Allow integrity attribute on rel=preload links

[domfarolino]

This PR "allows" the integrity attribute to be used by <link rel=preload>s. Previously, the integrity attribute would only be considered for <link rel=stylesheet>s, by spec. Link preloads, in this specification, use the default fetch and process the linked resource algorithm, so this should be the only change left.

I am fine with adding support to modulepreload too, but I think that should be a separate PR, since this is tightly coupled with the consensus we have from the preload folks.

Corresponding w3c/preload PR & issue: w3c/preload#127 & w3c/preload#137

• At least two implementers are interested (and none opposed):
  • Multi-vendor agreement/resolve at TPAC 2018 on this spec change can be found here
• [Tests] ...incoming...
• Bugs:
  • Chrome: https://crbug.com/677022
  • Firefox: https://bugzilla.mozilla.org/show_bug.cgi?id=1222633 (this this just extends the preload feature, I think putting the generic preload bugzilla bug here is sufficient)
  • Safari: (Currently handles this behavior correctly, as the WPTs will show)

/cc @kinu @nyaxt @yutakahirano

---

/semantics.html ( diff )

[domenic]

This is a document conformance change, so I don't think we usually require tests for this (although @sideshowbarker has been maintaining conformance checker tests). That said, if there are no tests for integrity="" on rel=preload yet, then we should indeed do those ASAP...

[annevk]

@domfarolino so yeah, looking at the processing model @domenic is correct and HTML already enforces integrity for everything that uses "default fetch and process the linked resource", probably including modulepreload? It would be good to have that tested.

[domfarolino]

probably including modulepreload? It would be good to have that tested.

Ah, actually modulepreload has its own overload of the "default fetch and process the linked resource" algorithm, but it also honors the integrity attribute as well, so I can include tests for that too I suppose. I'll also just update this PR to allow integrity for rel=modulepreload as well for document conformance, as it's simple enough. Thanks.

[domfarolino]

What are the next steps for getting this merged? Tests are WIP (branch here), and will land before Chrome's implementation, but not currently a blocker for the spec as Domenic and Anne have mentioned. The change seems pretty non-controversial. Currently it is technically blocking Blink's Intent to Implement and Ship, so I'm wondering if there's anything else I can do here?

[domfarolino]

Ah, just realized I never made the modulepreload change I mentioned :) Done!

[domenic]

I think we were being a little over-strict by using your document conformance change pull request as an opportunity to ask you to do processing-model tests. I agree processing-model tests should not be a document-conformance-PR blocker. But, hey, you did the tests anyway :). The WIP branch is good enough for me.

I'll merge this now!

[sideshowbarker]

heads-up @whatwg/documentation

[chrisdavidmills]

Documentation need recorded at https://trello.com/c/6YAMGL3B/132-html-link-element-fx-68