00:01
<MikeSmith>
http://stackoverflow.com/questions/33146029/seeing-if-a-request-succeeds-from-within-a-service-worker
20:27
<Domenic>
We should really get around to removing the outline algorithm
20:31
<tantek>
agreed
21:42
<MikeSmith>
Domenic: tantek Why?
21:43
<tantek>
insufficient implementation, having it be in the spec is misleading (affords it implied status that is inaccurate)
21:53
<MikeSmith>
there's actually nothing in the spec for browsers to implement
21:54
<tantek>
the outline algorithm itself presumably
21:54
<MikeSmith>
nothing exposed to web content
21:54
<tantek>
great, then move it to a non-normative appendix
21:54
<tantek>
though I thought it was somehow exposed in the DOM, no?
21:54
<MikeSmith>
nope
21:55
<tantek>
then a non-normative appendix is a good start to moving it out.
21:55
<MikeSmith>
So why not let's actually add a requirement?
21:55
<MikeSmith>
Element.createOutline()
21:55
<tantek>
because no implementer has cared enough to date to ask for it
21:55
<tantek>
features by default should be deleted
21:56
<tantek>
not solutions looking for problems
21:56
<MikeSmith>
or alternately document.createOuline(element)
21:57
<MikeSmith>
implementors are typically not the ones who ask for such features
21:57
<MikeSmith>
And it's not a solution looking for a problem
21:58
<MikeSmith>
The core problem is the fact that we added section and article
21:58
<tantek>
I'm kind of done with those too
21:59
<MikeSmith>
The outline algorithm is a necessary consequences of having done that
21:59
<tantek>
If it were necessary it would have been implemented. So no.
21:59
<MikeSmith>
tantek: well the language is stuck with those elements now
21:59
<tantek>
is it? if no one does anything particularly interesting with them they turn into aliases for div
22:00
<MikeSmith>
sometimes having discussions with you really bums me out man
22:02
<MikeSmith>
https://github.com/whatwg/html/issues/83#issuecomment-136882847
22:02
<MikeSmith>
Anyway I agree with what Hixie said there
22:05
<gsnedders>
MikeSmith: how do a11y tools get at the outline? do they get it from the DOM tree? from the a11y tree-thing?
22:05
<MikeSmith>
and otherwise in general for 2016 l think literally I'm going to seriously try to adopt a communication mode where I try very hard to project a positive vibe and not unintentionally bum people out
22:05
<gsnedders>
I should really understand how the a11y layers work better.
22:06
<MikeSmith>
They don't do anything with it
22:06
<gsnedders>
I thought most a11y stuff had ways to jump between sections?
22:06
<MikeSmith>
They just do a naive outline based on the headings
22:06
<gsnedders>
Or is that just jump between sections?
22:07
<MikeSmith>
naive headings level
22:07
<MikeSmith>
Yeah they provide jumping
22:07
<gsnedders>
So do they have *any* concept of an outline?
22:07
<MikeSmith>
But they do not understand nested h1
22:07
<gsnedders>
Right, that's not what I'm asking.
22:08
<MikeSmith>
no they do not
22:08
<gsnedders>
I'm asking do they have an outline or do they have "jump to next heading in document order"?
22:08
<MikeSmith>
They treat every h1 as a top level heading
22:08
<gsnedders>
Do they have a concept of levels of heading?
22:09
<gsnedders>
If they do, do we not need a conformance class for accessibility tools?
22:09
<MikeSmith>
They create a naive hierarchy based just on h1 to h6
22:10
<MikeSmith>
So you can navigate to/through all the H2 headings etc
22:10
<gsnedders>
Right. So all we're asking them to change (currently) is how they build their hierarchy. OK.
22:10
<MikeSmith>
yeah
22:12
<MikeSmith>
the way I address this problem in the HTML checker is to emit a warning if an h1 is used in a nested section or article
22:12
<gsnedders>
So again: Do we need a new conformance class for accessiblity tools?
22:12
<MikeSmith>
that is, other than at the top level
22:12
<gsnedders>
Because currently there isn't one that applies for stuff like that.
22:13
<MikeSmith>
We already have that conformance class, I think
22:13
<MikeSmith>
But if not, yeah, we should add it
22:13
<MikeSmith>
definitely
22:14
<gsnedders>
I guess half the problem is the mismatch of things exposed via the accessibility APIs on different platforms
22:14
<gsnedders>
and by different rendering engines
22:15
<MikeSmith>
btw I also tried to address this need in the HTML checker by adding a Show Outline feature. which many people do useー just as many did use the Show Outline feature of the legacy W3C markup validator
22:15
<MikeSmith>
gsnedders: yeah
22:17
<MikeSmith>
Steve Faulkner has been working to try to erase that mismatch and get more interoperability for a11y APIs
22:17
<MikeSmith>
he edits a related spec
22:20
<gsnedders>
I guess some of the problem is that much of that is OS level
22:25
<MikeSmith>
yup
22:32
<MikeSmith>
but IMHO another problem is that the a11y tree is not exposed to Web content
22:33
<MikeSmith>
it's one of the cases where the browser is doing something important and useful but Web developers have no way to observe it or examine it from script
22:35
<MikeSmith>
IMHO we need an a11y abstraction above the platform APIs, that is exposed for scripts to observe
22:38
<MikeSmith>
what I mean is, if you look at http://rawgit.com/w3c/html-api-map/master/index.html#el-article
22:39
<MikeSmith>
it shows that, depending on the platform API, <article> gets mapped to some thing called either ROLE_SYSTEM_GROUPING or ROLE_SYSTEM_DOCUMENT or ATK_ROLE_DOCUMENT_FRAME or AXDocumentArticle
22:40
<MikeSmith>
so we should just have one standard abstract defined thing/name that it gets mapped to, and that should be exposed to scripts
22:40
<MikeSmith>
and the tree of such things that browser engines create should be exposed to scripts
22:44
<MikeSmith>
and anyway I think what this shares with this outline-algorithm issue is that in general when you have specs defining things or stating requirements that normal Web developers have no way to observe and test from there Web application code, we're gonna have problems
22:48
<MikeSmith>
in other news http://stackoverflow.com/questions/34473930/service-worker-caching-requests-for-one-time-urls-without-ignoresearch is interesting
22:48
<MikeSmith>
one-cache-per-file SW caching
22:49
<MikeSmith>
"A few gotchas. With this approach, you’ll need to roll your own cache-expiration logic."
23:44
<Domenic>
MikeSmith: I think the problem is that HTML right now has an abstract notion of author-facing semantics which are actively in conflict with user-facing semantics as implemented by a11y tools
23:44
<Domenic>
I'd like to bring them in line
23:44
<Domenic>
Since a11y tools and browsers have stated they will not change, we should change the author-facing abstract semantics