[HazhoHuman]

Hey awesome folks;
any updates about this, what is the destiny of it?
can we safely implement it without worrying about backward compatibility if major changes happen later??

[annevk]

@HazhoHuman I'm afraid I don't understand the question. Could you provide more context?

[hawkinsw]

Hello -- I am terribly sorry that my comment was unhelpful. I was attempting to respond to @jakearchibald 's awesome note above. Again, I sincerely apologize for the noise.

[hober]

Closing as we've identified our position.

[jtbandes]

Hi, just wondering if there is another issue or bugzilla ticket tracking the implementation of this feature? I was not able to find anything in a quick search.