freenode :: #whatwg

15 Feb 2017
00:08jugglinmikeHello all. I have a question about Service Workers, if anyone here has a moment.
00:08jugglinmikeIf I understand correctly, it seems like the allocation of distinct task queues is implementation-defined
00:09jugglinmikeSo does that mean that one browser could implement postMessage and fetch tasks on the same queue, but another could define separate task queues for each?
00:26Mekjugglinmike: the spec defines separate task sources for fetch events and all other events, and additionally allows all those other event types to have their own task source per event type, yes. It is then also up to user agents how to map task sources to task queues (i.e. all events from the same task source end up in the same queue, but different sources might or might not share queues)
00:55Mekhuh, http://output.jsbin.com/gasinavezo for some reason seems to output "true, false, true" in firefox, while unless I'm crazy and don't understand javascript all three checks should be identical...
00:56jugglinmikethank you Mek!
00:59Mekor wait, maybe in my case the problem is that the about:blank iframe hasn't finished loading yet in the first assignment, so my bug rather than soms craziness
05:05annevkMek: within an event loop you can have multiple sources, but only one task loop
05:05annevkMek: you can just give priority to a source
07:12MikeSmithannevk: so I spend my entire Stackoverflow time budget today on trying to answer one question:
07:12MikeSmithhttp://stackoverflow.com/questions/42239643/when-does-firefox-set-the-origin-header-to-null-in-post-requests/42242802#42242802
07:13MikeSmithwhen you have time, as usual if I got anything wrong please either lemme know or comment/answer there
07:14MikeSmithI wrote all that without answering the one question the OP cares about most, which is I guess, Why Firefox doesnt send Origin for cross-origin POSTs
07:15MikeSmithwhich I dont know why Firefox does that, but if it does, its non-conforming, right?
07:22annevkKnown bug in Fx
07:23annevkWill look when I get back
08:10annevkzcorpan: hmm https://bugzilla.mozilla.org/show_bug.cgi?id=1338535
08:11zcorpanannevk: you get to decide if mozilla is interested in solving this problem? :-)
08:14zcorpanhttps://bugs.chromium.org/p/chromium/issues/detail?id=690927 is the issue for chromium
08:14annevkzcorpan: well, I'm not, not sure about Mozilla
08:14annevkzcorpan: but yeah, seems like my manager thought that would be a good idea
08:15annevkzcorpan: disabled does submit right? Why can't disabled be used everywhere?
08:16zcorpanannevk: it doesn't submit
08:16annevkzcorpan: oh okay, disabled means you can't copy I guess
08:16annevkIt's been like a decade since I looked at forms closely, I know nothing
08:17zcorpanannevk: disabled can be used for all controls, but not readonly
08:18annevkMikeSmith: your answer is slightly wrong
08:18annevkMikeSmith: we changed when Origin is required to be transmitted recently
08:19annevkMikeSmith: to be either for cross-origin CORS (not same-origin) or requests whose method is not GET/HEAD
08:19annevkMikeSmith: so same-origin POST used to be excluded, but is no longer
08:19annevkMikeSmith: cross-origin POST from a <form> (without CORS) used to be excluded, but is no longer
08:20annevkMikeSmith: therefore the behavior in Fx is considered a bug
08:23MikeSmithannevk: OK I thought my answer had the current state of things right at least, but I you mean I should update the answer to indicate that the behavior changed recently? (and note what the previous required behavior was)
08:24annevkMikeSmith: up to you I suppose, but see https://github.com/whatwg/fetch/commit/eb89fcd54bb39e81b11c569f6ad7ba615883f7b9 for the change
08:25MikeSmithOK, I guess this is another case where browsers have not caught up with the spec change yet?
08:32annevkMikeSmith: well, Chrome always implemented that behavior and WebKit too, but it never got into Firefox
08:32annevkMikeSmith: it&#39;s something Adam Barth pushed but it never really got standardized/accepted
08:32annevkAnd I kinda dislike it since it no longer makes CORS requests distinct
08:35MikeSmithyeah
08:36MikeSmithits also more confusing for authors/developers
08:37MikeSmithanyway I updated my SO answer:
08:37othermaciejannevk, zcorpan: for non-text controls like checkboxes or radio buttons, it&#39;s confusing to have it be read-only but not have the disabled appearance, and is generally not done in native toolkits. So original idea of extending readonly was bad. The submitanyway idea isn&#39;t broken u-wise but it seems super marginal, based on the stated use case. It&#39;s basically a conenience for something really rare.
08:37MikeSmithNote: The above describes how the Fetch spec currently defines the requirements, due to a change that was made to the spec on 2016-12-09. Up until then the requirements were different:
08:37MikeSmithpreviously no Origin was sent for a same-origin POST
08:37MikeSmithpreviously no Origin was sent for cross-origin POST from a <form> (without CORS)
08:37MikeSmithSo I think the Firefox behavior described in the question conforms to what the spec previously required, but not what the spec currently requires.
08:37MikeSmithhey othermaciej!
08:38othermaciejAlso don&#39;t like the guy saying &quot;WHATWG agrees to standardize&quot;, that doesn&#39;t seem a fair representation of discussion of the issue
08:38othermaciejHi MikeSmith!
08:39zcorpanI would agree to standardize *if* 2+ browsers want to implement.
08:40othermaciejyeah, I think a conditional statement like that would be justified, but that&#39;s not what his bugs say. They make it sound like ~WHATWG has already decided, so get on board~
08:40annevkothermaciej: thanks, that all sounds reasonable
08:41annevkothermaciej: in the Bugzilla issue I asked for evidence that the request is widespread, we&#39;ll see I suppose
08:41annevkBut I tend to agree that syncing with a hidden input is easy enough
10:40jochen__how&#39;d I go about getting https://github.com/w3c/wpt-tools/pull/164 merged?
10:44annevkjochen__: prolly talk to jgraham
10:49Ms2gerLook at the history for that runner and find someone who uses it?
10:53mkwst_MikeSmith has the most commits/reviews. But they&#39;re mostly from 2014. *cough* :)
11:05jgrahamjochen__: Merged, but really if your workflow involves the runner you should find a better workflow :)
11:36zcorpanannevk: thanks for fixing snapshots for quirks mode
11:37annevkDoes anyone know if there is a file system where the directory name can be empty?
11:39jgrahamannevk: Sounds implausible
13:28jochen__annevk: mind looking at the document.open PRs again (html/wpt)?
13:56annevkjochen__: are we closer to figuring out what the security model should be?
13:56annevkI&#39;ll have a look in a bit
14:22annevkzcorpan: so there are a lot of people asking about readonly-like semantics and the top answer is a bunch of JavaScript
14:22annevkzcorpan: top answer having like >300 votes, so maybe it&#39;s something worth considering
14:22zcorpanannevk: got a link?
14:23annevkhttp://stackoverflow.com/questions/368813/html-form-readonly-select-tag-input
14:23annevkhttp://stackoverflow.com/questions/37762671/how-to-create-readonly-select-box-but-i-will-post-that-value
14:23annevkhttp://stackoverflow.com/questions/155291/can-html-checkboxes-be-set-to-readonly
14:23annevkhttp://stackoverflow.com/questions/4727974/how-to-post-submit-an-input-checkbox-that-is-disabled
14:23annevkThey&#39;re from OP in the Firefox bug
15:39jochen__annevk: basically we all agree that we disagree with the one that is in the spec :)
15:40annevkjochen__: if you can guarantee an issue will be filed once you and bz are ready I&#39;ll go ahead and merge all the things
15:41jochen__let&#39;s file one immediately
15:41annevkI guess I&#39;ll remember it too since I don&#39;t like security mismatches
15:41annevkThanks, I&#39;ll go merging things then
15:43jochen__https://github.com/whatwg/html/issues/2360
15:45jochen__what&#39;s the bar for being allowed to merge stuff to wpt btw?
15:45jochen__or in other words, can I haz access plz?
15:47Ms2gerBeing trusted to review sufficiently well :)
15:47Ms2gerHave you done reviews?
15:51jochen__i reviewed the referrer policy stuff
15:51jochen__(/referrer-policy/)
15:52jochen__but happy to just review stuff for a bit more and come back later
16:09wanderviewannevk: was there some question about ancestorOrigins sticking around?
16:09annevkwanderview: there&#39;s a question of how it should work
16:10annevkwanderview: https://github.com/whatwg/html/issues/1918
16:11wanderviewannevk: I guess this blocks Client.ancestorOrigins as well then
16:11wanderviewin service worker pec
16:11wanderviewspec
16:11annevkwanderview: for us it does, yes
16:11annevkwanderview: it would be bad to undermine bz
16:14annevkDomenic: the numbers in https://github.com/w3c/web-platform-tests/issues/3018 seem quite good
16:14annevkDomenic: tweet?
16:14Domenicannevk: yeah :)
16:18annevkDomenic: I think I found a bug just browsing that coverage report
16:19DomenicOh neat! In the file stuff I&#39;d guess? The fact there were several uncovered branches there looked suspicious
16:19annevkDomenic: https://url.spec.whatwg.org/#file-host-state shouldn&#39;t have to check for &quot;includes credentials&quot; as that&#39;s not possible for file URLs
16:19annevkDomenic: yeah, that also looked very suspect, guess I&#39;ll dive into this tomorrow
16:20annevkDomenic: tooling is amazing
16:23wanderviewannevk: JakeA: how important is it to provide Client.ancestorOrigins if we remove Client.frameType?
16:23wanderviewdo I have to implement .ancestorOrigins when .frameType is removed to avoid breaking a critical use case?
16:26DomenicMikeSmith: botie is gone again
16:48JakeAwanderview: it isn&#39;t something I&#39;ve heard a lot of requests for
16:50wanderviewJakeA: I guess there is this: https://github.com/w3c/ServiceWorker/issues/1007
17:17zcorpanDomenic: can you have another look at the new changes in https://github.com/whatwg/html/pull/2355 (and the test)?
17:31tobieannevk: do all of the whatwg specs have auto pr branch deploys?
17:32annevktobie: pretty much
17:33MikeSmithDomenic: botie is basic
17:42tobieannevk: my code is assuming that is the case. So we need to somehow special case it that&#39;s not so.
17:53annevktobie: I don&#39;t understand, we simply only make it work for specs that have that setup
17:54annevktobie: in the end all specs should have that setup, those that don&#39;t are buggy
17:54tobieannevk: wfm
18:01tobieannevk: OK so you be able to go ahead and add repos now
18:01tobieannevk: if you ping me with the repo name, I can run a script to update all open PRs
18:02tobieannevk: it&#39;s too complicated to automate this for now, due to a GH API issue.
18:04annevkDomenic: so if you&#39;re not tweeting I can
18:05annevktobie: let&#39;s do it tomorrow, okay?
18:05tobieannevk: sure
18:05tobieannevk: ping me
18:05annevkwill do
18:05tobieannevk: (might be out of office tomorrow. we&#39;ll see)
18:07annevktobie: Friday or next Monday is fine too, will ping and we&#39;ll see
18:08Domenicannevk: oh, sure, I thought that was the plan
18:15annevkDomenic: done
18:28smaugno zcorpan
18:28smaughmm, perhaps foolip knows
18:28smaugabout MouseEvent.offsetX/Y
18:29smaugwebkit based engines seem to cache the value after first getter call
18:29smaugyet the spec doesn&#39;t hint anything like that
20:26wanderviewannevk: if I want to make a PR to service worker spec... is there a build step I need to run?
20:27annevkwanderview: not per se, but if you have bikeshed you could run that
20:27wanderviewannevk: I modify the .bs file and then run bikeshed?
20:28annevkwanderview: not really sure what their policies are though
20:28annevkwanderview: yeah
20:28annevkwanderview: if they don&#39;t have PR instructions that is worthy of an issue
20:44wanderviewannevk: I guess the CONTRIBUTING.md file does mention it in there
20:58wanderviewJakeA: did you forget to re-generate the index.html after you removed the skip-service-worker flag? when I run bikeshed on the pristine checkout of the repo it seems to make those changes...
20:59JakeAHmm, will check
21:19wanderviewJakeA: I guess I get this: LINK ERROR: No &#39;dfn&#39; refs found for &#39;skip-service-worker flag&#39;.
21:19wanderviewmaybe just the fetch side of the change landed?
21:46* wanderview made his first pull request against the ServiceWorker spec.
23:08MikeSmithwanderview: nice PR
23:49foolipsmaug: is this caching that can be observed?
23:51foolipsmaug: at a glance it looks like it depends on things that can&#39;t be changed with any init*Event(), but maybe I&#39;m missing something :)
23:51smaugfoolip: caching can be observed
23:52smaugfoolip: https://bugzilla.mozilla.org/show_bug.cgi?id=1339758#c11
23:53foolipsmaug: oh, so if the layout is invalidated by the event handler, the event&#39;s offsetX/Y isn&#39;t?
23:54smaugright, in Webkit and Blink at least
23:54smaugand Edge too, but it has apparently different behavior
23:54smaugmy interpretation is that Gecko works per spec, but bz thinks Edge might
23:55smaugbut webkit/blink are certainly wrong
23:55foolipsmaug: did offsetX go and disappear spec side? It used to be in https://w3c.github.io/uievents/#interface-MouseEvent I think
23:55smaugfoolip: it is not there
23:55smaugin CSS OM
23:56fooliphttps://drafts.csswg.org/cssom-view/#extensions-to-the-mouseevent-interface it is
23:57smaugso, do you happen to know the reason for caching? Sounds like no :) And I think we should figure out what actually should happen here
23:58smaugspecing webkit/blink behavior would be sad, since the caching happens rather random time
23:58foolipsmaug: no, this is the first time I&#39;ve seen it, but presumably because of perf at some point. I would ask dtapuska to investigate and explain
23:58smaug(whoever happens to access the getter first time)
23:59foolipwhat is https://drafts.csswg.org/cssom-view/#dom-mouseevent-offsetx trying to say though? &quot;position where the event occurred&quot; in past tense?
16 Feb 2017
No messages
   
Last message: 6 days ago