Add site and obtain a site #5354
Conversation
source
Outdated
| @@ -91733,11 +91733,11 @@ import "https://example.com/foo/../module2.mjs";</code></pre> | |||
| <p>A <dfn data-export="">scheme-and-registrable-domain</dfn> is a <span>tuple</span> of a <span | |||
| data-x="concept-url-scheme">scheme</span> and a <span data-x="concept-domain">domain</span>.</p> | |||
|
|
|||
| <p>An <dfn data-export="">agent cluster key</dfn> is an <span>origin</span> or a | |||
There was a problem hiding this comment.
It also wasn't clear to me why we exported agent cluster key. I didn't investigate though.
|
We should either remove "obtain an agent cluster key" or change it to take a BCG too, and move the "if cross-origin isolated, return origin" logic into there. |
|
That seems like a reasonable thing to do. That leaves the question of whether to distinguish site and agent cluster key more. We probably do not want to combine them as that will give the same kind of ambiguities. Should I just spell it out more that if a site is an origin it's either because it's opaque or its registrable domain is null? Whereas for an agent cluster that doesn't hold. |
|
That sounds like a pretty good clarification, yeah. |
It's useful for whatwg/fetch#943 to be able to distinguish between a site and agent cluster key as #4734 changes how agent clusters can be keyed.
annevk
force-pushed
the
annevk/expose-obtain-a-site
branch
from
6be1052 to
abbf781
Compare
|
I think this is ready now and would allow Fetch to build on it. And once merged I could rebase the other PR to take advantage of it as discussed. |
|
|
||
| <p class="XXX"><a href="https://github.com/whatwg/html/pull/4734">whatwg/html #4734</a> is | ||
| expected to widen <span>agent cluster key</span> (but not <span>site</span>) to encompass all | ||
| <span data-x="concept-origin-tuple">tuple origins</span>.</p> |
There was a problem hiding this comment.
Maybe move this one paragraph down so its clearer that it applies to both of the currently-trivial definitions? Not sure though.
| @@ -91744,11 +91744,13 @@ import "https://example.com/foo/../module2.mjs";</code></pre> | |||
| <p>A <dfn data-export="">scheme-and-registrable-domain</dfn> is a <span>tuple</span> of a <span | |||
| data-x="concept-url-scheme">scheme</span> and a <span data-x="concept-domain">domain</span>.</p> | |||
|
|
|||
| <p>An <dfn data-export="">agent cluster key</dfn> is an <span>origin</span> or a | |||
| <p>A <dfn data-export="">site</dfn> is an <span data-x="concept-origin-opaque">opaque | |||
There was a problem hiding this comment.
Per https://dictionary.cambridge.org/grammar/british-grammar/either that's for two choices.
|
I'm going to merge this, but I'm happy to revisit the editorial discussion in the follow-up PR. |
It's useful for whatwg/fetch#943 to be able to distinguish between a site and agent cluster key as #4734 changes how agent clusters can be keyed, but that should not impact the key used for caches.
However, #4734 does not change this algorithm directly. Perhaps we should remove "obtain an agent cluster key" entirely? (The other weird thing is that the key type doesn't really change, except insofar that the origin can have a host that has a non-null registrable domain.)
Thoughts?
cc @shivanigithub @domenic
/webappapis.html ( diff )