2017-02-01 [17:29:15.0000] what spec actually requires that the Origin header be sent for any request that is not a GET or HEAD request? [17:29:27.0000] I see the Fetch spec says this: [17:29:40.0000] A CORS request is an HTTP request that includes an Origin header. It cannot be reliably identified as participating in the CORS protocol as the Origin header is also included for all requests whose method is neither GET nor HEAD. [17:30:30.0000] …but the Fetch spec itself nowhere requires that Origin be sent in any request that is not a GET or HEAD request [18:04:40.0000] ah nm [18:04:45.0000] found it [18:04:52.0000] https://fetch.spec.whatwg.org/#dfnReturnLink-20 [18:04:58.0000] > If the CORS flag is set or httpRequest’s method is neither `GET` nor `HEAD`, then append `Origin`/httpRequest’s origin, serialized and UTF-8 encoded, to httpRequest’s header list. [01:00:04.0000] does anybody know if there's any downside in using ceche immutable? [01:00:07.0000] *cache [01:01:03.0000] XhmikosR: you need to change URLs whenever you update the resource [01:01:31.0000] annevk: I already need to do that in the time my cache is valid which is too long already :) [01:01:56.0000] XhmikosR: then you should be fine I think [01:02:03.0000] so my question should be, if one has cache busting and long cache headers, does immutable have any downsides? [01:02:54.0000] I see, thanks. I wanted to give immutable a go [03:17:26.0000] In Bikeshed, How would I link to the Audio named constructor of in HTML spec? {{Audio}} doesn't seem to work. [03:18:04.0000] s/of in/in the/ [03:23:22.0000] tobie: you'd have to PR HTML to export it appropriately [03:23:51.0000] annevk: ty [06:20:12.0000] annevk, rs? https://github.com/w3c/testharness.js/pull/238 [06:22:16.0000] Or foolip ^ [06:23:26.0000] Ms2ger: looks okay, do we really still need ArrayClass to exist at all? [06:23:38.0000] Ms2ger: I thought we could remove that, or did I forgot about something silly? [06:24:24.0000] /me can't find any GitHub issues [06:24:47.0000] Ms2ger, annevk: I merged it [06:25:00.0000] annevk, cssom still uses it, for better or worse [06:25:02.0000] Takk [06:25:12.0000] Ms2ger: can't CSSOM use iterable<>? [06:25:18.0000] No comment [06:25:30.0000] I can't find a single issue so I'm just going to raise one against IDL [06:28:35.0000] Ms2ger: https://github.com/heycam/webidl/issues/291 [06:28:56.0000] Thanks [09:37:18.0000] No Hixie here? annevk, what is the canonical pronunciation? what wee gee? [09:38:04.0000] foolip: I use it [09:38:54.0000] annevk: then I will too. I've alternated between what wig and what double-u gee. [09:39:40.0000] (and what ve ge in Swedish) [09:40:00.0000] annevk: working on bit.ly/blinkon7-wpt [09:42:32.0000] foolip: Domenic uses wig [09:43:17.0000] foolip: are things recorded? Some talks seem very interesting [09:43:45.0000] annevk: yes they are, but it seems like videos aren't available immediately after, might take a few days [09:44:14.0000] I'm going to claim that the goal of standards is roughly "Maximize throughput of interoperably implemented spec text" [09:44:18.0000] Does that seem sane? [09:58:57.0000] foolip: I think also the goal of allowing followers (including late followers like Servo) is important. [10:00:52.0000] Domenic: absolutely. but if we have 4 interoperable implementations, is it not quite likely that the spec and tests are in good shape for a new entry? [10:01:31.0000] s/entry/enterer/? [10:04:10.0000] Yeah, converge impls, enable competition [10:05:01.0000] Not sure what maximize throughput means exactly, sounds like streams terminology [10:06:44.0000] heh, it's just the utilitarian in me [10:07:51.0000] oh, I see what you mean with streams, this is probably confusing phrasing [10:07:54.0000] /me will tinker [10:10:40.0000] bz to the rescue in that confusing origin/domain thread, yay [10:11:28.0000] /me finds a recent blog post with the wording already tinkered [10:42:56.0000] Domenic: maybe you can prod some people in person for the blogpost [10:43:24.0000] Oh yeah [10:46:20.0000] Domenic: kinda want it to include pretty graphics now, but we probably shouldn't ask for too much [14:22:01.0000] rniwa: haven't confirmed myself but we have reports that Safari also fails on emoji custom element names: https://bugs.chromium.org/p/chromium/issues/detail?id=687678 2017-02-02 [19:42:52.0000] Hey TabAtkins, could you add https://w3c.github.io/webauthn/ to shepherd? [20:23:03.0000] jyasskin: Will do tomorrow. File a bikeshed issue to remind me? [22:40:24.0000] TabAtkins: It seems like autolinks to method arguments don't allow omitting the arguments in the method signature. That is, {{Foo/bar()/baz}} doesn't work; I need {{Foo/bar(quux,baz,xyzzy)/baz}}, which is frustratingly long. [22:40:31.0000] TabAtkins: Sure [22:41:52.0000] Ah, that's https://github.com/tabatkins/bikeshed/issues/634 [01:56:30.0000] TIL I can use www.software.hixie.ch for testing document.domain. https://github.com/whatwg/html/pull/2288#issuecomment-276913576 [03:25:07.0000] foolip: I think a further interesting policy change would be to put the standard in charge of test maintenance as well [03:41:31.0000] What does that mean, exactly? [03:54:35.0000] I was mostly thinking about issues [03:54:55.0000] So its an issue against the standard if the test suite is broken [03:55:20.0000] Or an issue against the standard if coverage is lacking [03:56:28.0000] This would make it clearer how stable a standard, hopefully provide incentives to get those issues cleared, etc [04:44:51.0000] Getting rid of submodule issues is the thing I hate most about web-platform-tests, followed by finding the documentation [04:45:22.0000] I sometimes need to run git reset --hard origin/master before running the git submodule thingy [04:45:26.0000] Ugh [05:13:59.0000] hmm. is it known that url/a-element.html crashes safari tp? [05:22:24.0000] annevk: wpt repo has labels for all specs... maybe an easy first step is to file a single tracker issue for a standard asking to drive down open tests issues [05:27:12.0000] annevk: We were discussing adding open test PRs to the top of specs (dunno if that was what you were refering to or not) [05:34:11.0000] zcorpan: does it do so consistently now? Crash Safari? [05:34:17.0000] zcorpan: worth filing a bug on [05:34:22.0000] zcorpan: I noticed it happening once [05:34:47.0000] jgraham: that sounds interesting too [05:35:21.0000] zcorpan: yeah, could start doing that for WHATWG I suppose [05:35:33.0000] zcorpan: I believe I'm mostly on top with stuff I maintain, but who knows [05:35:40.0000] zcorpan: HTML probably has a bunch [05:48:05.0000] annevk: seems to consistently crash for me with Release 22 (Safari 10.2, WebKit 12604.1.4.2) [05:49:03.0000] zcorpan: ah yeah, do you want to file a bug? [05:49:16.0000] sure [05:54:24.0000] https://bugs.webkit.org/show_bug.cgi?id=167730 [05:57:13.0000] tobie: how soon do you expect https://github.com/heycam/webidl/issues/188 to be fixed? [05:57:19.0000] tobie: or should I workaround? [05:57:58.0000] annevk: I think it's next on my todo list. [05:58:17.0000] annevk: I think we have a plan for it. [05:58:24.0000] annevk: must check my notes. [05:59:39.0000] annevk: got sidetracked by figuring out WebIDL testing this week [06:01:41.0000] tobie: so bz suggested that since the way forward is defining your own toJSON method, a very simple fix would be to drop the restriction on defining your own toJSON method [06:01:52.0000] tobie: I could create a PR for that [06:02:06.0000] tobie: and the the fix for the remainder can be done later [06:05:15.0000] tobie: btw https://github.com/heycam/webidl/pull/244 has been open for a long time [06:05:50.0000] annevk: yes, I'm not too sure how to proceed with it. [06:09:40.0000] annevk I don't yet have a clear enough vision of what this part of the spec does to know precisely how to reorg things so they make more sense. [06:10:15.0000] tobie: okay [06:10:24.0000] tobie: I'm in no rush with that one [06:10:36.0000] annevk: I wish we had a better understanding of what external stuff might point the different parts of the spec [06:11:15.0000] tobie: yeah that would be useful now and then [06:11:48.0000] tobie: would be neat to have tooling... [06:12:20.0000] annevk: wrt to the toJSON stuff, I think we just need to prevent interfaces which rely on jsonifier to have a parent that specifies a toJSON attribute [06:20:36.0000] tobie: ah yeah, I didn't try to go that far in my PR [06:21:37.0000] annevk: well, jsonifier's not specified yet… so I didn't really expect you to. :) [06:22:05.0000] I just got word bz merged it, so \o/ [06:23:40.0000] annevk: do you guys plot things in the background. ;) [06:24:17.0000] tobie: I was chatting with bz in #content [06:24:23.0000] tobie: on Mozilla IRC [06:25:39.0000] sounds like a fun place to hangout. [06:38:17.0000] Contrary to what you might have heard on Twitter, the URL Standard feels pretty close to done parser-wise [06:38:42.0000] The main thing that still needs to be sorted is IDNA and what code points to allow in hosts/domains [06:57:01.0000] annevk: stupid question: why isn't file:/// part of local scheme ? https://url.spec.whatwg.org/#local-scheme [06:58:01.0000] AutomatedTester: I think the way we use local scheme wouldn't want file to be included [06:58:02.0000] more, why is it a special scheme? [06:58:08.0000] ok [06:58:39.0000] AutomatedTester: I'm not sure if file still needs to be a special scheme, it's worth looking into whether refactoring it into just being its own category is worthwhile [06:59:11.0000] AutomatedTester: the URL syntax section already has three categories, mostly [07:01:20.0000] annevk: would you like me to raise an issue about file being a special scheme? [07:03:35.0000] Can an URL without a special scheme have an empty hostname? [07:03:49.0000] AFAIK no, so that's why file is special. [07:38:55.0000] annevk: yes, I think so too. we discussed something similar on Monday, concluding that a good start would be to adding a tests section to the top of the spec, linking to wpt, open issues/PRs with the spec's label, and when we have it perhaps the dashboard too [07:39:14.0000] annevk: https://wptdashboard.appspot.com/t/dom/events if you haven't seen it [07:39:20.0000] note that the data is not fresh [07:39:31.0000] AutomatedTester: yeah, please do that, I should investigate it at least [07:39:40.0000] nox: but file can have an empty host [07:39:52.0000] nox: oh wait, you have it reversed [07:39:57.0000] Mmh? [07:40:03.0000] nox: special schemes require a host name, except for file [07:40:07.0000] Oh. [07:40:28.0000] annevk: this came up in the context of trying to figure out who's job it is to go through all open PRs in web-platform-tests [07:41:30.0000] foolip: right, I'm starting to lean towards it being the editor who is finally responsible [07:42:05.0000] foolip: but delegation seems fine, e.g., I'm not personally an OWNER of any test effort I think [07:43:20.0000] foolip: hmm, I'm getting contradictory, better think about it some more [07:43:34.0000] annevk: right, although I think every implementer should feel respsonsible for sheperding any change all the way to interop, ultimately someone needs to be the fallback [07:44:26.0000] foolip: Mark Pilgrim once wrote a nice article about this, classifying a couple of approaches [07:44:51.0000] annevk: is that now gone from the interwebs? [07:45:29.0000] foolip: found a backup: http://www.diveintomark.link/2004/why-specs-matter [07:46:52.0000] foolip: probably more fun than informative, but still [07:46:54.0000] haha, classic Pilgrim, I remember this one [07:48:25.0000] I remember thinking that I certainly would aspire to the angel description and not understanding why it was so implausible to him. [07:51:06.0000] "and then thoroughly test it against the accompanying test suite before shipping their product" is key, given that those test suites usually don't exist [07:54:49.0000] I guess a problem is that the standard is often broken, and so are the tests [07:55:00.0000] And the coverage of the test is low, etc. [07:58:13.0000] AutomatedTester: actually, never mind, I found the reason [07:58:33.0000] AutomatedTester: the main reason is the backslash handling [07:59:41.0000] AutomatedTester: although I guess we could make that a separate group, but not sure if there would be much benefit [08:53:35.0000] That article is amazing [10:38:07.0000] foolip: you around? [10:38:19.0000] foolip: I'd like to land https://github.com/w3c/web-platform-tests/pull/4658 [10:39:59.0000] annevk: not quite here [10:40:53.0000] annevk: if you've looked at current behavior and think that it's plausible that we could converge on the spec, then sure [10:40:57.0000] now I'm gone [10:44:57.0000] foolip: ta [11:23:01.0000] annevk: Why does http://w3c-test.org/html/browsers/history/the-location-interface/location-protocol-setter-non-broken.html use a setTimeout before posting the reply in the data: case? That seems to upset Firefox (not sure why it does, but removing it fixed the problem) [11:59:33.0000] annevk: ahh thanks [12:22:44.0000] jgraham: can look Tuesday [13:00:39.0000] annevk: (asking about a worker that had DOM)- A) jsdom isn't eaxctly tiny B) the other major issue is debuggability. i've filed https://github.com/tmpvar/jsdom/issues/1719 to ask for Chrome Debugging Protocol Support in jsdom but that's getting super exotic super fast [13:54:46.0000] annevk: Anyone else that can look? It's blocking a wpt update 2017-02-03 [18:43:11.0000] Hello all :D [00:18:54.0000] jgraham: maybe bz, I recall those timers being significant though [03:00:28.0000]
can contain: [03:00:29.0000] Either: One figcaption element followed by flow content. [03:00:30.0000] Or: Flow content followed by one figcaption element. [03:00:30.0000] Or: Flow content. [03:00:43.0000] but flow content... how many tags? [03:00:49.0000] 1? 1+? [03:01:11.0000] it is ok to put 2 images inside a
? [03:23:26.0000] felixjet: 0+ [03:23:51.0000] text is also flow content [03:25:05.0000] felixjet: you can use https://checker.html5.org/ to check if something is ok [03:38:04.0000] apparently i used 15.38 Tebibytes with BigQuery in january [03:40:20.0000] Sounds expensive [04:23:08.0000] jgraham: $71 [04:25:35.0000] but most months i'm charged $0.01 since using bigquery is free to some threshold [07:03:54.0000] Good morning! 2017-02-04 [16:35:59.0000] Good evening! [16:37:25.0000] Good afternoon! [16:51:16.0000] gsnedders: how have you been? [16:59:20.0000] JonathanNeal: okay, been busy and at meetings/BlinkOn this week on my second trip to the US this year. [16:59:33.0000] Once a month now, huh? [17:01:21.0000] well, it's scarcely February! [17:01:30.0000] hopefully no more long-haul till September [17:01:41.0000] But I might have agreed to do one more before then :( [17:02:14.0000] What are you working on that keeps bringing you out? Or are these talks? [17:03:38.0000] JonathanNeal: CSS WG F2F at the start of the month, now here for a variety of meetings as well as BlinkOn (some with Google, some cross-vendor) [17:03:51.0000] getting paid by Google to work on test stuff nowadays [17:05:05.0000] getting paid by Google is always nice [17:05:15.0000] Yeah, can't say it's a bad hting. [17:05:17.0000] *thing :) [17:14:44.0000] fyi for travelers to US...fed judge in wash just issued temporary restraining order immediately halting trump's immigration ban...so customs lines should be shorter for now [17:15:36.0000] it'll probably be in 9th circuit appeals court tomorrow, though, so who knows how long it'll last [17:18:14.0000] s/customs/immigration/ I assume, but still, depends on the CBP actually following a court order which hasn't always happened… [17:20:03.0000] will be interesting to see whether it gets to a 4-4 supreme court [17:20:40.0000] oh, god, I'd never thought of the fact that there's a vacant position there… [17:21:48.0000] so, uh, what happens in that case? [17:22:46.0000] in case of 4-4 ties, appeals court decisions stand 2017-02-05 [17:40:12.0000] XhmikosR: I will need to a new checker release again soon [17:41:13.0000] XhmikosR: there was a pretty serious memory-leak/bloat problem in the code [17:42:05.0000] XhmikosR: https://github.com/validator/validator/commit/6bccda0b460bb2a5d29359a02a744d73913d64e0 [17:44:25.0000] XhmikosR: the code creates a buffer to hold the first 34KB of each document it checks (for use with the language detector) but neglected to clear that buffer between documents, so the more documents you check, the bigger that buffer gets [00:47:17.0000] MikeSmith: all right, just ping me and I'll do it as soon as I'm in front of my dev rig [08:49:33.0000] 1 [08:49:46.0000] oofs [08:50:03.0000] XhmikosR: 17.2.1 released [08:58:33.0000] MikeSmith: HTML 17.2.1? [08:58:46.0000] Times flies [08:59:57.0000] MikeSmith: done [09:03:05.0000] XhmikosR: thanks [09:54:37.0000] where would one put tests about canvas element [09:54:45.0000] not about context ,but the element itself [09:55:03.0000] I would expect somewhere under html/ but can found a nice place [10:01:20.0000] huh, there are very very few tests 2017-02-06 [17:01:02.0000] smaug____: yeah html/semantics/embedded-content/the-canvas-element directory [17:01:15.0000] smaug____: what kind of tests were you expecting there? [17:01:16.0000] oh there [17:01:19.0000] didn't find that [17:01:29.0000] well there are not many there either [17:01:30.0000] MikeSmith: someone was asking where to add new tests [17:01:33.0000] OK [17:02:22.0000] I guess the spec does not have many requirements in that section that are not requirements of the 2dcontext API [17:02:32.0000] I suggested html/dom [17:02:54.0000] ah [17:03:24.0000] there are no canvas tests under there [17:04:03.0000] sure [17:04:11.0000] but that doesn't really say much [17:04:12.0000] well except the HTMLCanvasElement interface [17:04:38.0000] (assuming that’s what it’s called) [17:04:41.0000] but ok, I commented that html/semantics/embedded-content/the-canvas-element could be used [17:04:47.0000] thanks [17:05:00.0000] cheers [17:06:52.0000] reading https://lists.webkit.org/pipermail/webkit-dev/2017-February/028694.html “Upstreaming Tests from WebKit to Web Platform Tests” from rniwa [18:50:37.0000] https://www.w3.org/community/httpslocal/ [01:24:16.0000] zcorpan: FYI https://bugzilla.gnome.org/show_bug.cgi?id=778192 (request to have libxslt output (as Saxon already does) [01:25:23.0000] zcorpan: also I discovered that if you use with libxslt it will output [01:25:36.0000] due to https://bugzilla.gnome.org/show_bug.cgi?id=747301 [01:26:33.0000] but if you use with Saxon or other XSL implementations, of course you will get a literal SYSTEM "about:legacy-compat" in the doctype, as you’d expect [01:27:27.0000] I don’t have much hope that Daniel Veillard will implement my request unless I provide a patch (and maybe not even then) [01:28:20.0000] But I prefixed the bug title with “Please consider supporting…” so maybe that will help :) [01:31:04.0000] MikeSmith: 👍 [04:33:43.0000] https://wiki.whatwg.org/wiki/FAQ#Is_there_a_process_for_adding_new_features_to_a_specification.3F looks like steps 7, 8, 9 are kinda backwards now [07:08:21.0000] Yeah it'd be good to get that updated [13:28:35.0000] SimonSapin: Dammit Simon, you ninja'd me by like 30 seconds. I was almost done with a nearly identical response. ^_^ [13:28:43.0000] haha [13:29:14.0000] TabAtkins: so how do you feel about implementing it? [13:29:27.0000] Writing up that response now - please god do it. [13:30:03.0000] TabAtkins: we had a servo PR for this last year, but it was kind wrong in lots of ways and I never got around to fixing it up. Then it got bitrotted enough that we closed it. [13:30:53.0000] (And it’s not like servo was gonna drive adoption. But now with Stylo it could be another story…) [13:42:56.0000] SimonSapin: Implementing what? [13:43:06.0000] nox: MQ4 new syntax [13:43:16.0000] https://lists.w3.org/Archives/Public/www-style/2017Feb/0035.html [13:43:52.0000] Specifically, adding the relational operators (<, >, <=, >=) to MQ rather than using min-*/max-* prefixes on things. [13:44:00.0000] Nice. [15:57:35.0000] element's browsing context isn't something defined anywhere, right? 2017-02-07 [16:00:30.0000] /me wonders if intersection observer has any good tests [16:11:55.0000] smaug: https://github.com/w3c/web-platform-tests/pull/4384 in theory... [16:12:58.0000] /me is thinking about cases like moving elements to other documents etc [16:18:18.0000] smaug: spec/test author says no. He also says "patches welcome!" :) [16:18:34.0000] heh [16:18:46.0000] (admitting it's a major omission) [16:20:03.0000] smaug: An element's browsing context would be the element's node document's browsing context, right? https://dom.spec.whatwg.org/#concept-node-document to https://html.spec.whatwg.org/#concept-document-bc [16:25:10.0000] jyasskin: sure, but it isn't defined anywhere what element's browsing context is, right? [16:25:41.0000] Yeah, HTML always says "foo's node document's browsing context" or "iframe element's browsing context". [16:26:55.0000] I believe the second means the nested browsing context, which makes it less straightforward to just define "element's browsing context". [00:43:38.0000] jyasskin: smaug: we should probably just change it to element's node document's browsing context, unless it's actually element's associated browsing context (the nested browsing context) [00:43:58.0000] jyasskin: smaug: I'd happily fix instances/take PRs [01:01:56.0000] zcorpan: do you have a pointer to the Safari TP crash issue? [01:03:10.0000] annevk: https://bugs.webkit.org/show_bug.cgi?id=167730 [01:05:26.0000] zcorpan++ for https://github.com/w3c/web-platform-tests/pull/4690 [01:05:36.0000] /me created that mess [01:18:58.0000] annevk: it was fun to see if my idea would work, which it did. but it's kinda non-obvious. if we need this sort of setup in more places we should think of better ways to do it in testharness [01:19:48.0000] zcorpan: I don't think we generally need this [01:19:56.0000] yeah [01:19:58.0000] zcorpan: this is mostly for aligning legacy setups [01:21:12.0000] although "missing" assertions because an early but less important assertion failed is something people find annoying sometimes, i think. [04:40:09.0000] what does step() mean with async tests? why is that needed, why not just use asserts? [04:40:34.0000] jgraham: ^ [04:42:20.0000] another question, why does http://searchfox.org/mozilla-central/source/testing/web-platform/tests/html/semantics/embedded-content/the-canvas-element/toBlob.jpeg.html work? http://searchfox.org/mozilla-central/source/testing/web-platform/tests/common/canvas-tests.js#53 seems to call t.done() before the callback to toBlob is called [04:47:03.0000] smaug: it helps catch the exception [04:47:37.0000] annevk: step() ? but it isn't needed, right? asserts without step work fine? [04:48:05.0000] smaug: no they don't work fine, since there's no global exception handler, unless you're creating a standalone test [04:48:37.0000] annevk: I mean if one has simple test without any possible exceptions [04:49:11.0000] smaug: if it's a single test you can use http://testthewebforward.org/docs/testharness-library.html#single-page-tests [04:49:42.0000] and async.. [04:49:53.0000] smaug: yes, you can still use that [04:50:04.0000] so, step isn't needed [04:50:11.0000] it can be useful in some cases [04:50:19.0000] smaug: only if you explicitly call async_test() [04:50:58.0000] sure, t = async_test(); assert(); t.done() [04:51:18.0000] /me is still puzzled why asserts work even _after_ calling done(); [04:51:21.0000] smaug: no, that's not okay [04:51:29.0000] why not? [04:51:39.0000] smaug: because then it's unclear to what test the assert belongs [04:55:52.0000] /me assumes if there is just one async tests not-done, asserts would be bound to that [04:56:05.0000] and based on a patch that works [04:56:38.0000] and http://searchfox.org/mozilla-central/source/testing/web-platform/tests/html/semantics/embedded-content/the-canvas-element/toBlob.jpeg.html [04:56:51.0000] that calls assert asynchronously [04:56:58.0000] certainly outside any step [04:57:05.0000] (and even after t.done()) [04:58:08.0000] smaug: that it works after t.done() should be enough of an indication that it isn't bound to the test whatsoever... [04:58:33.0000] what you mean with that [04:58:49.0000] exactly what I wrote [04:59:07.0000] harness says the test passed [04:59:31.0000] testname "toBlob with image/jpeg returns a JPEG Blob" passed [04:59:57.0000] _addTest does special things [05:00:30.0000] Oh yeah, that resource doesn't look like a normal assert either [05:00:43.0000] it uses a* [05:01:06.0000] smaug, I'm not sure I follow; done() is called after step() [05:01:21.0000] Ms2ger: done is called before the assert [05:01:28.0000] in http://searchfox.org/mozilla-central/source/testing/web-platform/tests/html/semantics/embedded-content/the-canvas-element/toBlob.jpeg.html [05:01:30.0000] No... [05:01:39.0000] Oh, wait [05:01:41.0000] I didn't say done is called before step [05:01:50.0000] Is toBlob sync? [05:01:54.0000] no [05:01:58.0000] I guess that would make no sense [05:03:00.0000] Yeah, that test is wrong [05:03:40.0000] I should find some time to make those canvas tests make sense [05:04:16.0000] Ms2ger: so why does the test work? [05:04:59.0000] in other words, what is the meaning of done() ? [05:05:01.0000] "work" how? [05:05:11.0000] I don't think the test checks anything [05:05:48.0000] oh, hmm [05:06:04.0000] do we get pass even without any asserts? [05:06:14.0000] Yes [05:06:33.0000] aha, that is a bug in framework [05:06:35.0000] ok, I see [05:06:48.0000] and asserting after done() should cause failure [05:09:47.0000] That would mean the test could never finish [05:09:52.0000] Anyway, I'll go and fix these [05:10:47.0000] Ms2ger: mochitest does warn if you don't have any asserts, and it does warn if you try to use ok() or such after finish() [05:14:58.0000] https://github.com/w3c/web-platform-tests/pull/4753 [05:15:23.0000] * [05:15:27.0000] focus fail [05:28:54.0000] smaug: Warning for assert-after-done is probably possible, but it's not possible to make it affect the test outcome because it would obviously be racy. [05:29:49.0000] A test with no asserts seems legit e.g. img.onload = done; img.onerror = () => assert_unreached("image should not have errored"); [05:29:59.0000] smaug: done() and t.done() and different things [05:30:02.0000] *are [05:30:36.0000] One says that a particular async test is finsihed, the other says that the entire page is finished (assuming there are no more tests which are not themselves finished) [05:30:49.0000] If you have only a single test in a file you can just write bare asserts [05:31:30.0000] If you want multiple tests in a file they have to be called inside a specific step otherwise they will simply cause the page to have status ERROR without failing any specific test [07:54:30.0000] zcorpan: count me a bit amazed but genuinelyhappy https://git.gnome.org/browse/libxslt/commit/?id=5f472f85d074828316ecfa8e5df91e6be4163aeb [07:56:21.0000] MikeSmith: 🎉 [07:56:58.0000] time to write some sweet XSLT [08:01:02.0000] zcorpan: My condolences. [08:01:47.0000] zcorpan: haha 😆 [08:06:56.0000] jgraham: why is test without asserts ok? I would expect img.onload to require something like function() { assert(true); done(); } [08:09:15.0000] smaug: Congratulations you are the first person ever to ask for more boilerplate in a wpt :) [08:09:32.0000] ha [08:09:47.0000] I'm trying to ask API I could understand and which would be less error prone [09:18:25.0000] smaug: a test passes if no assertions fail. so no assertions means it passes [09:18:50.0000] no assertions most likely mean buggy tests [09:19:02.0000] like in toBlob.jpeg.html case [09:41:57.0000] MikeSmith: so XSLT now depends on non-standardized features? [09:43:48.0000] I guess XSLT itself doesn’t, but in practice getting the two most-commonly used XSLT engines to output does [09:43:51.0000] now [09:46:47.0000] anyway XSLT overall is a “doctor it hurts when I do this” thing but a lot people still use it so it seems useful to mitigate bad consequences of it when we can [09:47:33.0000] I just want people to be able to use in their docs without their toolchain getting in teh way [09:49:07.0000] make fugly the forgottent footnote in history it deserves to be [09:50:06.0000] and further eliminate the misunderstanding that people continue to have that the doctype means anything or that browser do anything with it other than just using it as signal to not switch into quirks mode [10:03:28.0000] why am I trawling through MO-archive drama? [10:03:31.0000] this isn't productive. [10:11:13.0000] gsnedders: work on WPT docs 🙌 [10:36:15.0000] MikeSmith: I made a mistake in my shadow tree review [10:36:18.0000] MikeSmith: you were right [10:36:26.0000] MikeSmith: a shadow tree has a root that's a shadow root [10:36:33.0000] MikeSmith: and that is the thing that has a host [10:36:42.0000] MikeSmith: sorry [10:39:35.0000] annevk: this seems like a perfect case for pushing a fixup for him :) [10:40:09.0000] Domenic: MikeSmith: I'm happy to fix it tomorrow [10:42:03.0000] MikeSmith: are you trying to say that about:legacy-compat should be used? [10:50:51.0000] MikeSmith++ [11:04:36.0000] annevk+++ 👻 [11:04:47.0000] boogyman: saying that it should not be used [11:05:22.0000] saying we should help it along to death [11:11:30.0000] right now the spec says, Here’s what a doctype looks like: . It’s just that simple!... Oh, except here’s this other crazy-ass bit that literally only 0.1% of you reading this will even understand the purpose of but that we’re gonna just drop it in here anyway [11:12:35.0000] Domenic: https://github.com/whatwg/fetch/issues/479 could use your input somewhere today [11:12:42.0000] (your timezone) [11:22:50.0000] annevk: FWIW, I agree with your preference that `{}` empty the header list and `undefined` pass it through, and with bz's preference for having a general rule that absent braced things are equivalent to `{}`, so I'm no help. [11:25:49.0000] jyasskin: good times [11:26:12.0000] jyasskin: maybe we need [AllowOmission] [12:08:55.0000] Gecko peeps: https://developer.mozilla.org/en-US/Firefox/Releases/51 and MDN claim Storage API's navigator.storage.estimate() is in 51 but I'm not seeing navigator.storage existing. Patch is https://bugzilla.mozilla.org/show_bug.cgi?id=1267941 - am I missing something? [12:15:14.0000] jsbell: Secure context only, perhaps? [12:15:29.0000] jgraham: was trying it in console from https site, hrm... [12:16:13.0000] jsbell: WFM in nightly at least [14:10:43.0000] MikeSmith: botie has gone missing, making it hard for me to deal with zcorpan having gone missing :P 2017-02-08 [16:05:41.0000] Domenic: oof will try restarting [16:11:25.0000] Domenic: botie is alive again [16:11:31.0000] \o/ [16:11:44.0000] In the meantime I have resorted to GitHub pinging [16:12:01.0000] (https://github.com/whatwg/html/pull/2332 is the context for the record) [16:13:10.0000] "predecoding" really rolls off the tongue [22:29:52.0000] fucking hell the wiki is doing the “A database query error has occurred. This may indicate a bug in the software.” thing to me again [22:31:27.0000] in any browser so it’s got to be some misguided IP-blocking thing [23:01:10.0000] MikeSmith: works here, sigh [23:04:20.0000] annevk: yeah not sure what triggers it but the only solution just seems to be to wait for it to unblock [23:21:57.0000] MikeSmith: maybe time to move more stuff to whatwg.org and get that autopublished [23:22:22.0000] MikeSmith: it's ready to go, just need to coordinate with Ian [23:58:16.0000] annevk: yeah thanks for getting all that prepared, I think it will work out a lot better [01:04:36.0000] public-webappsec is slowly sliding towards not having any meaningful discourse either [01:33:48.0000] annevk: http://caniuse.com/#feat=url says URL is supported since Edge 12? [01:33:58.0000] re https://github.com/whatwg/url/issues/137#issuecomment-278275342 [01:34:13.0000] Oh my [01:35:29.0000] zcorpan: it totally is [01:35:37.0000] zcorpan: filing a bug now [01:35:46.0000] 👍 [01:35:53.0000] annevk: public-webappsec has plenty of meaningful discourse! For instance, just today we're discussing imaginary timelines for "completion" of documents that will likely never actually be complete. Totally useful. [01:36:14.0000] mkwst: yes [01:37:04.0000] annevk: should URLSearchParams also have toJSON? [01:38:34.0000] zcorpan: dunno [01:40:05.0000] mkwst, that must be the only w3c list where that happens [01:41:19.0000] Ms2ger: It seems to only happen when groups' charter expires. We make up dates, folks tell us that our made-up dates aren't made-up enough, and we make up new dates. Very worthwhile. [01:42:10.0000] mkwst: you might enjoy https://github.com/whatwg/url/issues/222 [01:42:40.0000] (brought to you by Amazon's recommendations system) [01:45:20.0000] can't all standards be done already? how hard can it be? [02:25:55.0000] > The @W3C Web Cryptography Working Group is closed, it produced Web Cryptography API [02:26:03.0000] zcorpan, clearly that one is done! [02:26:30.0000] there we go [02:27:17.0000] Hmm, except for the requirement about secure contexts that is being added? [02:27:21.0000] Time to set it up on whatwg.org for maintenance [02:29:05.0000] Seems like Multimodal Interaction's charter expired, so I guess EmotionML is done! [02:45:49.0000] mkwst: heh, just occurred to me you could block these base elements in CSP too [02:46:59.0000] annevk: If you're happy with me bypassing HTML by abusing the hooks you've so generously helped me add, I can do that. ;) [02:47:25.0000] mkwst: or CSP can block relative URLs :-D [02:48:32.0000] The web would be so much more secure (and less usable) if I had my way. [02:48:41.0000] Thankfully for everyone, I'm not the decider. :) [02:49:14.0000] (and block http: URLs) [02:51:30.0000] Obviously. That's step 1. [02:52:31.0000] annevk: Regarding this subresources thing, would you accept a patch that changed the definition of `subresource request` and `non-subresource request` to include/exclude requests targeting nested browsing contexts? That might be simpler than burying the distinction in a substep. [02:52:55.0000] mkwst: that would require coordination with JakeA and jungkees [02:53:03.0000] mkwst: I think [02:53:29.0000] Ok. There's no rush, right? Coordination doesn't seem like a blocker. :) [02:53:29.0000] mkwst: I would be happy with refactoring things though or introducing new terminology that helps explain this [02:53:57.0000] annevk: Even more terminology? Seems like we already have too much. :) [02:54:31.0000] mkwst: yeah, if only people stopped adding little exceptions to the rules here and there [02:55:10.0000] Exceptions are great! They let me break things I don't like without breaking things I do like! [02:55:35.0000] (Also: should I reflow entire paragraphs to 100 characters if I touch them?) [02:55:54.0000] mkwst: (yeah) [02:56:13.0000] mkwst: you like ? You monster! [02:57:21.0000] I like developers not storming my castle with pitchforks. [02:57:40.0000] Also, Google Cache uses ``, so killing it seems unlikely. [02:57:48.0000] (I looked into it) [03:01:24.0000] Hrm. Actually, this "subresource"/"non-subresource" distinction doesn't really match the way I think about the terms. Why are workers non-subresources? Also, why does "potential-navigation-or-subresource request" exist? [03:01:40.0000] I assume this is used in service workers, given your pointer to relevant folks. [03:02:42.0000] mkwst: a worker creates its own environment [03:02:51.0000] mkwst: the potential thing is for and [03:03:22.0000] mkwst: and yeah, service workers uses this to some extent [03:19:28.0000] tobie: I keep accidentally writing IDL patches against master :/ [03:20:11.0000] annevk: only heycam|away can fix this. [03:20:34.0000] tobie: okay, did we ping him recently? [03:27:56.0000] annevk: I check for him here daily, but he seems to be marked as away all the time. [05:14:49.0000] Can anybody please give me wording for a one-sentence definition of the concept of Service Worker scope? [05:15:27.0000] “The *scope* of a service worker is the...” [05:15:56.0000] MikeSmith: "... set of URLs that cause the service worker to run (roughly)" [05:17:05.0000] nice [05:17:07.0000] annevk: thanks [05:17:08.0000] MikeSmith: maybe "set of browsing context/environment URLs", to make it clear it's not subresources that matter [05:17:48.0000] OK [05:18:16.0000] but hmm in terms of how it’s declared, it’s a single URL [05:19:27.0000] “...a URL that defines a set of browsing context/environment URLs...” :/ [05:19:29.0000] MikeSmith: maybe say something like "substring" or "URL pattern" [05:19:46.0000] MikeSmith: there's a matching algorithm [05:20:10.0000] OK [05:21:48.0000] So a problem I have with Fetch is that it's hard to get review [05:22:13.0000] E.g., https://github.com/whatwg/fetch/pull/476 hasn't had much attention thus far [05:22:50.0000] /me assigns a bunch of people in the hope that'll work [05:24:53.0000] hah that’s one way to do it [05:26:25.0000] mkwst: when are we going to tackle https://github.com/whatwg/url/pull/72 again? [05:26:37.0000] mkwst: what's the current feeling at Google around public suffix? [05:26:52.0000] mkwst: it seems WebAuthn(sp?) is using it too now [05:27:58.0000] MikeSmith: I'd appreciate your input on https://github.com/whatwg/url/pull/228 [05:28:20.0000] /me hurries to respond before annevk assigns him [05:28:26.0000] MikeSmith: in particular since we moved from "validator" to "checker", should we also not use "valid"? [05:29:12.0000] MikeSmith: Domenic and zcorpan are suggesting that instead of "URL string" we say "valid URL string" [05:29:23.0000] yes [05:29:28.0000] I agree [05:29:35.0000] I like "valid URL string" [05:29:57.0000] “valid” by itself does not have the baggage that “validator” does [05:30:17.0000] but I guess it could also be “conforming URL string” [05:30:32.0000] or “conformant” [05:30:58.0000] but still in this context “valid” seems fine to me [05:31:04.0000] Okay [05:31:10.0000] wellformed vs valid? [05:32:11.0000] nox: seems like we have no wellformed-ness concept in this case [05:32:21.0000] hmm [05:32:31.0000] "Parser doesn't fail" [05:32:46.0000] vs "Parser doesn't fail and doesn't emit syntax violations" [05:33:02.0000] OK yeah [05:33:19.0000] There's also doesn't fail, doesn't emit syntax violations, and URL makes sense [05:33:32.0000] E.g., mailto://tralala/ [05:33:42.0000] annevk: When does it make no sense but there is no syntax violation? [05:33:45.0000] I don't know how many types we want to distinguish [05:33:55.0000] well anyway, zcorpan point is that we are using “valid” everywhere else already, so for consistency we should use “valid” for this too [05:33:55.0000] "mailto://tralala" seems perfectly legit to me. [05:34:05.0000] nox: it's not a valid mailto URL [05:34:41.0000] if we want to switch to using something other than “valid” then we should change it all globally [05:34:47.0000] which is closer to the XML distinction between well-formed and valid [05:35:04.0000] XML uses well-formed for parses into something, and valid for "makes sense per vocab" [05:35:13.0000] But I don't really care about being analogous to that [05:35:18.0000] yeah [05:37:06.0000] well-formed seems like it has mental baggage we probably want to avoid [05:41:03.0000] annevk: “The scope of a service worker is a single URL against which addresses much match in order to cause the service worker to run.”? [05:42:12.0000] MikeSmith: s/addresses/environment URLs/? [05:43:24.0000] “environment URL” is a dfn in the spec? [05:43:30.0000] /me looks [05:43:59.0000] hmm no [05:44:24.0000] this is pulling at threads [05:44:54.0000] hard to define it conceptually without also needing to mint other definitions it can reference [05:45:59.0000] annevk: “document addresses”?... [05:46:03.0000] https://html.spec.whatwg.org/#environment exists [05:46:20.0000] MikeSmith: "address" is not a thing, what's wrong with URL? [05:46:26.0000] MikeSmith: "document" doesn't capture workers [05:46:36.0000] MikeSmith: you could do document/worker URLs maybe [05:48:50.0000] yeah [05:48:52.0000] https://html.spec.whatwg.org/#navigating-across-documents:concept-environment-creation-url-2 [05:49:04.0000] “The created environment's active service worker is set in the handle fetch algorithm during the fetch if its creation URL matches a service worker registration.” [05:49:11.0000] that is not bad wording [05:49:32.0000] “creation URL” is dfn’ed [05:50:01.0000] meaning “environment URL” [05:50:53.0000] anyway I guess I will take a break now and go to the sento and meditate on it there [07:27:59.0000] JakeA: I seem to recall you told me in irc that you were ok adding a "serviceworker" ClientType. Am I remembering that right? https://github.com/w3c/ServiceWorker/issues/1036 [07:28:19.0000] just looking for a head nod in the issue so I can add it to gecko [07:28:32.0000] wanderview: I'll nod [07:28:37.0000] thanks [07:28:55.0000] JakeA: the reserved client thing is trickier than I thought, btw [07:29:19.0000] Which bit? [07:29:53.0000] JakeA: getting stuff initialized before the network load just breaks existing assumptions in the code [07:30:04.0000] maybe it will be easier in blink [07:30:30.0000] I've just had to fix a bunch of stuff in our worker implementation to support doing this [07:30:33.0000] which is good I suppose [08:18:26.0000] TabAtkins: any idea why the following returns 400? https://api.csswg.org/bikeshed/?force=1&url=https://raw.githubusercontent.com/tobie/sensors/fd4501a5c2bb7fb727fc2faff7f2c14f14227b74/index.bs [08:38:19.0000] Ask plinss? (Over in w3c#css) [08:38:20.0000] will do [08:38:34.0000] tobie: ^^^ [08:38:45.0000] ty [10:00:26.0000] reading https://webkit.org/blog/7380/next-generation-3d-graphics-on-the-web/ [11:18:28.0000] annevk: why https://github.com/w3c/web-platform-tests/pull/4702 is right? [11:18:47.0000] or I guess I should ask why the spec is right [11:19:04.0000] per spec href getter and toJSON return the same value [11:19:34.0000] but the test hints that value starts with " and ends with " [11:23:37.0000] smaug: JSON.stringify adds that [11:24:51.0000] annevk: ok, why do we have toJSON() then? [11:25:16.0000] oh, I see [11:25:20.0000] rather silly setup [11:25:29.0000] but I guess that is what toJSON is [11:26:16.0000] but url.toJSON() == url.href [11:32:49.0000] smaug: yeah [11:33:51.0000] I was just looking at a patch to implement that in Gecko and patch was clearly wrong since it was trying to json-stringify and return that value [12:13:02.0000] smaug: that's the behavior toJSONString() had (back when it was the first iteration of Crockford's JS library with the funky "don't be evil" license). 2017-02-09 [19:34:40.0000] annevk: I just noticed https://bugzilla.mozilla.org/show_bug.cgi?id=500713 has been waiting on a needinfo from me since last year re mimesniff. You wanna take it? [00:01:50.0000] GPHemsley: done [00:02:02.0000] GPHemsley: that is, I replied and cleared the flag for you [02:15:11.0000] annevk: if you have a minute to look at http://stackoverflow.com/questions/42131714/cors-access-control-max-age-works-for-same-origin-or-just-same-request-url please post a comment if I got anything wrong, or another answer even [02:15:45.0000] or if you have a place to cite in the spec for that [02:16:54.0000] I couldn’t see a specific dfn or other anchor to cite, so other than saying “Read through the spec” don’t know what else to point to [02:18:26.0000] hsivonen: can you take a look at https://github.com/whatwg/html/pull/2319 ? [02:19:17.0000] MikeSmith: that looks good, but it's slightly more complicated [02:19:36.0000] MikeSmith: it's per-URL and per-originating-origin [02:20:18.0000] MikeSmith: https://fetch.spec.whatwg.org/#concept-cache-match is the reference [02:20:23.0000] MikeSmith: again, roughly [02:34:35.0000] annevk: OK thanks [03:34:46.0000] nox: https://github.com/whatwg/url/pull/228#issuecomment-278617035 [03:35:21.0000] nox: and https://github.com/whatwg/url/pull/228#discussion_r100102850 [03:35:45.0000] I suggest calling a valid URL a crispy dark baguette, [03:35:58.0000] and a non-conforming one underbaked garbage. [03:36:10.0000] ❤️ like [03:45:57.0000] heh [04:36:39.0000] BenjaminSchaaf: the EOF thing is tested by /webvtt/webvtt-file-format-parsing/webvtt-file-parsing/support/signature-space.vtt at least, correct? [07:02:28.0000] zcorpan_: sorry I caused conflicts with your PR [07:03:38.0000] Domenic: 😡 😆 should be easy enough to resolve [07:09:58.0000] but i'll wait for hsivonen's review [07:10:09.0000] then i can squash and then resolve conflicts [07:14:56.0000] zcorpan_: FWIW, I have a slight preference for doing reviews based on non-conflicted code [07:15:29.0000] zcorpan_: chance for missing a typo increases if you edit after [07:15:41.0000] annevk: yeah, true [07:16:53.0000] annevk: but also becomes harder to review if i squash and then rebase. i suppose i can rebase without squashing though [07:17:17.0000] zcorpan_: that's what I tend to do, not ideal either of course [09:12:02.0000] MikeSmith: is there a canonical explanation for why "validator" is wrong? [09:12:40.0000] MikeSmith: I haven't really internalized the rationale and that seems to be affecting how I name things [09:18:22.0000] annevk: It is not canonicaly wrong so there canonical explanation. And in other contexts I think there is nothing wrong with it, but in the specific context of checking HTML documents I quit using it because it triggers assumptions from people about what the checker should do for them. [09:19:17.0000] in the context of checking HTML documents, people see it as meaning, e.g., they should be able to get a badge that says “Your document is valid HTML5” [09:20:24.0000] that is, it tends to cause people to see the purpose of checking your HTML to be just to get a binary pass/fail [09:20:42.0000] instead of the purpose of the checker being to help you find unintended mistakes [09:21:41.0000] and sometimes people intentionally use markup that causes the checker to emit an error or warning, but they know what they are doing and why [09:22:24.0000] so that’s the reason I added the message-filtering feature, to let people suppress checker errors and warnings they do not find useful [09:23:09.0000] and if they are using that feature, it is wrong for the checker to emit a message saying, Your document is valid HTML [09:23:43.0000] so instead I have the checker say, There were no errors or warnings to report. [09:24:40.0000] that is, to give control back to user about whether the document has errors as far as they are concerned [09:25:55.0000] MikeSmith: okay, I guess that's reasonable and way more nuanced than things were in my head [09:26:06.0000] OK [09:26:08.0000] MikeSmith: thanks for the help thus far [09:26:19.0000] thank you for caring about this stuff [09:26:23.0000] words are important [09:27:30.0000] oh man branch name annevk/url-ß [09:27:50.0000] that reminds me of something that zcorpan would do [09:28:01.0000] MikeSmith: I now want to try it everywhere [09:28:02.0000] intentionally, to see what it would break [09:28:09.0000] haha yup [09:30:58.0000] Yeah, so expecting unix command line environments to have sane unicode setups is a little optimistic [09:31:37.0000] I think travis does at least [09:32:28.0000] but if you mean every wpt user having a sane unicode setup I guess these days if they don’t they are gonna run into lots of other problems [09:33:12.0000] I have long go lost patience with Python2 on this [09:33:30.0000] I never run into annoying problems with Python except for this unicode thing [09:34:55.0000] with python I run into few other problemsーthings just workーand when I do run into problems it always otherwise turns out to be my fault [09:35:12.0000] but this unicode thing is python’s fault [09:49:16.0000] MikeSmith: It seems like this might be much easier to debug if you can try to reproduce locally [09:54:17.0000] jgraham: let me know if your recommendation is to submit again with an ASCII branch [09:54:25.0000] name [09:57:42.0000] tobie: PR preview seems to work for whatwg/url [09:57:45.0000] tobie: pretty nice [09:58:38.0000] Yeah. Fixed the auth today. [10:00:26.0000] tobie: what's the diff between "Diff w/ current ED | Diff w/ base" [10:01:54.0000] tobie: oh I see master vs branch point [10:02:28.0000] Happy to reword those. [10:03:29.0000] tobie: Is there a term for branch point? [10:04:04.0000] tobie: it would be nice if Preview could point to https://url.spec.whatwg.org/branch-snapshots/annevk/concept-relations/ instead [10:05:04.0000] tobie: I'd name the first "Diff with Standard" and I'm not quite sure what the use of the second is [10:05:13.0000] tobie: it seems you'd only ever want to review against master [10:12:11.0000] tobie: the HTML diffing is amazing [10:12:46.0000] annevk: it's keyboard navigable, btw. :) [10:12:50.0000] tobie: though I did find that diff with current ED doesn't seem to rebase [10:13:13.0000] tobie: so it instead acted as if the PR would revert changes made on master [10:13:32.0000] (I rebased myself locally now so it's no longer visible there) [10:14:34.0000] 19:12 tobie: though I did find that diff with current ED doesn't seem to rebase <-- not sure what you mean. [10:15:05.0000] tobie: on master I had changed "terminate these steps" to "return" [10:15:23.0000] tobie: the diff for the PR with current ED suggested that would change back to "terminate these steps" [10:15:57.0000] tobie: which is not what GitHub would actually do for squash/rebase [10:16:24.0000] Another question about origin equality... [10:16:24.0000] in step 4 of https://html.spec.whatwg.org/multipage/webappapis.html#dom-document-open, which equality is intended? [10:16:39.0000] same-origin-domain? [10:16:50.0000] tobie: so the diff is different from GitHub's diff, basically [10:17:03.0000] ajeffrey_: there's an open bug on that by jeisinger (also a PR) [10:17:13.0000] annevk: oh, no. It's because it got stale I guess. [10:17:30.0000] annevk: ah (goes and searches) [10:17:36.0000] That's the issue with caching. :( [10:18:35.0000] annevk: need to figure out something. [10:19:10.0000] annevk: that would be https://github.com/whatwg/html/pull/2288 then? [10:19:41.0000] annevk: thanks! [10:23:35.0000] https://wicg.github.io/IntersectionObserver/#calculate-intersection-rect-algo step 3 doesn't make sense, right? container is a containing block, but intersection root is a node [10:23:39.0000] or am I missing something here? [13:39:49.0000] is there something similar to the "unloading document cleanup steps" but for workers? (context: currently as spec'ed blob URLs are revoked when the document that created them is unloaded. but the spec doesn't say anything about blob URLs created by workers...) [14:18:06.0000] And yet another dumb question... [14:18:07.0000] https://html.spec.whatwg.org/multipage/browsers.html#dom-history-go says "if delta is zero, the user agent must act as if the location.reload()", [14:18:07.0000] does that include the same-origin-domain check at https://html.spec.whatwg.org/multipage/browsers.html#dom-location-reload? [14:40:24.0000] annevk: ^ [15:43:55.0000] annevk: You should now get the right diffs in pr-preview + link to the branch when it exists. [15:45:00.0000] annevk: I still need to fix the z-index of the HTML diff so the UI shows up above the warning on WHATWG specs, [15:45:11.0000] annevk: like this: https://s3.amazonaws.com/pr-preview/heycam/webidl/3834774..tobie:interface-objs:7dfd134.html [15:45:43.0000] annevk: (bottom right corner and "j" and "k" keys) 2017-02-10 [20:53:33.0000] for using to register a service worker, if it has what does the SW’s scope url get set to? the document’s base URL? [20:53:40.0000] https://w3c.github.io/ServiceWorker/#ref-for-element-attrdef-link-scope-2 [20:53:46.0000] > If the scope attribute is present, set scopeURL to the result of parsing the scope attribute with the link element’s node document’s document base URL. [20:54:58.0000] so as far as document conformance, should scope="" be allowed? [20:55:19.0000] and should it be allowed to contained leading/trailing space? [20:57:22.0000] in other words should it be a >valid URL< or a >valid URL potentially surrounded by spaces< or a >valid non-empty URL< or a >valid non-empty URL potentially surrounded by spaces< [21:16:31.0000] MikeSmith: I would base that decision on precedent of other such URLs, e.g. link href="" [21:16:48.0000] Off the top of my head I would guess "valid URL" but I don't know when we allow surrounding spaces in general [21:17:02.0000] I don't think the empty string should be excluded [21:17:54.0000] Domenic: thanks [23:58:46.0000] https://developer.microsoft.com/en-us/microsoft-edge/platform/issues/10552455/ \o/ [23:58:55.0000] First Edge bug I filed that is fixed I think [23:59:12.0000] And if it's actually fixed, that's a rather big one as it changes quite a few things [23:59:25.0000] But I wonder if they fixed it in depth or just surface [00:44:34.0000] annevk: Morning! Was https://github.com/whatwg/fetch/issues/486#issuecomment-278880798 directed at me or at Domenic? [00:45:49.0000] yoav: nobody in particular [00:46:24.0000] annevk: Did you refer to type handling in option (1) or option (2)? [00:46:31.0000] yoav: just got curious about it, since is just a high-level API in the end and it needs to map to fetch() somehow [00:46:55.0000] yoav: I guess 2, since 1 doesn't have a type attribute [00:47:10.0000] yoav: the parenthesis bit was about both though [00:47:28.0000] OK, I think I understand your question, I'll answer on the issue [01:08:12.0000] tobie: wow, you fixed all the nits with PR preview [01:08:27.0000] tobie: and you improved upon my wording too [01:08:41.0000] tobie: really cool [01:15:48.0000] yoav: thanks [01:16:16.0000] yoav: I guess I maintain that Node.js was right, they should have had their own MIME type [01:16:30.0000] yoav: but I suspect that will never fly with the JavaScript folks [01:16:43.0000] yoav: ugh, such a mess [01:17:24.0000] annevk: I'm in favor of them being their own destination [01:17:47.0000] yoav: given that MIME type is unlikely to be an option here, agreed [01:18:23.0000] also spent a few hours last week discussing how modules would actually be loaded in browsers, and I agree with your "ugh" [01:18:55.0000] heh [01:19:33.0000] yoav: btw, there is an HTML modules outline online, but it's dated perhaps [01:20:08.0000] yoav: https://github.com/w3c/webcomponents/blob/gh-pages/proposals/HTML-Imports-and-ES-Modules.md [01:20:10.0000] pointer? "HTML modules" is ungoogleable [01:20:15.0000] cool, thanks [01:56:58.0000] annevk: Unrelated, but is there a notion of "supported mime type" that I can link to from a spec? [02:00:37.0000] yoav: what does that mean? [02:03:13.0000] e.g. "image/webp" supported in an image destination context in Chrome, but not in WebKit [02:03:27.0000] I guess it's slightly related [02:03:47.0000] I'm trying to fix the preload spec to define what `type` currently does [02:04:52.0000] annevk: I have no trouble defining what `media` does (which was off as well), but not sure where mime types are defined in the platform [02:04:55.0000] We don't have overarching terminology for that I think [02:05:06.0000] hmm [02:05:07.0000] We do have terminology for media in HTML [02:05:25.0000] MIME types are somewhat defined in https://mimesniff.spec.whatwg.org/ but they're not at all interoperable [02:05:39.0000] But specific MIME types varies a lot [02:05:40.0000] OK, I'll take a look [02:06:00.0000] E.g., image loading ignores MIME types generally, even X-Content-Type-Options doesn't change that [02:06:09.0000] The only MIME type image loading looks for is image/svg+xml [02:06:12.0000] https://mimesniff.spec.whatwg.org/#supported-by-the-user-agent [02:06:35.0000] I don't care about mime types in processing the response [02:06:39.0000] Interesting, I guess that does do what you want [02:06:44.0000] just need them in order to avoid sending the request [02:06:49.0000] But again, I wonder if that's implemented as-is [02:06:56.0000] But maybe if we just start using it at some point it will be [02:07:09.0000] yeah. [02:07:23.0000] It's way better than what I have now in preload, which is nothing... [02:07:41.0000] heh [02:08:06.0000] My main problem with that spec is that we don't have active maintenance, solid testsuite, and commitments from implementers [02:08:45.0000] But some parts of it do, so we can't just scrap it, it does actually define some underpinnings [02:09:49.0000] yeah, not ideal [02:10:14.0000] morning all, in specs, do we need to link to a definition of what a promise is or is it accepted that everyone knows what it is? [02:10:37.0000] I noticed Fetch doesnt link to Promises [02:11:22.0000] so not sure what i should do in WebDriver. [02:13:32.0000] tobie: how do i propose changes to https://www.w3.org/2016/10/htmldiff-nav.js ? [02:15:01.0000] zcorpan: oh boy [02:15:04.0000] AutomatedTester: it's a problem IDL needs to solve at some point [02:15:11.0000] AutomatedTester: the current solution is to assume everyone knows [02:15:37.0000] annevk: ok, I will go with the current solution then :) [02:15:47.0000] zcorpan: fork https://gist.github.com/tobie/39c068f2b58422b5d20a5e24941c9f17 and the email dom? [02:16:13.0000] s/the/then/ [02:16:29.0000] what is it you want to change? [02:16:38.0000] tobie: outline instead of border [02:17:25.0000] tobie: there's also ":focus { border: thin red solid}" in the diff stylesheet inserted for the url standard, which seems unnecessary [02:18:35.0000] zcorpan: you're making the diff thing even better? Hurray [02:19:18.0000] tobie: I guess diffing for HTML would be hard to provide, right? [02:19:30.0000] zcorpan: I have some weird spacing around links in the WebIDL spec too. [02:20:51.0000] zcorpan: e.g.: https://s3.amazonaws.com/pr-preview/heycam/webidl/3834774..tobie:interface-objs:7dfd134.html#dfn-serialization-behavior [02:21:30.0000] zcorpan: look how the link underlining bleeds. [02:21:56.0000] tobie: that seems like a problem of the differ, not a styling problem [02:22:31.0000] zcorpan: :( [02:22:43.0000] tobie: it inserts a newline before or . [02:23:29.0000] and possibly in general lots of newlines [02:23:34.0000] zcorpan: oh well. [02:24:52.0000] zcorpan: I can totally live with this. :) [02:24:58.0000] would be better if it inserted newlines right before ">" to avoid messing with the Infoset [02:25:15.0000] where is the source for the differ? [02:25:39.0000] zcorpan: I think the underlying lib is an old C program. [02:26:30.0000] zcorpan: it's gnu diff: https://services.w3.org/htmldiff [02:26:59.0000] annevk: not sure what you mean by diffing for html [02:27:26.0000] tobie: HTML Standard [02:27:48.0000] annevk: oh! [02:28:04.0000] There's only one, so HTML is totally unambiguous, right? [02:29:11.0000] annevk: yeah, I could special case granted there's a Web service I can use for the build [02:29:34.0000] annevk: or I can run the software to build it on the heroku platform [02:30:00.0000] tobie: we do have that, the real nice thing to offer would be multipage diffing, with links only to the multipage files that changed [02:30:32.0000] Not sure how many stretch goals we want though [02:31:15.0000] tobie: having this for the smaller specs is already amazing [02:31:34.0000] tobie: this is a big complaint from folks we ask to review [02:31:55.0000] annevk: I built it for reviewing WebIDL PRs [02:32:12.0000] annevk: there was no way I could be effective reviewing stuff otherwise [02:32:18.0000] tobie: glad you did [02:32:51.0000] annevk: wrt to HTML, I don't think it would be too hard to build the multipage thing [02:34:18.0000] annevk: but I don't want to focus on that just now [02:55:06.0000] annevk: i'm pretty sure HTML standard is too big for htmldiff :-( [02:56:15.0000] zcorpan: multipage HTML wouldn't be? [02:56:30.0000] ah right. yeah multipage would probably work fine [02:56:34.0000] Anyway, seems like something to tackle later when there's revived interest [02:57:06.0000] Meanwhile we can add this feature to other WHATWG standards if nobody has issues with the way it works for the URL Standard [03:07:07.0000] /me finds the most bizarre spec language [03:07:20.0000] "Every millisecond, as long as the stop timeout flag is unset, queue a microtask to run these subsubsubsteps: " [03:08:11.0000] smaug: that looks buggy [03:08:34.0000] using microtask there is wrong [03:09:05.0000] annevk: if you can wait a bit before you extend to other specs, it would be amazing. [03:09:09.0000] Doing something every millisecond also seems suspicious [03:09:22.0000] tobie: sure, ping me whenever [03:10:41.0000] https://github.com/whatwg/xhr/issues/112 [03:14:17.0000] smaug: heh, I wonder if I wrote that [03:15:13.0000] zcorpan: what does w3c-test:mirror do? Make the bot upload the PR to WPT? [03:15:20.0000] that happens. Looking at the blame in Gecko wondering who has written some odd stuff and realizing hey, it was me :) [03:15:24.0000] zcorpan: to w3c-test.org? [03:15:27.0000] annevk: yes [03:15:33.0000] zcorpan: cool [03:16:43.0000] smaug: turns out I wrote it two years ago [03:18:11.0000] "Every full moon, queue a microtask..." [03:18:57.0000] smaug: so I think the reason it does it that way is to able to check the current value of the timeout attribute [03:19:38.0000] smaug: but it's indeed rather silly, having said that, I don't immediately know what a good fix would be, defining timers isn't easy [03:20:47.0000] why is that using microtasks? [03:21:02.0000] I doubt browsers use microtask there [03:21:50.0000] smaug: I suspect browsers proxy the other way, message whenever timeout updates [03:21:59.0000] smaug: instead of polling it every second [03:22:37.0000] yeah, fire a timeout which queues a task to cancel the fetch [03:23:09.0000] smaug: so you have a "process" that runs the timer [03:23:27.0000] smaug: but then whenever the timeout attribute is set, that influences when the timer is done [03:23:43.0000] smaug: so those two things need to talk to each other in some way [03:23:52.0000] if timeout attribute is set to a new value, the previous timer is cancelled and new one created [03:25:00.0000] smaug: wait so if you start with 2000 and then set timeout to 3000 at 1000, you'd effectively allow the request to go for 4000? [03:25:13.0000] smaug: I thought the idea was 3000 [03:25:46.0000] when doing reset, you do take into account the time already spent after send [03:26:02.0000] okay [03:26:40.0000] I guess I should define something more aligned with that model then, seems reasonable and doable [03:27:09.0000] how does html spec define setTimeout handling... /me reads [03:31:37.0000] "wait ... timeout milliseconds" [03:33:00.0000] Yeah, that's probably reasonable, but we'll need to store how many milliseconds have passed when we reset [03:33:21.0000] Arguably we should define an abstraction that both standards could use, but maybe not for now [03:54:53.0000] who has an iphone and can test http://software.hixie.ch/utilities/js/live-dom-viewer/saved/4882 [03:57:26.0000] (interested in whether the native controls have an airplay button for one but not the other) [04:02:38.0000] hmm maybe it should be x-webkit-wirelessvideoplaybackdisabled or disableremoteplayback these days [04:09:49.0000] zcorpan: big play buttons [04:10:12.0000] annevk: and when you play? [04:12:04.0000] I see now that disableRemotePlayback has already been discussed in https://github.com/whatwg/html/issues/2293 so what I wanted to get to is not really relevant [04:12:08.0000] zcorpan: only the first goes through Apple TV display wise [04:12:21.0000] annevk: ok, thanks [04:12:25.0000] zcorpan: both are happy to play sound over it [04:12:40.0000] https://irccloud.mozilla.com/file/mm8hohrV/IMG_4367.PNG [04:12:42.0000] intredasting [04:12:59.0000] As you can see rendering is also somewhat different [04:13:44.0000] Finally a way to listen to hsivonen in the living room [04:14:50.0000] That's what any home needs [04:29:13.0000] 😆 [04:45:06.0000] hah maybe Henry Story can lobby Trump to sign an Executive Order requiring all browsers to fully implement https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/z_qEpmzzKh8/2hkaJdtsCAAJ [05:09:43.0000] annevk: why is https:example.org without base valid? [05:12:20.0000] zcorpan: it's not [05:12:58.0000] annevk: then https://url.spec.whatwg.org//branch-snapshots/annevk/validity-example/#example-url-parsing is wrong :-) [05:13:07.0000] yeah [05:13:22.0000] zcorpan: only that first one, right? [05:13:24.0000] fixing [05:14:15.0000] annevk: i think so. but haven't tried to verify that the others match the normative text [05:30:35.0000] zcorpan: thanks for the additional examples, adding them now [05:52:44.0000] lol annevk/url-ß [05:53:56.0000] at least good to see github seems to handle it [05:54:24.0000] 💩 is next, just need a good PR for it [06:24:25.0000] annevk: other than XHR and fetch() and the crossorigin attribute, are there any other cases where browsers will use the CORS protocol? [06:24:39.0000] MikeSmith: @font-face [06:24:45.0000] ah yeah [06:24:57.0000] OK, that’s the whole list? [06:25:00.0000] (currently) [06:25:21.0000] MikeSmith: , if that ends up with one fetch, it's not a resource that's returned [00:32:40.0000] Resource is the internal representation of a fetch (which may be in flight) [00:32:46.0000] right [00:33:26.0000] it doesn't sound like an awfully big project to add that kind of "fetch cache" [00:33:55.0000] so writing down what happens in Blink/WebKit during those checks is probably doable [00:34:22.0000] but I'm not familiar with Firefox's code in that area (and have no visibility into Edge) [00:36:15.0000] and I suspect each one of those checks will also require a (much needed but lengthy) discussion of "do we really need that?" [00:38:19.0000] annevk: I can probably start tackling that next week or the week after that, and see how it goes [00:38:33.0000] starting by documenting the existing implementations [00:38:56.0000] which will probably require feedback from Mozilla folks [00:39:01.0000] Yeah I think having a description of Chrome/WebKit's model would be good as reference [00:39:12.0000] And then we can figure out if others want to implement that too [00:39:17.0000] For non-preload purposes [00:39:34.0000] And we can figure out if the Chrome/WebKit model has holes with respect to CSP and such [00:39:45.0000] yeah [00:40:09.0000] I suspect it won't, as it tends to err on the side of "reload the resource" [00:40:25.0000] mkwst_ probably audited the code [00:40:27.0000] I hope so anyway [00:40:33.0000] oh yeah [01:17:14.0000] Can someone look at https://travis-ci.org/w3c/web-platform-tests/builds/201282365 and tell me what is going wrong? jgraham? [01:43:12.0000] annevk: all I'm getting internally about the HTML outline is "we think it might break sites" but no evidence as yet [01:43:25.0000] annevk: anything from Moz? [01:43:35.0000] JakeA: basically lack of time [01:44:03.0000] annevk: kinda sad that browsers did all of the bits except the a11y bit [01:44:06.0000] I guess that's nothing new [01:44:10.0000] JakeA: the breaking site thing seems super weird to me [01:44:27.0000] JakeA: we rolled out this feature somewhat poorly [01:44:46.0000] annevk: I mean, it'll change the outline of some sites, but I don't know if it'd be "bad" [01:44:50.0000] JakeA: everyone was keen on getting some new HTML elements, without doing all the work (e.g., as of yet no CSS selectors either that make it easy to style heading levels) [01:45:31.0000] annevk: right! And that's what's happening on that w3c thread. Everyone's super excited about the prospect of a ****new tag**** [01:46:00.0000] JakeA: yeah, new features are shiny, fixing infrastructure is not [01:46:43.0000] JakeA: which reminds me, browsers still haven't removed appcache... [01:46:55.0000] *siiiiigh* [01:47:17.0000] annevk: there's a vague worry internally that it's being used by intranets, but again, not seen any evidence [01:47:31.0000] JakeA: can't they do the long term removal for intranets? [01:47:40.0000] JakeA: with some kind of flag for enterprises [01:47:51.0000] JakeA: and just let that flag exist for 1.5 years [01:48:00.0000] JakeA: isn't that how showModalDialog got removed? [01:48:11.0000] annevk: yeah, that seems fair [01:48:43.0000] JakeA: it's long, but it's better than just letting the code rot and be a source of vulnerabilities [02:16:03.0000] annevk: I'm not sure what else I need to do to get https://github.com/w3c/web-platform-tests/pull/4518 merged. Travis is barfing, but it isn't clear on what, or if it's related to the PR [02:17:38.0000] JakeA: it seems the results for Chrome are unstable? [02:18:14.0000] JakeA: I second the problem with Travis though, the output is not great [02:20:37.0000] annevk: yeah, I can't figure out if it's specific to my tests [02:23:09.0000] JakeA: That seems pretty clear from the GitHub comment? Chrome timed out on 2 runs out of 8, so the tests are unstable and can't yet land [02:23:40.0000] annevk: It looks like for some reason travis thought that branch contained hundreds of commits [02:23:45.0000] I'm not sure why [02:23:56.0000] maybe because I rebased [02:24:31.0000] jgraham: ah, I didn't spot the relevant part in the wall of text [02:25:16.0000] Domenic: "I wonder if we can do it purely through CSS (I doubt it)." - was this intended to trigger a knee-jerk on my part to fix it just to prove you wrong? 🤔😁 [02:25:20.0000] jgraham: I think what's going on is that he's trying to merge into the MicrosoftEdge branch rather than master [02:25:45.0000] JakeA: Anything you suggest to make it more obvious? It's at the top in bold. doesn't work these days… [02:26:22.0000] annevk: Oh, that could be it [02:26:58.0000] jgraham: your comment tipped me off somehow, even though I had seen it trying to apply hundreds of commits [02:28:06.0000] annevk: We can probably make that work in the bot, but I don't know why you would do that in general [02:28:19.0000] jgraham: :P I guess fold all the passes into a click-to-reveal thing. The "everything's ok" parts of the messages are huge compared to the "here's the actual problem" [02:28:34.0000] I hope Microsoft haven't got the idea that they should put all their commits in a MicrosoftEdge branch [02:28:48.0000] JakeA: Hmm, does nested
work? Could try it [02:29:01.0000] jgraham: I dunno :( [02:29:37.0000] jgraham: in this case the unstable tests are nothing to do with my PR. Should it block me from merging? [02:30:08.0000] jgraham: I'm hoping it was just a mistake, we'll hear tomorrow hopefully [02:30:53.0000] JakeA, is it the rebase or the test-helpers.sub.js change? [02:31:19.0000] jgraham: nested works [02:31:51.0000] annevk: if I got anything wrong at https://stackoverflow.com/questions/42222937/using-fetch-api-to-post-xml/42223732#42223732 lemme know or comment/answer there when you got time [02:32:24.0000] Ms2ger: hmm, maybe, I'm not sure how it works [02:33:17.0000] I think it's accurately detecting that your change affected all the things [02:34:20.0000] MikeSmith: seems accurate enough [02:34:32.0000] MikeSmith: I suspect he doesn't like CORS and thought fetch() would give a way out [02:34:45.0000] JakeA, I guess I can merge [02:35:43.0000] Ms2ger: I can't see how that change could make other tests unstable, but browsers have surprised me before [02:36:03.0000] JakeA, it probably just made the checker run a test it hadn't before [02:39:32.0000] Thanks for merging! [02:42:15.0000] annevk: thanksーand yeah I have no clue what he might actually be trying to do. hmm but yeah I guess he saw mode: no-cors and thought that meant no CORS restrictions [02:42:35.0000] I wonder now how often other people might think that [02:42:45.0000] it isn't the first time I've seen it [02:43:09.0000] naming :-( [02:43:12.0000] Seen a few people think it's a "security? Pfft, no thanks" switch [02:43:31.0000] hah [02:44:11.0000] I guess we should have some better simple documentation (at MDN) that clearly explains what "no-cors" really means. There doesn’t actually see to be something like that at MDN yet [02:45:15.0000] Fetch is poorly understood by developers [02:45:39.0000] I can't say I understood it well before service worker [02:45:49.0000] Presumably understanding was worse before it was written down [02:45:58.0000] yeah [02:46:32.0000] There used to be no answer to what happens for redirects to