2014-10-01 [17:03:18.0000] SimonSapin: maybe you can just have an informational slot for the original? [17:04:22.0000] jgraham: I think it's likely as true as many sentences of the form "Browser X has shown that Y is possible" [17:04:38.0000] but those sentences are said incorrectly in many cases ;) [17:07:16.0000] I believe the original sentence by dherman was of the form "Servo is doing some interesting experiments that seem to be working ok so far. They may be worth looking into" [17:15:00.0000] That formulation of the original seems reasonable. The broader form is less reasonable for servo since it is generally compatible with almost no web content so it's impossible to measure compat losses [17:25:44.0000] I don't think that's an accurate assessment re: compat [17:25:51.0000] but you may have a very high bar [17:26:29.0000] how can you not have a very high bar after Array.prototype.contains [17:26:30.0000] /me cries [17:27:38.0000] (https://bugzilla.mozilla.org/show_bug.cgi?id=1075059 for those following along at home) [17:29:09.0000] what features are looking at being secure sites only? [17:29:17.0000] i know Service Worker is one [17:30:05.0000] terinjokes: annevk is hoping we can move privacy-sensitive ones like geolocation and videocam access into secure only over the next year or so [17:30:19.0000] terinjokes: probably EME [17:30:50.0000] terinjokes: in Chrome, WebCrypto, since our security guys say it is bad to give the impression of a site using crypto when in reality they could be MitMed. But Firefox disagrees and is shipping it everywhere. [17:31:43.0000] interesting [17:33:14.0000] also in general http://www.chromium.org/Home/chromium-security/prefer-secure-origins-for-powerful-new-features [17:33:57.0000] speech API should be probably secure only [17:34:02.0000] ah thanks, I saw that at Edge, but didn't have a link for it [17:34:20.0000] i wonder if requestAutocomplete is restricted... [17:50:16.0000] for some reason i thought es6 modules, but looks like I'm wrong [00:21:43.0000] about fetch/CORS behavior, is the server supposed to send the Access-Control-Allow-Origin for a 304? [00:42:55.0000] Domenic: it is [00:43:09.0000] Domenic: https://wiki.whatwg.org/wiki/TLS tracks what I could find [00:44:54.0000] terinjokes: ^ [00:45:24.0000] MikeSmith: yeah, why not? [02:05:44.0000] i guess it's too late to make webrtc https only [02:08:09.0000] I dunno, is it? [02:12:15.0000] Depends if anyone is using in production on non https, doesn't it? [02:13:00.0000] And how long that will last if we deprecate it, but I thought P2P stuff was already TLS-restricted [02:23:17.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/saved/3223 - getUserMedia seems to be allowed on insecure in gecko/blink/presto at least (but it's prefixed in gecko/blink) [02:27:54.0000] zcorpan: oh, getUserMedia(), yes, there's a thread about that on public-media-capture that I still need to reply to [02:58:46.0000] annevk: for http://people.w3.org/mike/tests/fetch/ an Access-Control-Allow-Origin is sent wwith 200 responses as expected but not for 304 responses [02:59:21.0000] MikeSmith: so you won't get to read the 304 response [03:00:10.0000] oh? [03:02:41.0000] annevk: ah because the browser just revalidates the cache entry? [03:03:50.0000] MikeSmith: so this might actually be wrong in Fetch I guess [03:04:22.0000] MikeSmith: currently we do a CORS check before we handle response statuses [03:05:49.0000] MikeSmith: but even for redirects we require CORS checks to be done [03:05:57.0000] MikeSmith: and this is effectively a redirect, so... [03:06:38.0000] ok [03:07:29.0000] MikeSmith: are you encountering an issue? [03:11:03.0000] annevk: I had been thinking it was a bug because of seeing "XMLHttpRequest cannot load ... No 'Access-Control-Allow-Origin' header is present on the requested resource." when testing [03:11:59.0000] but now I realize maybe it was just because I was getting a stale cached copy from prior to when the .htaccess file was changed to add the header [03:35:24.0000] jungkees: I'm not sure if it's worth repeating, but what you should focus on is the model [03:35:39.0000] jungkees: what objects are in play, what is their lifetime, what objects are they associated with [03:35:52.0000] jungkees: and then define that model in detail [03:36:20.0000] jungkees: and then when you define something like the installing attribute, it will simply return one of the objects in that model [03:37:09.0000] jungkees: the way you are writing the service worker specification now will not get you to the finish line [03:37:39.0000] Anyone here happen to have a coupon code for the W3C validator suite? registering now >_> [03:40:57.0000] I suspect MikeSmith does, but he's prolly not allowed to share those :-) [03:41:20.0000] annevk: Thanks for the comment. I'll soon have time to revisit that point. [03:42:04.0000] jungkees: no rush btw, please do enjoy your time off or whatever it is you're doing now :-) [03:42:21.0000] is there a room similar to #whatwg but for CSS spec? [03:42:29.0000] jungkees: if you need help let me know [03:42:41.0000] alystair: irc.w3.org has #css [03:42:49.0000] I need to know who to blame for inner box-shadow rendering on [03:42:53.0000] annevk: definitely. I need your help over time :-) [03:43:13.0000] (or lack of it) [03:43:18.0000] thanks [03:44:15.0000] alystair: I Just use the free validator, and for batch validation, the jar from https://github.com/validator/validator.github.io/releases/latest [03:46:58.0000] Woah, the W3C moved into validator sales? [03:48:06.0000] perhaps W3C mug sales were not as brisk as they'd hoped? ;) [03:48:33.0000] Oh does that explain why the W3C is still full of mugs? [03:48:37.0000] jgraham: apparently getting paid by members was not enough [03:48:40.0000] (sorry ;) [03:48:58.0000] too much coffee not enough tea~ [03:49:04.0000] I use my W3C mug regularly for tea [03:49:20.0000] I got mine from Amy though [03:49:58.0000] (I didn't actually know they had mugs, more humorous than I expected) [03:50:10.0000] Now I kind of want a WHATWG mug [03:50:44.0000] Has anyone figured out how WHATWG green translates to print? [03:51:06.0000] I imagine it looks just as bad as on screen :) [03:51:06.0000] probably just about as ugly? [03:51:09.0000] heh [03:52:09.0000] "ugly mug" would have a much more literal meaning :D [03:52:12.0000] We made a t-shirt once, that almost nobody understands. I think I still have mine somewhere and I see hober wearing it sometimes [03:52:18.0000] (back on the subject of the validator suite, those "testimonials" look like filler content) [03:52:51.0000] createMathMLDocument() huh. http://www.w3.org/TR/MathML2/appendixd.html [03:53:12.0000] jgraham: hey I spent a lot of time writing those testimonials [03:53:44.0000] It looks they're still for sale: http://five-gt-two.spreadshirt.com/ [03:56:47.0000] We even have wallpapers: http://whatwg.majda.cz/wallpapers/ [03:58:18.0000] I was hoping for the kind of wallpaper you can use to decorate your home [04:10:06.0000] heh [04:51:53.0000] I looked into the style sheet for specifications other than HTML. It seems that for wider screens we could float the table of contents to the right. Perhaps put a border around it or some such or some a wide enough margin. [04:52:22.0000] We could also remove the gap on the left. HTML uses that for infoboxes but no other spec has infoboxes at this point and unless Hixie generalizes the API I don't see that happening. [04:52:54.0000] zcorpan does place the "select text to file a bug" widget in that left margin, but I'm sure we can think of something. [04:53:25.0000] Removing the wide left margin would also allow us to fit more text on the screen for mobile devices. Apparently people do read specs on their phones. [04:59:51.0000] Oh, back in '96 W3C used the MIT license: http://www.w3.org/TR/REC-png-961001#Credits [05:00:58.0000] Of course, even that was a regression from: http://www.w3.org/Policy.html [05:06:08.0000] wow 5 > 2 that is kind of subtle these days [05:06:44.0000] i think i don't have mine anymore [05:07:17.0000] Domenic: yeah, it comes off almost ironic now [05:07:53.0000] I guess today it comes down to a battle between 5 > null and 5 > undefined ;) [05:07:56.0000] heh yeah [05:11:48.0000] broken link at the bottom of https://whatwg.org/specs/. /cc Hixie I guess [05:16:21.0000] jgraham: oh man those testimonials really are about completely unispiring as they could be [05:16:48.0000] jgraham: I think it just needs pictures, like at http://www.theonion.com/articles/new-antifacebook-social-network-ello-boasts-lack-o,37035/ [05:17:29.0000] jgraham: hey, could even borrow even do a variation on the “No advertising is nice, but what really appeals to me is the lack of users.” quote there. Good fit [05:20:20.0000] So, when you iterate over FormData you get an Array that consists of DOMString and a File? Does that seem appropriate, Domenic? [05:20:27.0000] what testimonials? [05:21:02.0000] annevk: not an array, an iterator. but yes. [05:21:47.0000] Domenic: well each iterator value would be an array due to lack of tuples, no? [05:21:50.0000] zcorpan: middle of https://validator-suite.w3.org/ [05:21:55.0000] annevk: ah right i misread [05:22:28.0000] For URLSearchParams, it would be DOMString, DOMString; Headers would be ByteString, ByteString [05:22:36.0000] /me peruses https://github.com/whatwg/loader [05:25:26.0000] annevk: agree [05:25:42.0000] annevk: that is the default iterator; you also want keys() and values() iterators [05:28:48.0000] MikeSmith: LOOOOOOOOOOOOOOOOOOL [05:30:04.0000] "enhanced validity" [05:30:24.0000] Too bad Extended Validation is already a thing [05:31:09.0000] oh man I guess I'd be better off having not tried to actually read that page [05:31:38.0000] annevk: "enhanced validity" is going to be the tagline for W3C's new unsolicited email marketing campaign [05:32:25.0000] jgraham: that was considered but I think the consensus decision was to go with "now with turbo boost" [05:34:18.0000] annevk: trying to figure out how to fit "Reasonably unfiltered" into the messaging there [05:34:20.0000] "can't perform in your text editor? Women laugh at your markup? Get enhanced validity now" [05:34:32.0000] hahah [05:36:23.0000] Domenic: https://www.w3.org/Bugs/Public/show_bug.cgi?id=26183#c17 [05:36:56.0000] jgraham: that's a bit sexist [05:38:04.0000] annevk: I was thinking of viagra spam, which is generally marketed at men [05:43:36.0000] annevk: seems similar to the MapLikeIterator from that other bug [05:43:44.0000] annevk: which was blocked on figuring out what the abstract model should be [05:43:59.0000] annevk: "blocked" = "I was too lazy to propose that part because it is harder" [05:48:24.0000] Domenic: perhaps at this point IDL should just expose the hooks for iterators as I suggested in the end and we can define the templates ourselves [05:48:42.0000] Domenic: templates for multimap-like, set-like, etc. [05:48:57.0000] Domenic: and then much later uplift the whole thing to something better [05:50:20.0000] seems reasonable I guess [06:09:06.0000] annevk, what exception for incorrect invocation? [06:09:16.0000] we have no exceptions on that code path (N.requestPermission) [06:09:29.0000] beverloo: e.g. Notification.requestPermission("TEST") throws I think [06:09:43.0000] beverloo: you haven't studied the full code path then ;-) [06:10:14.0000] ah, binding magic [06:23:10.0000] Isn't the conversation that's currently playing out on whatwg@ exactly what Hixie complained about with the use of rejection for exceptional conditions? [06:34:52.0000] jgraham: don't think so [06:35:49.0000] annevk: OK, well I mailed the list so feel free to explain why I'm wrong [06:37:39.0000] maybe later [06:41:41.0000] Now I feel rejected :p [06:50:08.0000] annevk: FWIW I feel like it comes down to names, and "request" is ambiguous [06:50:22.0000] If it it was "getPermissionFromUser()" then I would say not getting permission is exceptional [06:50:54.0000] if it was "doesUserAllowCameraAccess()" then it would be clearly unexceptional [06:51:03.0000] but "requestPermission()" could go either way [07:13:23.0000] link to the bug on speccing animation frame tasks and tying CSS animations into all that? [07:30:00.0000] jgraham, hsivonen: my gut says that the fragment parsing tests should probably be in a separate test file; thoughts? [07:57:29.0000] Domenic: https://www.w3.org/Bugs/Public/show_bug.cgi?id=26839 is the latest on animation frame tasks [07:58:17.0000] Domenic: if you ignore the name, since whether we call it requestMoon() or getMoon() seems hardly relevant, what behavior do we want? What code do we want people to write? [07:58:36.0000] jgraham: hah, just deferred [08:24:55.0000] JakeA: https://github.com/slightlyoff/ServiceWorker/issues/412 ping [08:29:30.0000] looking [08:30:26.0000] JakeA is out (vacation) IIRC [08:31:41.0000] slightlyoff: you're more than acceptable as stand-in :p [08:41:23.0000] TabAtkins: I feel like you should talk to Domenic about how to sort that mismatch [08:41:52.0000] TabAtkins: I agree with you that it makes a lot of sense to group all "failure", but then the async/await path really sucks [08:59:17.0000] I think that the described behaiour is fine, annevk [08:59:32.0000] slightlyoff: and the new name? [08:59:34.0000] annevk: is the name change related to the value of Cache-Control? or of the option to fetch? [08:59:43.0000] slightlyoff: fetch [08:59:51.0000] cache-control is not really affected in any way I hope [08:59:59.0000] I'd like it to not have a dash = ) [09:00:06.0000] but otherwise I don't mind [09:00:25.0000] it seems fine. Defer to Horo-san on actual behavior, but this seems fine WRT our impl [09:14:11.0000] https://twitter.com/CDN_Antitheist/status/517252789680877568 [09:15:02.0000] Ms2ger: those replies lol [09:29:53.0000] annevk: I agree that it sucks. [09:30:34.0000] annevk: Though we could always add a Promise.prototype method that shifts *some* rejects into fulfills, when they're tagged as "not fatal" or something. [09:30:50.0000] await geo.request().lessThrowing() [09:31:04.0000] Or something else like that, I dunno. [09:31:29.0000] I guess Domenic's point is that it should be the other way around [09:31:32.0000] Maybe a different keyword, one that only rethrows "fatal" rejects, and one that throws all of them. [09:32:04.0000] annevk: Yeah, but his "other way around" is optimized for `await`, not promises. We can fix one of them, the other's frozen. [09:38:22.0000] I'm rather aligned personally with the return/throw analogy. [09:38:59.0000] (And to be clear, a network error seems like something you'd throw for since that is exceptional. Normally that works.) [09:41:07.0000] Hmm, depends what you mean by "network error" [09:41:48.0000] I'm familiar with APIs that throw for socket errors, but less so with ones that would throw for e.g. a 500 HTTP response [09:42:23.0000] jgraham: network error is defined in https://fetch.spec.whatwg.org/ [09:42:30.0000] annevk: I thought that fetch() fulfilled for failure responses. [09:42:44.0000] TabAtkins: failure being? [09:43:46.0000] Hm, from a quick reading of the definition of "network error" in the spec, I'm not sure. It doesn't go into any detail about what sorts of things cause network errors. [09:44:04.0000] Yeah, you need to read through the spec [09:44:20.0000] UGH WHY IS IT SO HARD [09:44:31.0000] But typical scenarios include the network failing, CORS failing, infinite redirects [09:44:45.0000] HTTP 500 is not a network error [09:45:00.0000] The network is fine, it's the server that's not [09:45:05.0000] Ah, kk [09:45:43.0000] Really? https://fetch.spec.whatwg.org/#cors-preflight-fetch-0 sounds like it is, but I might be missing something [09:47:54.0000] jgraham: that would be CORS failing [09:48:30.0000] jgraham: CORS is a protocol layered on top, but any failures have to be masked as network failures for security [09:48:33.0000] Oh I see [11:47:14.0000] so i'm working on replacing the in-spec feature boxes with just importing the caniuse data [11:47:27.0000] anyone have any opinions on this? [11:47:37.0000] e.g. should I just look at the state of the latest version? [11:47:40.0000] most widely used version? [11:47:45.0000] any version? [11:47:50.0000] (version of a browser, i mean) [11:47:57.0000] i'm leaning towards "latest" [11:48:19.0000] Latest [11:48:24.0000] k [11:49:47.0000] i'm thinking that i should get the N most popular browsers, probably N=5 since we have 5 icons right now in the box, and then i just show the icons of the the subset of those browsers that support the feature [11:49:57.0000] instead of the current hidden icon / shown icon thing [12:05:53.0000] Loving unimpressed marcosc [12:06:32.0000] Hixie: while you do that, can you consider a change in styling that makes them floats and uses the sidebar for text rather than mostly whitespace? [12:06:52.0000] Hixie: in particular on devices with smaller screens, it feels a bit like a waste [12:07:20.0000] Hixie: I might fork the style sheet (or override) for a while to do something else on other specs [12:07:52.0000] how do you mean? [12:07:59.0000] the margin's roughly the same on both sides, no? [12:08:02.0000] Hixie: I think what you should show is since when the feature is supported on a per browser basis [12:08:20.0000] Hixie: no, there's an 8em left margin [12:08:34.0000] Hixie: very clear in e.g. https://fetch.spec.whatwg.org/#acknowledgments [12:08:59.0000] Hixie: outside of HTML that margin is only used for that black arrow [12:09:20.0000] oh, i see, there's a max-width and i happen to have a monitor wide enough that it gives a margin on both sides [12:10:14.0000] i guess we could move the margin to the right, but that doesn't seem like it'd help much... [12:10:24.0000] floats don't work very well so i'm reluctant to use those [12:10:42.0000] no shifting the margin is not good :-) [12:10:44.0000] and floats don't let you pin the box to the top of the screen [12:10:57.0000] hmm true [12:11:41.0000] if you want i can point the spec to a style sheet you control so you can play with it and see if you can come up with something that works but is better [12:11:42.0000] Hixie: do you care to maintain the style sheet for non-HTML specs as well? perhaps the boxes layout can be opt-in? [12:12:09.0000] imho we should keep all the specs consistent [12:12:16.0000] ideally, we'd have all the specs have this caniuse data [12:12:39.0000] Hixie: that sounds great to me, can you write the software more generically this time? :-) [12:12:43.0000] though since i'm baking it in at spec generation time to make the html spec load faster, it's not something you'll get for free [12:12:54.0000] the html spec generator is pretty generic, i think? [12:13:12.0000] is that public now? sorry I haven't been paying attention much [12:13:34.0000] I'm happy to do work to make this work in other specs, either through a gives / for me [06:32:45.0000] rubys: in IE10 [06:32:57.0000] rubys: using Live DOM Viewer [06:34:27.0000] /me checking my doctype [06:35:11.0000] (adding doesn't change anything) [06:35:42.0000] here is my script: http://intertwingly.net/stories/2014/10/05/urltest [06:36:14.0000] "In IE10 Standards mode, the host and pathname DOM properties no longer return unexpected results." [06:36:22.0000] http://blogs.msdn.com/b/ieinternals/archive/2012/03/01/ie10-beta-consumer-preview-minor-changes-changelist.aspx [06:36:36.0000] any idea why my script wouldn't be treated in standards mode? [06:38:44.0000] annevk: I'm running IE11, but I would hope that MS wouldn't have reverted that. [06:39:26.0000] rubys: I think the results might differ if you create the element through script [06:40:27.0000] [06:40:31.0000] gives "" in the log [06:40:40.0000] (I also tried setting .href, same result) [06:40:41.0000] annevk: yuk [06:40:48.0000] agreed [06:42:47.0000] And you're testing in IE11? [06:43:03.0000] Wow, they're bad at fixing bugs, perhaps because Eric left the company [06:43:56.0000] Or maybe they fixed it but then decided they could not ship it due to compat [06:44:10.0000] annevk: I'm testing using a VM from here: https://www.modern.ie/en-us/virtualization-tools#downloads [06:44:26.0000] IE11 / Win8.1 [06:46:04.0000] annevk: other things I have done recently: I've added Rust and Node.js to my results: http://intertwingly.net/stories/2014/10/16/urltest-results/ [06:46:24.0000] cool [06:46:43.0000] annevk, SimonSapin: I would appreciate somebody giving http://intertwingly.net/stories/2014/10/16/urltest.rs a glance to see if I got it right, or if it could be done better. [06:46:58.0000] After all, I didn't know Rust as of two days ago. [06:47:35.0000] I haven't written any Rust yet so I'll let SimonSapin do that [06:47:43.0000] I also attempted a manual three-way merge of the urltestdata from rust-url and web-platform-tests [06:48:22.0000] https://github.com/servo/rust-url/pull/39 and https://github.com/w3c/web-platform-tests/pull/1293 [06:50:40.0000] For the spec there's these projects on the todo: 1) Rewrite parsing in a more functional style 2) Introduce IPv4 parsing 3) Investigate if we can allow non-browser schemes make use of "relative url" parsing facilities in a generic way [06:51:57.0000] since rust-url seems to be a motivating factor for #1, I'm looking into rust-url. [06:53:26.0000] If I can wrap my head around it, I may try to help with #1 [06:55:24.0000] Yeah, SimonSapin proving it could be done just as efficient with a clearer style is the main factor for that one [07:20:55.0000] rubys: I can reproduce my IE10 results in IE11 [07:21:37.0000] so works, document.createElement doesn't? [07:39:14.0000] yup [07:39:46.0000] I don't want to know what their code looks like [07:40:07.0000] luckily you can't :-) [07:41:26.0000] rubys: what do the colours mean now? Can you include a legend on the page? [07:42:27.0000] I'll add a legend. Meanwhile, here is what the colors mean: http://intertwingly.net/blog/2014/10/02/WHATWG-URL-vs-IETF-URI#c1412684307 [07:42:39.0000] ... where 'testdata' is substituted for 'whatwg' [07:43:28.0000] I got pushback when I called testdata 'whatwg'. I got pushback when I called anything not testdata 'whatwg'. So I gave up :-) [07:45:51.0000] rubys: ericlaw suggests a workaround, appendChild() the element and then check its pathname [07:47:06.0000] Or, I can simply change the test to use an element. [07:56:26.0000] rubys: what does the "notes" field contain? I'm guessing, the environments which fail to match? [07:57:17.0000] if you click through, you will see one or more issues with that user agent [07:57:30.0000] so essentially, yes [07:58:27.0000] cool. will be interesting to see the results with IE fixed [07:59:07.0000] I'll run that in a few minutes [07:59:49.0000] rubys: "addressable" is what was formerly IETF? [08:00:01.0000] annevk: yes. [08:03:22.0000] rubys: any reason why we include Opera Presto? [08:03:47.0000] s/we/you/ [08:04:37.0000] zcorpan felt it should be included. http://intertwingly.net/blog/2014/10/02/WHATWG-URL-vs-IETF-URI#c1412536749 http://intertwingly.net/blog/2014/10/02/WHATWG-URL-vs-IETF-URI#c1412583899 [08:07:05.0000] addressable is just a Ruby library right? [08:07:46.0000] yes, but it is one that I have confidence that it attempts to faithfully implement the IETF RFCs. [08:12:03.0000] IE results updated, generating new results [08:14:00.0000] annevk: any reason why resolving a URL against a base couldn't be factored out in the URL standard? In other words, define what parsing a URL means,and define what resolving a URL against a base means? [08:15:15.0000] annevk, Domenic: new results uploaded. Changes in IE results, table header, and new legend at the bottom of the index page. [08:17:04.0000] rubys: that would have a pretty major impact on what a URL was [08:17:10.0000] is* [08:18:11.0000] rubys: I also haven't found that any browser actually does that [08:18:13.0000] annevk: can you give an example where it would be difficult to do and/or operationally it would produce different results? [08:18:52.0000] rubys: well e.g. new URL("/test") would no longer throw, presumably? [08:19:27.0000] presumably. Do any browsers expose URL? [08:19:54.0000] all of them do by this point. except IE maybe [08:21:08.0000] Still so many IE failures :( [08:23:36.0000] rubys: I get .protocol to work in IE only for an element that was already in the document [08:23:51.0000] rubys: so many bugs there :-( [08:24:47.0000] rubys: at least Firefox/Chrome/Safari support new URL() [08:25:04.0000] rubys: also Opera, not sure about Presto [08:25:05.0000] annevk: I presume that you have whatever access you need to do @ericlaw suggested, i.e. "Please file a bug on Connect?" [08:25:29.0000] Yeah I guess I can do that again... [08:31:19.0000] https://github.com/whatwg/fetch/issues/11 is interesting [08:31:25.0000] -1 on ES5 syntax of course [08:31:31.0000] unsure on whether early examples make sense [08:31:40.0000] given that the spec is only in a small part about developer-facing API [08:31:44.0000] https://connect.microsoft.com/IE/feedbackdetail/view/1002884/protocol-attribute-does-not-always-work [08:32:49.0000] Hmm I thought I had issues disabled [08:33:15.0000] yeah that was surprising as well [08:47:20.0000] rubys: note that new URL() not dealing with it does not necessarily preclude us from structuring the specification in that way, though there might be other pitfalls given that implementations have no such thing [08:48:31.0000] My thoughts are that it might be simpler to spec that way; but it will take actually trying to determine if there are any pitfalls. [08:50:15.0000] my line of thinking: I was reading rust-url, and was wondering what it would look like in ABNF. Then I realized that I don't know how to describe parsing two inputs in ABNF. Also, many libraries aren't structured this way. [08:52:40.0000] If you exclude the most widely deployed libraries? [08:54:02.0000] annevk: no question new URL() behavior shouldn't change. The question is whether a cleaner and simpler spec could be structured differently. [08:55:00.0000] No I meant that browser implementations have those operations not decoupled either [08:57:07.0000] But it's worth figuring out I suppose. A handful of people have asked for that and there might be some cases where you want to parse a URL without a base URL. (We want to be certain on that might before doing this, obviously.) [08:57:56.0000] Doesn't it make more sense to just make parsing "without a base URL" a special case of parsing with a base URL? [08:58:05.0000] e.g. parsing "without a base URL" = parsing with base URL "about:blank" [08:59:05.0000] Domenic: see above. "Then I realized that I don't know how to describe parsing two inputs in ABNF." [08:59:16.0000] why is that important [08:59:28.0000] seems conceptually less simple and clean [08:59:57.0000] Breaking a hard problem into two smaller problems is less simple and less clean? [09:00:12.0000] creating two separate problems instead of making one a special case of the other, yeah [09:02:15.0000] Domenic: rubys is suggesting that parsing "/foo" would result in some kind of construct that the specification currently does not acknowledge [09:02:51.0000] Domenic: the specification does indeed default to something akin to about:blank at the moment, return failure for "/foo" as input [09:02:55.0000] returning* [09:41:06.0000] annevk: yeah, seems very much like two problems, as now you have to create a whole new set of data structures to represent not-really-URLs like "/foo" [09:42:00.0000] Is "scheme" exposed by browsers, or only "protocol"? [09:42:51.0000] looking at the Rust implementation, it has an intermediate representation for relative URLs. [13:42:26.0000] http://www.apple.com/ipad-air-2/wireless/ This new Apple SIM card thing seems interesting. Seems they finally found enough support to do the thing they have wanted to do since the beginning, although more modestly still. Can't wait for SIM cards to finally die in a fire. [13:43:11.0000] (I wanted to link directly to the SIM card bit, but although there's many data and class attributes there, no id attribute to be found.) [13:45:11.0000] that is nice [13:48:30.0000] would like feedback on spec-writing matters: https://github.com/whatwg/streams/pull/233 [13:51:29.0000] annevk: http://www.apple.com/ipad-air-2/wireless/##Apple%20SIM [13:52:08.0000] tantek: how does that work? [13:52:52.0000] annevk: http://indiewebcamp.com/fragmention [13:53:05.0000] It doesnt' work, he's just using a fragmention scheme that could maybe work in the future. [13:53:08.0000] it doesn't seem to actually work [13:53:18.0000] e.g. http://indiewebcamp.com/fragmention##Chrome%20extension works [13:54:04.0000] works with polyfill just fine TabAtkins Domenic [13:54:30.0000] sure but you implied that the link you actually sent worked [13:54:54.0000] Domenic: not at all, merely answering the use-case of "I wanted to link directly to the SIM card bit" - which I did [13:55:27.0000] and if your browser supports it (e.g. via an extension), then that apple.com link even "works" [13:55:56.0000] hence why I linked specifically to the Chrome extension [13:58:37.0000] Domenic: gave some feedback, not sure if it helps [13:59:35.0000] annevk: heh, yeah, not sure either. thanks though. [14:02:26.0000] Domenic: looking closer, it seems your current approach avoids duplication [14:02:53.0000] annevk: I don't think so though. Because I only ever call each method once in the spec. It avoids branching in favor of polymorphism I guess. [14:03:48.0000] Domenic: the behavior of %DefaultReadableStreamStrategy% would not be duplicated if you branch? [14:03:59.0000] annevk: right it would just be inlined [14:04:13.0000] Domenic: I'd prolly do that then [14:05:21.0000] Domenic: but if it's not observable and both are sufficiently clear, it's really up to you [14:05:40.0000] annevk: yeah I am leaning toward inlining. [14:05:50.0000] (updated the post with links to the places where it's assigned and used) [14:06:00.0000] annevk: but yeah I think it is subjective which is why I am asking around [14:11:14.0000] Domenic: I appreciate the attention to detail, I hope in due course we come up with some rules to align other specifications with [14:11:29.0000] annevk: that's the hope ^_^ [14:23:08.0000] TabAtkins: I am trying to port over the self-link styling to the WHATWG style sheet. Do you remember where font-size: 83% comes from? [14:23:59.0000] It looked okay. [14:24:08.0000] figured [14:24:21.0000] Also, it's 5/6 [14:42:14.0000] TabAtkins: are any of *.spec.whatwg.org using bikeshed? [14:43:13.0000] foolip: https://streams.spec.whatwg.org/ ! [14:43:38.0000] Just streams so far, I think. [14:43:51.0000] But more of w3 than I thought is using it now. [14:45:44.0000] cool, Streams looks very WHATWGy, like the rest [14:45:48.0000] I also like the logo :) [14:46:08.0000] Yeah, style has nothing to do with Bikeshed, you can provide whatever boilerplate and styling you want. [14:46:16.0000] Speaking of which, refresh the page to check out the sweet new link anchors next to every header and [14:46:19.0000] context: I'm not so happy with ReSpec for the WebVTT spec [14:46:32.0000] And since Domenic set up all the boilerplates and PR'd them into Bikeshed, anyone else wanting WHATWG style can use them too. ^_^ [14:46:51.0000] I've had to fix a phantomjs crash to make it actually work for automatic updates :/ [14:47:06.0000] TabAtkins: how I remember it, I pretty much just flailed around ineptly until you ended up doing the work for me :P [14:47:33.0000] Domenic: oh yeah, I love those anchors :) [14:47:33.0000] Eh, close enough. [15:23:03.0000] can anybody think of a good reason to ever use a value for tabindex other than 0 or -1? https://www.w3.org/Bugs/Public/show_bug.cgi?id=27076 [15:25:13.0000] when the controls aren't in DOM order? [15:26:55.0000] consider e.g. a paragraph with two form controls in it, and a float in the middle, with controls in it as well. You would want to navigate the controls in the paragraph, then those in the float, then those in the next paragraph, probably. [15:29:22.0000] Hixie: ok, but Marco's comment is that in practice authors who put values other than 0 or -1 are almost always (or always even) misusing it based on misunderstanding [15:29:50.0000] Hixie: so it would be good to have some actual data [15:30:05.0000] by that reasoning, we should drop most of HTML [15:31:04.0000] Hixie: That's the logic behind dropping longdesc. ^_^ [15:31:14.0000] not quite [15:31:43.0000] for longdesc the logic is that implementing it is bad for accessibility so it should be dropped in implementations [15:31:57.0000] MikeSmith wasn't suggesting dropping support for tabindex>1, just making them non-conforming [15:32:43.0000] Hixie: well short of that we can do the obsolete-but-conforming thing and I can actually have the correspnding validator warning for it include a link to the bug. i.e., in the way we talked about before for nested footnotes (I think it was that one, right?) [15:33:12.0000] yeah, non-conforming or obsolete-but-conforming [15:33:33.0000] what would the warning be? "If you are actually trying to put controls in a different tab order than the default, let us know so we don't drop support for this feature"? [15:33:40.0000] script@type is a related kind of case I guess [15:33:51.0000] Hixie: yeah, seriously [15:34:22.0000] I think it would be good to start using the validator to collect actual feedback for cases like this [15:34:26.0000] i don't know how you'd actually find out if people are using this correctly [15:34:31.0000] if they are, you'd never hear of it [15:34:32.0000] it'd just work [15:34:38.0000] and you'd never know the DOM didn't match rendering order [15:36:09.0000] dunno either, just suggesting a possible resolution for that bug [15:36:54.0000] Hixie: ah i get your point -- as an end user, you wouldn't know [15:37:41.0000] (this differs from longdesc, for example, because with longdesc you can programmatically detect the vast majority of bad uses, and compared to the good uses, they're the vast majority of all uses) [15:39:08.0000] so Marco's statement there about only seeing it misused is more about him only noticing the cases where it's misued byt not noticing the cases where values>1 are used properly, because things just work as expected for him in those cases [15:39:39.0000] Hixie: yeah (about comparison to checking longdesc misuse) I see now [15:39:55.0000] /me will add another comment to the bug [15:40:52.0000] TabAtkins: what non-standard attribute names does bikeshed .bs markup use. Is there a list somewhere? [15:41:12.0000] MikeSmith: All of its output attributes are data-* [15:41:18.0000] oh [15:41:20.0000] It only uses non-standard ones in its input format. [15:41:28.0000] ah yeah that's what I meant [15:41:32.0000] the input format [15:41:44.0000] Ah, no, there's no list. Why? [15:42:28.0000] because I can imagine that is sometimes might be useful to check the input files to find errors [15:42:33.0000] check with the validator [15:43:23.0000] and one of things I'm thinking about is, adding some standard filtering mechanism to allow users to specify certain errors to ignore [15:43:23.0000] Bikeshed's input format isn't really HTML. [15:43:30.0000] ah ok [15:43:43.0000] nm then [15:43:45.0000]