Is a XHR.getAllResponseHeaders change need any intent process?

[Yutaka Hirano]

Hi,

Raphael is proposing a CL which changes XHR.getAllResponseHeaders behavior. With the CL, all header names will be lower-cased. It's a developer-facing change but small, do we need to send any intent (to ship? to change?) for this change?

The change is based on a spec change.

Thanks,

[Takeshi Yoshino]

The example in the XHR spec is not yet updated. See

- https://github.com/w3c/web-platform-tests/blob/master/XMLHttpRequest/getallresponseheaders.htm

- https://github.com/w3c/web-platform-tests/blob/master/XMLHttpRequest/resources/headers.asis

for examples.

Here's WebKit's opinion on this: https://bugs.webkit.org/show_bug.cgi?id=169286#c7

Takeshi

[Rick Byers]

The questions I ask myself to try to decide between bug fix / "trivial" vs. platform change are:

• Is there a spec and web-platform-test change involved?
• Is this something this is significant enough that we'd update MDN docs for?
• Is it reasonable to expect that some (non-trivial number of) web pages might be broken by this change?

I haven't looked deeply into the details here, but big picture we trust blink contributors and their code reviewers to use good judgement here, and it's OK to sometimes get things on the edges wrong in either direction (there's always going to be bug fixes with unintended compat consequences).

[Mike West]

I'd suggest erring on the side of sending an intent. It doesn't take much more effort than sending a question about whether to send an intent. :)

For this change in particular, it's a developer-visible change in behavior that you could imagine sufficiently brittle code relying on: `x == "Content-Type"` will fail tomorrow when it didn't fail today. I think lowercasing is a reasonable change to make, but it's something we'd also want to tell developers about in one way or another, and evaluate whether other vendors will hop on board.

-mike