Update the “Link type "serviceworker"” section #1110

sideshowbarker · 2017-04-09T02:16:07Z

Add new section “Declaring a "serviceworker" Link header”
Drop normative requirement to fire an error if link[workertype] value is
not a valid worker type; instead just abort without firing an error.
Drop the partial interface IDL for the HTMLLinkElement interface (the
definitions for the IDL attributes have already been merged directly
into the HTMLLinkElement interface IDL definition in the HTML spec).

sideshowbarker · 2017-04-09T02:21:43Z

This needs to wait for #1107 (Changing useCache boolean to updateViaCache enum) and needs a further tweak/update if/when whatwg/html#2517 is merged.

sideshowbarker · 2017-04-09T02:33:11Z

I am wondering if there’s an inconsistency in behavior here that we should fix.

Specifically, in this PR we’re dropping the following normative requirement:

If workerType is not a valid WorkerType value, queue a task to fire an event named error at the link element, and abort these steps.

That’s for the case of the link element with an a workertype attribute whose value is not a valid WorkerType value. But in contrast, in the case of Link header with a workertype attribute whose value is not a valid WorkerType value, the spec still says this:

Let workerType be the "workertype" target attribute of the Link header, or "classic" if no such attribute is present.
If workerType is not a valid WorkerType value, abort these steps.

So we’re aborting when the Link header provides a workertype value that’s not valid, but we’re not aborting in when a link element provides a workertype value that’s not valid.

Is that an inconsistency in behavior that we should fix?

domenic · 2017-04-09T03:14:04Z

Is that an inconsistency in behavior that we should fix?

I think we could go either way. It's very common for HTML to have an invalid value default for enumerated attributes; I can't think of any that fall back to an error behavior. But I'm not sure about HTTP link parameters. Maybe @mnot can lend us his HTTP knowledge.

Re-stating things: I think an inconsistency is fine if each behavior makes sense within its own domain (HTML vs. HTTP).

domenic · 2017-04-09T03:15:23Z

Wait, no, now I am not so sure, because I started comparing it to HTML on another dimension: script type="" attributes. Those are no-ops for unknown types. (They are not errors though.) Maybe we should be consistent with that...

ylafon · 2017-04-09T05:50:35Z

Marked as non-substantive for IPR from ash-nazg.

jungkees · 2017-04-27T10:47:47Z

@mkruisselbrink, do you have bandwidth to review this PR?

jungkees · 2017-06-07T06:20:13Z

docs/index.bs

@@ -1778,8 +1803,7 @@ spec: webappsec-referrer-policy; urlPrefix: https://w3c.github.io/webappsec-refe
      1. Let |scriptURL| be the result of <a lt="URL parser">parsing</a> the <{link/href}> attribute with the <{link}> element's <a>node document</a>'s <a>document base URL</a>.
      1. Let |scopeURL| be null.
      1. If the <{link/scope}> attribute is present, set |scopeURL| to the result of <a lt="URL parser">parsing</a> the <{link/scope}> attribute with the <{link}> element's <a>node document</a>'s <a>document base URL</a>.
-      1. Let |workerType| be the <{link/workertype}> attribute, or "<code>classic</code>" if the <{link/workertype}> attribute is omitted.
-      1. If |workerType| is not a valid {{WorkerType}} value, <a>queue a task</a> to <a>fire an event</a> named <code>error</code> at the <{link}> element, and abort these steps.


Removing this condition entirely seems not desirable. Keep it as-is or at least abort would be needed I guess.

jungkees · 2017-06-07T06:24:41Z

@sideshowbarker, both #1107 and whatwg/html#2517 were merged. So, I think we can continue on this PR. It looks good with needing to have a decision on #1110 (comment).

/cc @domenic, @mnot

sideshowbarker · 2017-06-08T00:47:44Z

both #1107 and whatwg/html#2517 were merged. So, I think we can continue on this PR. It looks good with needing to have a decision on #1110 (comment)

OK, great—so @domenic, what do we want to do for #1110 (comment)?

domenic · 2017-06-08T00:52:05Z

Hmm, I did seem to forget that.

I have started to think that the best thing to do is to no-op on invalid worker types. This means we need another HTML spec tweak (maybe update whatwg/html#2740) to add a third state, invalid, and make that the invalid value default. Then this spec can no-op when it encounters the invalid state.

domenic · 2017-06-08T14:48:34Z

whatwg/html#2740 is updated so you can now count on the attribute being in one of three states: classic, module, or invalid.

(At least, once that PR is merged.)

sideshowbarker · 2017-06-10T22:25:49Z

-      1. Let |workerType| be the <{link/workertype}> attribute, or "<code>classic</code>" if the <{link/workertype}> attribute is omitted.
-      1. If |workerType| is not a valid {{WorkerType}} value, <a>queue a task</a> to <a>fire an event</a> named <code>error</code> at the <{link}> element, and abort these steps.
Removing this condition entirely seems not desirable. Keep it as-is or at least abort would be needed I guess.
I have started to think that the best thing to do is to no-op on invalid worker types. This means we need another HTML spec tweak (maybe update whatwg/html#2740) to add a third state, invalid, and make that the invalid value default. Then this spec can no-op when it encounters the invalid state.
whatwg/html#2740 is updated so you can now count on the attribute being in one of three states: classic, module, or invalid.

OK, see 18e52c5 for my attempt at hooking into that

domenic

HTML parts LGTM. I guess it needs a rebase though.

sideshowbarker · 2017-06-12T03:41:01Z

@jungkees after you (re)review this I’ll rebase so it’ll be ready to merge

jungkees · 2017-06-12T07:00:29Z

docs/index.bs

+    A <a>serviceworker link</a> can be declared using a `Link` header</a> [[!RFC5988]] with "`serviceworker`" as the value of the "`rel`" parameter and a [=service worker/script url=] as the <a>target IRI</a>, and the following optional <a>target attributes</a>:
+
+    : `scope`
+    :: Value: A [=service worker registration/scope url=].


I'm not sure but would like to check if some serialized url should be used for this? scope url is a URL record.

Can @annevk help?

@sideshowbarker, checking https://tools.ietf.org/html/rfc5988#section-5, I think we can just say ":: Value: A [=service worker registration/scope url=] as a URI-Reference" or something like that. WDYT?

And for the script url above, "a [=service worker/script url=] as a URI-Reference inside angle brackets ("<>")" seems clearer I guess. Please correct me if I'm wrong.

I think it's better not to reference obsoleted concepts like URI. Instead I'd do ", serialized", linking to https://url.spec.whatwg.org/#concept-url-serializer (so I think that's [=URL serializer|serialized=]).

jungkees · 2017-06-12T07:02:35Z

@sideshowbarker, LGTM with checking #1110 (comment). Thanks!

* Add new section “Declaring a "serviceworker" Link header” * Drop normative requirement to fire an error if link[workertype] value isn’t a valid worker type; instead just abort. * Drop the partial interface IDL for the HTMLLinkElement interface (the definitions for the IDL attributes have already been merged directly into the HTMLLinkElement interface IDL definition in the HTML spec). Fixes #1073

mnot · 2017-06-14T14:07:30Z

Sorry, missed that. There aren't any special considerations for Link header attributes regarding defaults; if you want to define them, and make them different in different places, that's up to you.

jungkees · 2017-06-23T06:59:37Z

@sideshowbarker, could you address #1110 (comment) to merge?

sideshowbarker · 2017-06-24T08:35:31Z

@sideshowbarker, could you address #1110 (comment) to merge?

Done in f3ffc85

Sorry for having taken so long to get back to this

jungkees · 2017-06-24T11:53:22Z

Merged. Thanks a lot @sideshowbarker!

sideshowbarker force-pushed the sideshowbarker/link-rel-serviceworker branch 2 times, most recently from 959067b to c09d3bd Compare April 11, 2017 12:39

jungkees reviewed Jun 7, 2017

View reviewed changes

domenic mentioned this pull request Jun 8, 2017

Tweak <link>.workerType reflection whatwg/html#2740

Merged

domenic approved these changes Jun 11, 2017

View reviewed changes

jungkees reviewed Jun 12, 2017

View reviewed changes

sideshowbarker force-pushed the sideshowbarker/link-rel-serviceworker branch from a77b22f to 79437cd Compare June 12, 2017 08:57

jungkees mentioned this pull request Jun 14, 2017

Switch useCache to an enum #1104

Closed

Reference URL serializer for URLs in Link header

f3ffc85

sideshowbarker force-pushed the sideshowbarker/link-rel-serviceworker branch from 555acd3 to f3ffc85 Compare June 24, 2017 08:37

jungkees merged commit f8e8103 into master Jun 24, 2017

sideshowbarker deleted the sideshowbarker/link-rel-serviceworker branch June 24, 2017 11:53

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update the “Link type "serviceworker"” section #1110

Update the “Link type "serviceworker"” section #1110

sideshowbarker commented Apr 9, 2017 •

edited

Loading

sideshowbarker commented Apr 9, 2017

sideshowbarker commented Apr 9, 2017

domenic commented Apr 9, 2017

domenic commented Apr 9, 2017

ylafon commented Apr 9, 2017

jungkees commented Apr 27, 2017

jungkees Jun 7, 2017

jungkees commented Jun 7, 2017

sideshowbarker commented Jun 8, 2017

domenic commented Jun 8, 2017

domenic commented Jun 8, 2017 •

edited

Loading

sideshowbarker commented Jun 10, 2017

domenic left a comment •

edited

Loading

sideshowbarker commented Jun 12, 2017

jungkees Jun 12, 2017

jungkees Jun 14, 2017

domenic Jun 14, 2017

jungkees commented Jun 12, 2017

mnot commented Jun 14, 2017

jungkees commented Jun 23, 2017

sideshowbarker commented Jun 24, 2017 •

edited

Loading

jungkees commented Jun 24, 2017

Update the “Link type "serviceworker"” section #1110

Update the “Link type "serviceworker"” section #1110

Conversation

sideshowbarker commented Apr 9, 2017 • edited Loading

sideshowbarker commented Apr 9, 2017

sideshowbarker commented Apr 9, 2017

domenic commented Apr 9, 2017

domenic commented Apr 9, 2017

ylafon commented Apr 9, 2017

jungkees commented Apr 27, 2017

jungkees Jun 7, 2017

Choose a reason for hiding this comment

jungkees commented Jun 7, 2017

sideshowbarker commented Jun 8, 2017

domenic commented Jun 8, 2017

domenic commented Jun 8, 2017 • edited Loading

sideshowbarker commented Jun 10, 2017

domenic left a comment • edited Loading

Choose a reason for hiding this comment

sideshowbarker commented Jun 12, 2017

jungkees Jun 12, 2017

Choose a reason for hiding this comment

jungkees Jun 14, 2017

Choose a reason for hiding this comment

domenic Jun 14, 2017

Choose a reason for hiding this comment

jungkees commented Jun 12, 2017

mnot commented Jun 14, 2017

jungkees commented Jun 23, 2017

sideshowbarker commented Jun 24, 2017 • edited Loading

jungkees commented Jun 24, 2017

sideshowbarker commented Apr 9, 2017 •

edited

Loading

domenic commented Jun 8, 2017 •

edited

Loading

domenic left a comment •

edited

Loading

sideshowbarker commented Jun 24, 2017 •

edited

Loading