27482 2014-12-02 13:10:51 +0000 Have appcache be bypassed when the navigation is under the scope of a service worker 2015-09-29 16:21:54 +0000 1 1 1 Unclassified WHATWG HTML unspecified Other other RESOLVED MOVED P2 normal Unsorted 1 jungkees jungkees d ian mike contributor oldest_to_newest 115784 0 jungkees 2014-12-02 13:10:51 +0000 Service worker folks have had a discussion about what the UA should do when both a service worker registration and an application cache are available for the navigation of resources: https://github.com/slightlyoff/ServiceWorker/issues/275 Particularly for the following requirement: > If the navigation is under the scope of a SW, appcache is completely bypassed, regardless of whether the navigation uses the SW. I presume the following condition in brackets added to the otherwise-branch of the below algorithm will satisfy that: https://html.spec.whatwg.org/multipage/syntax.html#the-before-html-insertion-mode - A start tag whose tag name is "html" "; otherwise, if there is no such attribute, or its value is the empty string, or resolving its value fails, [or the result of running Match Scope algorithm passing the Document's address (url) is non-null], run the application cache selection algorithm with no manifest." Match Scope returns a matching service worker registration taking a url as an argument, and is defined here: https://slightlyoff.github.io/ServiceWorker/spec/service_worker/index.html#scope-match-algorithm Could you comment on it? 115851 1 ian 2014-12-03 19:56:00 +0000 Seems fine to me in principle. Could we rename "match scope" to something that explicitly references Service Workers, so it's more clear what it means? Also, why would we run the match scope algorithm, rather than just checking whether there's an assigned service worker? 115900 2 jungkees 2014-12-04 02:51:15 +0000 (In reply to Ian 'Hixie' Hickson from comment #1) > Seems fine to me in principle. > > Could we rename "match scope" to something that explicitly references > Service Workers, so it's more clear what it means? > Yes, I'll rename it to "match service worker registration". Actually, a service worker registration is keyed by scope, so it not only clarify the name to the external caller context but also explains what it exactly does. > Also, why would we run the match scope algorithm, rather than just checking > whether there's an assigned service worker? The "match service worker registration"algorithm given a url actually checks whether there exists a service worker registration that can cover the url space. The part that makes this matching registration's active service worker be the service worker controlling the given client is defined elsewhere (in handle fetch algorithm). 115906 3 jungkees 2014-12-04 04:13:39 +0000 (In reply to Jungkee Song from comment #2) > (In reply to Ian 'Hixie' Hickson from comment #1) > > Seems fine to me in principle. > > > > Could we rename "match scope" to something that explicitly references > > Service Workers, so it's more clear what it means? > > > > Yes, I'll rename it to "match service worker registration". Actually, a > service worker registration is keyed by scope, so it not only clarify the > name to the external caller context but also explains what it exactly does. > Addressed: https://github.com/slightlyoff/ServiceWorker/commit/2b123e8177a2dd1e56a41c0fe2ea29c8fb97d731 123052 4 d 2015-09-02 19:03:41 +0000 Assigning to Jungkee since in https://www.w3.org/Bugs/Public/show_bug.cgi?id=28737#c3 he said he'd be willing to make a PR. 123356 5 d 2015-09-29 16:21:54 +0000 https://github.com/whatwg/html/pull/184