Switch useCache to an enum

[jakearchibald]

We're worried about fetch all importScripts on every navigation. It sounds really heavy and a waste of bandwidth.

New proposal. useCache values:

• "never"
• "imports"
• "all"

It's a bit late in the day, but we all think this is a much safer change.

[wanderview]

Gecko bug to avoid shipping boolean useCache: https://bugzilla.mozilla.org/show_bug.cgi?id=1353638

Gecko bug to switch to enum: https://bugzilla.mozilla.org/show_bug.cgi?id=1353636

[jakearchibald]

I'm worried it isn't clear that useCache applies to updates, so I'm going to rename it updateViaCache. Objections welcome.

[jakearchibald]

Service worker spec PR #1107

[wanderview]

Just to clarify, is this also doing FB's request to bypass cache on register? The new name just made me wonder.

[jakearchibald]

Nah, I haven't added that. But you're right, the naming makes even more sense considering that.

[jakearchibald]

PR: #1107

[jungkees]

SW change: c92c148. (merged)
HTML change: whatwg/html@b5fcec0. (merged)
Tests: web-platform-tests/wpt#5515. (underway)

[mfalken]

We're confused that the service worker spec and the HTML spec seem to have contradicting definitions of updateViaCache in HTMLLinkElement.
(SW spec)

partial interface HTMLLinkElement {
  ...
  [CEReactions] attribute ServiceWorkerUpdateViaCache updateViaCache;
};

(HTML spec)

interface HTMLLinkElement : HTMLElement {
  ...
  [CEReactions] attribute DOMString updateViaCache;
};

It seems the HTMLLinkElement extension in the service worker spec is not needed since the HTML spec has it.

[jakearchibald]

Ahh yes, we need to remove the one from the service worker spec. The HTML spec is the correct one.

[jungkees]

Yes, it's being addressed in #1110 (comment).

[jakearchibald]

Tests landed web-platform-tests/wpt#5515