| 00:00 | <Hixie> | csuwldcat: how do you do an HTTP server with web components? |
| 00:00 | <csuwldcat> | jigga what? |
| 00:00 | <Hixie> | an HTTP server. Like Apache. How do you do that with Web components? |
| 00:00 | <csuwldcat> | I smell a rouse... |
| 00:00 | <csuwldcat> | not to be confused with rose |
| 00:01 | <csuwldcat> | ill bite |
| 00:01 | <csuwldcat> | "you dont" |
| 00:01 | <Hixie> | that's a copout |
| 00:01 | <csuwldcat> | I had binocculars on that one ;) |
| 00:01 | <csuwldcat> | c8 |
| 00:01 | <csuwldcat> | c* |
| 00:02 | <zewt> | Hixie: other people having a different view on what a proposed feature is for doesn't make them nonsensical (which is what the "http server" analogy is suggesting) |
| 00:02 | <csuwldcat> | we have a x-localstorage tag |
| 00:02 | <zewt> | :| |
| 00:02 | <csuwldcat> | and x-indexeddb |
| 00:02 | <csuwldcat> | oh, they'll blow your hair back |
| 00:03 | <Hixie> | zewt: my point was just that saying "that's a copout" is an insufficient answer because it can be applied to things that are clearly not copouts but that we both agree are valid. |
| 00:04 | <csuwldcat> | what I'm saying, is that you can say any element "shouldn't be a component" |
| 00:04 | <csuwldcat> | just to reinforce a narrative that says "most custom elements have native kin" |
| 00:04 | <Hixie> | maybe i'm wrong and HTML should be presentational again after all, but I don't see why. Without an argument explaining why Web Components in markup are a valid answer for presentational stuff, rather than being just the easy solution today but not the right solution, the answer "it's a copout" is not valid. |
| 00:05 | <Hixie> | csuwldcat: you have my cause and effect backwards |
| 00:06 | <Hixie> | csuwldcat: i'm not starting from the assumption that most custom elements have native building blocks and using the argument that presentational stuff should be in CSS to support it. I'm starting from the assumption that HTML is not presentational, and concluding from that that thus it is likely that most components will have semantic building blocks to start from. |
| 00:06 | <csuwldcat> | I have the agony of a thousand unborn developers on my mind who will curse forced, fail-prone, obtuse, fallaciously-inferred semantics and inheritance |
| 00:07 | <Hixie> | i don't know how to debate in those terms |
| 00:09 | <jamesr> | what's expected rendering for this testcase: <!DOCTYPE html><div style="width: 6px; height:50px; overflow: scroll; border: 1px solid blue"></div> |
| 00:09 | <Hixie> | jamesr: expected per spec or per reality? |
| 00:10 | <jamesr> | i'm seeing the overflow controls sticking out of the div, but in inconsistent ways |
| 00:10 | <Hixie> | jamesr: (i.e. how much detail do you want regarding scrollbar rendering) |
| 00:10 | <jamesr> | i'm looking at code so i think i want to know 'bout reality :) |
| 00:10 | <jamesr> | bonus points - rtl! |
| 00:11 | <Hixie> | well basically i think the answer is you want cropped scrollbars that are visible except on recent MacOS with the legacy pref off and on mobile iOS and Android |
| 00:12 | <Hixie> | horizontal scrollbar invisible, vertical one aligned to the left on rtl and right on ltr |
| 00:12 | <jamesr> | ok. from what i'm seeing so far doubtful i'll find anyone that matches that |
| 00:12 | <Hixie> | invisible because the bottom right (left in rtl) corner is bigger than 6px square. |
| 00:12 | <Hixie> | on recent macos the scrollbars overlap, so you'd see the bottom |
| 00:12 | <zewt> | i don't recall ever seeing a browser actually crop scrollbars, at least in windows |
| 00:12 | <Hixie> | and on mobile the scrollbars are invisible anway |
| 00:12 | zewt | tries it |
| 00:13 | <jamesr> | zewt, on linux, scrollbars aren't cropped anywhere |
| 00:13 | <jamesr> | in chromium they're sticking out of the "wrong" end |
| 00:13 | <jamesr> | and they aren't cropped |
| 00:13 | <jamesr> | in firefox-linux, they stick out of the "correct" end in LTR but don't flip in RTL and aren't cropped |
| 00:14 | <zewt> | https://zewt.org/~glenn/test-windows-ff.png gross |
| 00:14 | <jamesr> | hah |
| 00:14 | <jamesr> | that's even worse that linux, amusingly enough |
| 00:14 | <jamesr> | what flavor of windows is that? |
| 00:15 | <zewt> | win7 |
| 00:15 | <zewt> | does gecko use native controls for windows scrollbars, or does it fake it? |
| 00:15 | <jamesr> | zewt, and if you put that in <body dir="rtl">, do you get the same thing but stuck to the right edge of the screen? |
| 00:15 | <csuwldcat> | Hixie: I think you're missing the built-in motivation for developers to use extends="" |
| 00:15 | <Hixie> | csuwldcat: extends="" isn't in the page, so i'm not even looking at it. |
| 00:15 | <csuwldcat> | huh? |
| 00:16 | <Hixie> | csuwldcat: extends="" is in the web component's file, not the main HTML file. |
| 00:16 | <csuwldcat> | why? |
| 00:16 | <csuwldcat> | so? |
| 00:16 | <csuwldcat> | the UA knows |
| 00:16 | <csuwldcat> | crawlers know |
| 00:16 | <jamesr> | opera matches webkit - it sticks to the "wrong" end, isn't cropped, but does flip for RTL |
| 00:16 | <csuwldcat> | who else is there that can't know? |
| 00:16 | <Hixie> | <Hixie> requiring that tiny spiders implement web components just to be able to tell that it's an input is imho putting the burden on the wrong side of the balance |
| 00:16 | <zewt> | https://zewt.org/~glenn/test-windows-ff-rtl.png identical, just right-aligned |
| 00:16 | <Hixie> | <Hixie> my problem is that a UA that doesn't know anything about Web Components, when it sees <x-tag-completer>, has no idea what this is <Hixie> such a UA needs to see that as an <input> |
| 00:17 | <Hixie> | etc |
| 00:17 | <Hixie> | see above |
| 00:17 | <jamesr> | zewt, interestingly in the non-overconstrained case firefox does flip the scrollbar to the correct size |
| 00:17 | <jamesr> | so it's just the overconstrained case that does the LTR behavior always |
| 00:17 | <csuwldcat> | true, but I'm pretty sure people aren't going to be using components much until the browsers support them |
| 00:17 | <zewt> | https://zewt.org/~glenn/test-windows-chrome.png is that the same as what you see in linux? |
| 00:17 | <csuwldcat> | again, it's chicken and egg |
| 00:18 | <csuwldcat> | the conceptual/process overhead for using components AND completely different fallbacks is great |
| 00:18 | <jamesr> | zewt, yup, same with safari/mac. if you rtl then you should see that same thing flipped around a vertical axis |
| 00:18 | <csuwldcat> | no matter how it is specified |
| 00:18 | <jamesr> | and right-aligned |
| 00:18 | <jamesr> | in chromium, at least, we use the native theming assets but render them ourselves so we're technically capable of clipping |
| 00:19 | <jamesr> | not sure that's true for all other browsers |
| 00:19 | <zewt> | gecko looks like it does the same thing |
| 00:19 | <csuwldcat> | my bet: it's mostly an API that devs use to make more significant apps |
| 00:19 | <csuwldcat> | that said |
| 00:19 | <zewt> | since it looks exactly the same, but doesn't actually behave like native scrollbars |
| 00:19 | <zewt> | (eg. right-clicking doesn't give you the scrollbar context menu that real ones do) |
| 00:19 | <csuwldcat> | the only two browsers that have app initiatives |
| 00:19 | <csuwldcat> | are evergreen |
| 00:20 | <csuwldcat> | so the scope and duration of the issue you're trying to prevent is fleeting |
| 00:20 | <Hixie> | csuwldcat: it's not just browsers. We shouldn't require every HTML processor to also implement Web Components just to grab out what the fallback element is. |
| 00:20 | <Hixie> | imho. |
| 00:20 | <Hixie> | this is for all time, not just fleeting transition pain. |
| 00:21 | <zewt> | really wish there was a thing i could paste some HTML into, and get back screenshots from all major browsers in each OS |
| 00:21 | <csuwldcat> | are there many HTML processors that ignore link tags currently? |
| 00:21 | <csuwldcat> | not just special link rel |
| 00:21 | <csuwldcat> | normal ones that have been available for years |
| 00:21 | <csuwldcat> | are there many out there? |
| 00:21 | <Hixie> | csuwldcat: sure. Many tiny HTML processors don't fetch scripts or styles. |
| 00:22 | <csuwldcat> | what about next/previous page prefetch? |
| 00:22 | <Hixie> | hardly any browsers support that, let alone tiny UAs |
| 00:22 | <csuwldcat> | or any other number of ways docs are brought in? |
| 00:22 | <Hixie> | here's a concrete example: |
| 00:23 | <Hixie> | say i wanted to extend <h2>-<h6> in the HTML spec to do something cool like have a bug report form |
| 00:23 | <Hixie> | right now to generate the Table of Contents I run it through anolis, a tiny HTML processor that knows about <h1>-<h6> |
| 00:23 | <Hixie> | if my <h2>s start being <x-fancy-h2>, it won't work anymore |
| 00:23 | <Hixie> | if they become <h2/fancy-header> it'll work fine still |
| 00:24 | <csuwldcat> | will it work if you put <H1/fancy-header>? |
| 00:24 | <csuwldcat> | or sorry |
| 00:24 | <csuwldcat> | <dt/fancy-header> |
| 00:24 | <Hixie> | what would that mean? |
| 00:24 | <csuwldcat> | or any other tag that has a header like application? |
| 00:25 | <csuwldcat> | oh, as a developer, I'm just wondering why I can |
| 00:25 | <Hixie> | i would assume that my "fancy-header" component would be defined to extend only h1-h6 |
| 00:25 | <csuwldcat> | can't go applying that to any ol' element |
| 00:25 | <csuwldcat> | can I? |
| 00:25 | <Hixie> | i would assume that web components wouldn't allow it |
| 00:25 | <Hixie> | but that's orthogonal |
| 00:25 | <csuwldcat> | oh, so a completely obtuse fail to potential users? |
| 00:25 | <csuwldcat> | sounds fun |
| 00:26 | <Hixie> | wouldn't be obtuse |
| 00:26 | <csuwldcat> | surely it is |
| 00:26 | <csuwldcat> | the interface begs people to mix and match |
| 00:26 | <csuwldcat> | beggggs |
| 00:26 | <Hixie> | it would be an error message like "The 'fancy-header' component applies to h1, h2, h3, h4, h5, h6. Application to <dt> is not defined and the binding has not been applied." or some such. |
| 00:26 | <Hixie> | but if the author wants it to apply to <dt>, and the component is defined to apply to it, then sure, why not |
| 00:27 | <Hixie> | i would imagine many components would apply to almost any phrasing element |
| 00:27 | <csuwldcat> | lol |
| 00:27 | <Hixie> | not sure why that's funny, but glad i can entertain :-) |
| 00:27 | <csuwldcat> | if that was the case, semantics are irrelevant :) |
| 00:28 | <Hixie> | why? |
| 00:28 | <csuwldcat> | if I can apply it to anything...then what use is <div/foo> ??? |
| 00:28 | <Hixie> | depends what "foo" is |
| 00:28 | <Hixie> | if "foo" is a specific kind of control, then it probably only applies to <input> |
| 00:28 | <csuwldcat> | sure does, but not to this implementation |
| 00:28 | <Hixie> | if "foo" is "makes the title="" attribute have a fancy tooltip", then it would apply to any element with a title="" attribute |
| 00:28 | <csuwldcat> | it doesn't care if you need a kinship or now |
| 00:28 | <csuwldcat> | not |
| 00:28 | <Hixie> | not really sure what you're arguing here |
| 00:28 | <zewt> | you can apply css styles to any element, that doesn't make semantics irrelevant |
| 00:29 | <csuwldcat> | it'll force you to regurgitate it arbitrarily |
| 00:29 | <Hixie> | do you not think it should be possible to extend any element to have fancy tooltips? |
| 00:29 | <csuwldcat> | im arguing |
| 00:29 | <csuwldcat> | that if you are saying most/many could be applied to any element |
| 00:29 | <csuwldcat> | then what in the world is the leading tag name worth? |
| 00:29 | <Hixie> | "many" and "most" are not the same |
| 00:29 | <csuwldcat> | sure, a UA can read it |
| 00:30 | <Hixie> | and i just gave you a concrete example of how the tag name is critical |
| 00:30 | <csuwldcat> | but it is meaningless, worse, it could actually be *wrong* |
| 00:30 | <csuwldcat> | if slidebox has no semantics |
| 00:31 | <csuwldcat> | and I can apply it to any tag... why are you forcing me to write a random tag name in? |
| 00:31 | <csuwldcat> | or are you contending that I declare a special intent to be universally used? |
| 00:31 | <csuwldcat> | extends="all"> |
| 00:31 | <csuwldcat> | ? |
| 00:32 | <csuwldcat> | SMALLEST_POSSIBLE_TAG_NAME_HERE/foo |
| 00:33 | <csuwldcat> | "mr speaker, why do I have to write that tag name for these 5 elements? They aren't really sections, or divs?" "Well Billy, someone made us do that regardless for a few other use cases" |
| 00:34 | <csuwldcat> | "Doesn't HTMLUnknownElement have a similar proto profile as Span?" |
| 00:34 | <zewt> | ... obviously a tag name without special semantics could be recommended for use when no existing html tag corresponds to it (but that's obvious and you know that already, and your sarcasm makes it hard to take your arguments seriously) |
| 00:34 | <csuwldcat> | "Sure billy, but that just isn't good enough for fringe use-cases" |
| 00:34 | <gsnedders> | Ugh, HTMLUnknownElement. |
| 00:34 | gsnedders | wants rid of that. |
| 00:35 | <csuwldcat> | Im more on board with x/foo |
| 00:35 | <csuwldcat> | than is="" |
| 00:35 | <gsnedders> | Just another way to break existing content by introducing elements. |
| 00:35 | <csuwldcat> | Hixie: I like that the proposal changed |
| 00:35 | <csuwldcat> | because everyone will just do x/foo |
| 00:35 | <csuwldcat> | and the inheritance comes from the <element extends="input"> anyway |
| 00:36 | <csuwldcat> | either way, brevity will win out, bring on the slashes ;) |
| 00:36 | <csuwldcat> | x/slidebox |
| 00:36 | <csuwldcat> | x/flipbox |
| 00:36 | <csuwldcat> | x-slidebox |
| 00:36 | <csuwldcat> | same shizz |
| 00:36 | <csuwldcat> | well not the same |
| 00:37 | <csuwldcat> | but developer-ergo wise, it will be |
| 00:39 | <csuwldcat> | how about all the downstream fall out? |
| 00:39 | <csuwldcat> | regexps, client parsers |
| 00:39 | <csuwldcat> | etc |
| 00:39 | <csuwldcat> | what happens when some detect what they believe is an end tag? |
| 00:40 | <csuwldcat> | slash is not expected by a large number of scripts I venture |
| 00:43 | <csuwldcat> | Hixie: did you consider all that? ^ |
| 00:44 | <csuwldcat> | unintended consequence... |
| 00:45 | <csuwldcat> | if that is common, we could *literally* be breaking the web, as in scripts would error out that used to work |
| 00:45 | <csuwldcat> | never assuming both start and end tags have slashes |
| 00:45 | <csuwldcat> | I could be wrong |
| 00:45 | <csuwldcat> | are slashes allowed? |
| 00:45 | <csuwldcat> | and widely know to possibly be in start tags? |
| 00:46 | <csuwldcat> | I haven't seen it, I actually not sure on that, do you know? |
| 00:51 | <Hixie> | sorry, was afk |
| 00:51 | <Hixie> | oh, he left |
| 00:52 | <Hixie> | well anyway, i should be off. bbiab. |
| 01:11 | <JakeA> | Hey all. What effect should .innerHTML = "" have on the nodes previously in the element? |
| 01:11 | <TabAtkins> | JakeA: They're deleted. |
| 01:11 | <JakeA> | should each child be removed from the node, or each descendant removed from its parent? |
| 01:12 | <TabAtkins> | I suspect it's just each child, but I'm not sure - you'd have to check with a Mutation Observer. |
| 01:12 | <JakeA> | Trying to follow http://dom.spec.whatwg.org/#concept-node-replace-all |
| 01:13 | <JakeA> | there's a loop through descendants, but I think it's just for adding to the mutation observer |
| 01:13 | <JakeA> | Chrome, Firefox, Opera remove the children. IE removes descendants |
| 01:15 | <JakeA> | Yeah, IE's in the wrong, the descendants bit is just for adding to the mutation childList. Cheers! |
| 08:27 | <hsivonen> | interesting. the validator.nu process on html5.validator.nu had managed to die |
| 08:27 | <hsivonen> | relaunched |
| 08:37 | <Ms2ger> | gsnedders, what's up with HTMLUnknownElement? |
| 08:50 | <MikeSmith> | do you know what caused the process to dies? |
| 08:50 | <MikeSmith> | *dies |
| 08:50 | <MikeSmith> | oops |
| 08:50 | <MikeSmith> | d i e |
| 09:11 | <hsivonen> | MikeSmith: I have no clue |
| 09:12 | <MikeSmith> | hsivonen: OK |
| 11:17 | <Stevef_> | mikesmith: darobin: links to all html5 elements etc are broken in https://dvcs.w3.org/hg/aria-unofficial/raw-file/tip/index.html because URLS have changed for HTML5 editors version any chance of a fix? |
| 11:24 | <darobin> | Stevef_: they're broken but they redirect |
| 11:24 | <darobin> | I don't think we're going to reorganise the way the spec is split up, I'm pretty happy with the current result |
| 11:25 | <Stevef_> | darobin: sure, i didn't wait ;ong enough for the redirect just saw the 404 and started to whine :-) |
| 11:25 | <Stevef_> | thanks, thats fine |
| 11:25 | <darobin> | Stevef_: if you want to update that doc to have all the links to HTML be correct, it's actually possible to automate, we have the data for that |
| 11:25 | <darobin> | hehe |
| 11:26 | <Stevef_> | darobin: will ping you when i think about ding it |
| 11:26 | <darobin> | Stevef_: this script has all the data: http://www.w3.org/html/wg/drafts/html/master/fragment-links.js |
| 11:26 | <Stevef_> | cheers |
| 11:27 | <darobin> | fragment_links contains a mapping from IDs to the file in which they are contained |
| 13:08 | <annevk> | https://dvcs.w3.org/hg/webcomponents/raw-file/tip/spec/shadow/index.html#shadow-dom-example probably illustrates much of the disconnect |
| 13:38 | <hsivonen> | fwiw, next time someone complains about using non-fictitios data in examples in a Hixie spec, cite http://www.w3.org/TR/vocab-regorg/ for precedent |
| 13:38 | <hsivonen> | *fictitious |
| 13:58 | <MikeSmith> | hsivonen: what's an example of non-fictitious data in one of Hixie's specs? |
| 13:58 | <MikeSmith> | Stevef_: fwiw, for the author view I plan to keep the file splits as they were/are |
| 13:59 | <Stevef_> | mikesmith: ok thanks for the info |
| 13:59 | <hsivonen> | MikeSmith: I'm not sure, but my vague recollection is that google and yahoo were named somewhere at some point. |
| 14:00 | <MikeSmith> | hsivonen: ah, OK |
| 14:17 | <MikeSmith> | NTT Software is offering bounties to anyone who implements some open WebKit bugs related to Japanese layout / vertical text |
| 14:17 | <MikeSmith> | http://www.ntts.co.jp/offer/css-bunkakai/pop01.html |
| 14:18 | <MikeSmith> | 20万円 or most are 10万円 |
| 14:19 | <MikeSmith> | 2200 USD or 1100 USD each |
| 14:23 | <darobin> | can someone offer a bounty on the life of people who publish tables as images? |
| 14:24 | <darobin> | MikeSmith: do you know who at NTT is driving that though? |
| 14:25 | <MikeSmith> | darobin: no I don't but I'll ask |
| 14:26 | <darobin> | if you I'd love to ask them some question |
| 14:26 | <MikeSmith> | NT Software is a different business unit from the people who are actively participating in W3C WGs currently |
| 14:26 | <MikeSmith> | *NTT |
| 14:27 | <darobin> | well yeah, NTT is more like a small country than a large company |
| 14:27 | <MikeSmith> | yeah |
| 14:28 | <MikeSmith> | I think it might be a separate company, the way NTT Docomo is, only partly owned by NTT |
| 14:29 | <darobin> | the entire conglomerate is set up that way AFAIK |
| 14:29 | <darobin> | ditto NTT Data |
| 14:29 | <MikeSmith> | yeah |
| 14:29 | <darobin> | I used to work for a company where we sold stuff to NTT several times because it was easier for them to do that rather than figure out how to get the same internally |
| 14:29 | <MikeSmith> | hah |
| 14:30 | <darobin> | MikeSmith: do you know if there's a page for the testing meeting? (And are you going?) |
| 14:31 | <MikeSmith> | umm, remind me which testing meeting? |
| 14:31 | <MikeSmith> | the one in California? |
| 14:33 | <darobin> | yeah that one |
| 14:33 | <darobin> | I see pretty much no details |
| 14:34 | <MikeSmith> | darobin: so I told plh I thought you should go to that |
| 14:34 | <MikeSmith> | not me |
| 14:34 | <MikeSmith> | and he agreed |
| 14:34 | <MikeSmith> | so it's settled! |
| 14:34 | <darobin> | MikeSmith: haha |
| 14:34 | <darobin> | well I'm planning to go that's for sure |
| 14:34 | <darobin> | but you could have been there too, you never know |
| 14:34 | <darobin> | I'm just worried that since I haven't heard from it that it's still taking place — I need to book travel... |
| 14:35 | <MikeSmith> | yeah I don't know the precise details |
| 14:35 | <MikeSmith> | but I think it's the end of this month |
| 14:35 | <MikeSmith> | so yeah you should book your travel |
| 14:36 | <MikeSmith> | I guess you gotta ping plh about the date |
| 14:36 | <Ms2ger`> | Speaking of tests |
| 14:36 | <Ms2ger`> | Hi darobin :) |
| 14:36 | <darobin> | Ms2ger`: I haven't forgotten you sweetheart |
| 14:36 | <darobin> | it'll happen this week |
| 14:37 | <MikeSmith> | darobin: btw http://spec.nttdocomo.co.jp/spbss/ |
| 14:38 | <MikeSmith> | ringmark-ish type thing |
| 14:38 | <MikeSmith> | you gotta push some buttons to see the results |
| 14:38 | <MikeSmith> | but I think there's English version |
| 14:38 | <MikeSmith> | http://spec.nttdocomo.co.jp/spbss/english/index.html |
| 14:39 | <darobin> | interesting |
| 14:39 | <MikeSmith> | I haven't looked yet to see what tests it's actually running |
| 14:39 | <darobin> | might be worth looking inside to see if the tests are better than ringmark |
| 14:40 | <MikeSmith> | I suspect they are not tests that they came up with themselves |
| 14:40 | <MikeSmith> | but instead got from somewhere else |
| 14:41 | <MikeSmith> | I hope not from html5test at least |
| 14:43 | <darobin> | MikeSmith: mmmm, they're testing for the Calendar API... |
| 14:43 | <darobin> | this doesn't bode well for the rest |
| 14:43 | <MikeSmith> | hmm yeah |
| 14:43 | <darobin> | it's just an API that doesn't exist |
| 14:46 | <darobin> | they've helpfully minified all their scripts, too :) |
| 14:51 | <Ms2ger`> | Sounds exactly like ringmark, then |
| 15:01 | <annevk> | TabAtkins: thanks for explaining, but I still don't get it :/ |
| 15:02 | <annevk> | I think my Mac has this problem https://discussions.apple.com/thread/4421353 which suggests I need to replace the SSD |
| 15:02 | <annevk> | aarghfhfd |
| 15:26 | <gsnedders> | Ms2ger`: I hate it (HTMLUnknownElement). It is in the spec because I got it there. |
| 15:30 | <gsnedders> | jgraham: Any opinion on how far back we should support html5lib on Py2? |
| 15:32 | <Ms2ger`> | Fwiw, we require 2.7 now and it makes us happy :) |
| 15:33 | <gsnedders> | I think most of what we care about goes back to 2.5, and I think the 3to2 generated code deals with that. |
| 16:35 | <MikeSmith> | darobin: regardless of the NTT Software bounty thing, I think Taichi Kawabata is definitely the guy you want to talk to |
| 16:35 | <MikeSmith> | he was at TPAC |
| 16:35 | <MikeSmith> | dunno if you met him |
| 16:35 | <MikeSmith> | I don't know him well |
| 16:36 | <darobin> | the name vaguely recalls something, but I have a terrible memory of people I've met briefly vs names I've just seen in passing in email |
| 16:36 | <darobin> | in any case, good to know! |
| 16:37 | <dglazkov> | good morning, Whatwg! |
| 16:41 | <gavinc> | "just change the date on the spec every time someone checks something in" ... seriously? Don't we have computers to do that sort of thing? |
| 16:43 | <MikeSmith> | dglazkov: annevk was having a hard time understanding what the Reprojection part of the Shadow DOM spec is saying |
| 16:43 | <MikeSmith> | or what it's trying trying to say |
| 16:43 | <MikeSmith> | http://logbot.glob.com.au/?c=freenode%23whatwg&s=8%20Jan%202013&e=8%20Jan%202013#c760296 |
| 16:43 | <MikeSmith> | oh I see in the logs now you saw his comments already |
| 16:45 | <MikeSmith> | darobin: e.g., http://lists.w3.org/Archives/Public/www-style/2011Feb/0804.html has an example of double-sided ruby use case that Kawabata-san contributed |
| 16:46 | <darobin> | gavinc: where's that from? |
| 16:46 | <gavinc> | darobin: rdf-wg, so no, no one else here cares ;) |
| 16:47 | <darobin> | MikeSmith: thanks, that's an interesting case indeed (though it does seem to belong to the more rare variety) |
| 16:48 | <darobin> | gavinc: heh. is this a spec using ReSpec on the dvcs.w3 server? if so it's a known bug that the date does not match the last change automatically |
| 16:48 | <dglazkov> | MikeSmith: reprojection is a hard subject anyway. |
| 16:48 | <dglazkov> | MikeSmith: I propose shopping |
| 16:48 | <gavinc> | darobin: ding ;) |
| 16:50 | <Ms2ger`> | Well, respec is the known bug ;) |
| 16:51 | <darobin> | and that, ladies and gentlemen, is how you trap Ms2ger` |
| 16:52 | <darobin> | Ms2ger`: actually the bug is that hgweb doesn't set last modified dates correctly |
| 16:52 | <Ms2ger`> | darobin, so about those tests :) |
| 16:52 | <darobin> | I said this week my love, so it'll be this week |
| 16:52 | <Ms2ger`> | My week ends on Wednesday |
| 16:52 | <Ms2ger`> | :) |
| 16:53 | <darobin> | well I've got a whole seven days ahead of me then! |
| 16:54 | <Ms2ger`> | Dammit :) |
| 16:55 | <Ms2ger`> | Wait, I thought someone at Mozilla already wrote tests for case sensitivity |
| 16:55 | <Ms2ger`> | Why more? |
| 16:56 | <dglazkov> | annevk: insertion point can certainly be a child of a shadow host. insertion point from tree A can be a child of a shadow host of tree B. |
| 17:06 | <MikeSmith> | dglazkov: I think Anne's point was that the spec should make that clear but doesn't |
| 17:08 | <dglazkov> | MikeSmith: gotcha |
| 17:17 | <MikeSmith> | dglazkov: you coining a lot of novel terms in that spec man |
| 17:17 | <MikeSmith> | I like "reprojection" though |
| 17:18 | <MikeSmith> | I think there was a song by the 13th Floor Elevators, "I've got reprojection" |
| 17:18 | <Ms2ger`> | Was that one of your bands? |
| 17:19 | <dglazkov> | MikeSmith: thanks! I think... |
| 17:19 | <MikeSmith> | Ms2ger`: that was the band to end all bands |
| 17:19 | <dglazkov> | :P |
| 17:19 | <MikeSmith> | one dude played the electric jug |
| 17:20 | <MikeSmith> | dglazkov: to truly compete with Hixie in name-coining you need to get the term "palpable" into your spec |
| 17:20 | <dglazkov> | MikeSmith: challenge accepted |
| 17:21 | <MikeSmith> | yes! |
| 17:21 | <MikeSmith> | this is why I love standards |
| 17:21 | <MikeSmith> | well, this and the HTML working-group chairs |
| 17:21 | <MikeSmith> | hard to decide what I love more |
| 17:21 | <Ms2ger`> | Booze |
| 17:22 | <dglazkov> | hard not to love those who are brave enough to admit: "I am not a person, I am a chair." |
| 17:23 | <gsnedders> | b'<!DOCTYPE HTML>\n<meta http-equiv="Content-Type " content="text/html; charset=iso8859-2">' |
| 17:24 | <gsnedders> | Why does that expect windows-1252 as the encoding? |
| 17:24 | <gsnedders> | Because of the whitespace? |
| 17:26 | <MikeSmith> | bugga wha |
| 17:26 | <MikeSmith> | who's that b' |
| 17:26 | <MikeSmith> | some pythong thing like u' |
| 17:26 | <gsnedders> | MikeSmith: Yup. |
| 17:26 | <gsnedders> | MikeSmith: For a binary string. |
| 17:28 | <MikeSmith> | does the spec actually say waht do for that? |
| 17:28 | <gsnedders> | Sure. |
| 17:28 | <Ms2ger`> | Let's hope so! |
| 17:36 | <annevk> | dglazkov: the spec is very vague with respect to the model I think |
| 17:36 | <annevk> | dglazkov: there's some algorithms, but no model |
| 17:36 | <annevk> | dglazkov: I think defining what this means in an implementation would be much better |
| 17:36 | <annevk> | dglazkov: and would not lead to problems as we've seen thus far with e.g. relatedTarget |
| 17:37 | <annevk> | dglazkov: and if relatedTarget is a problem, so are the other MouseEvent and TouchEvent properties |
| 17:37 | <dglazkov> | annevk: sounds good. |
| 17:37 | <dglazkov> | I like the notion of explicitly defining a model in a spec. |
| 17:38 | <dglazkov> | it would fit well with the algorithms, too |
| 17:38 | <annevk> | what we want in the end for all specs is to map relative closely to implementations |
| 17:38 | <annevk> | because whenever they don't we hit issues |
| 17:42 | <dglazkov> | annevk: +1 |
| 17:42 | <dglazkov> | :) |
| 17:45 | <dglazkov> | annevk: (an excuse follows) I've been trying to get to the point where the spec is complete and correct in terms of functional machinery, and concentrated on additions and edits. Now it's time to take a holistic view and refactor. |
| 17:47 | <tantek> | is there a groundhog day meme generator? (case-sensitivity) |
| 18:05 | <annevk_> | dglazkov: looking forward to the refactor |
| 18:06 | <dglazkov> | annevk: I thought you were doing it! |
| 18:06 | <dglazkov> | :P |
| 18:06 | <annevk> | hehe |
| 18:07 | <csuwldcat> | dglazkov: thought of something |
| 18:07 | <annevk> | need to know a bit more about the details for that |
| 18:07 | <csuwldcat> | would <div/foo> break a lot of script on the web? |
| 18:07 | <csuwldcat> | im thinking regexp and parsing code for all sorts of things |
| 18:08 | <csuwldcat> | do you think it would false id the start tags as end tags? |
| 18:08 | <dglazkov> | csuwldcat: that's the concern that hsivonen raised I think. |
| 18:08 | <csuwldcat> | hmm, how can we test that? |
| 18:08 | <dglazkov> | csuwldcat: I am still brewing on this. I apologize for not having a hard opinion. |
| 18:09 | <csuwldcat> | np |
| 18:09 | <csuwldcat> | you're the calm cool and collected dimitri! |
| 18:09 | <csuwldcat> | :) |
| 18:13 | <annevk> | dglazkov: fwiw, it might help to point Hixie to the kind of example at the end of the spec |
| 18:14 | <dglazkov> | annevk: oh? |
| 18:14 | <annevk> | why is the style sheet broken on https://dvcs.w3.org/hg/webcomponents/raw-file/tip/spec/shadow/index.html btw? |
| 18:15 | <jamesr_> | annevk: [blocked] The page at https://dvcs.w3.org/hg/webcomponents/raw-file/tip/spec/shadow/index.html ran insecure content from http://www.w3.org/StyleSheets/TR/W3C-ED. |
| 18:15 | <jamesr_> | dglazkov: should be https://www.w3.org/StyleSheets/TR/W3C-ED (or just //www.w3.org/StyleSheets/TR/W3C-ED ftw) |
| 18:15 | <dglazkov> | annevk: need to make link rels point to https as well |
| 18:15 | <annevk> | oh a Chrome thing |
| 18:15 | <dglazkov> | yop |
| 18:16 | <annevk> | CSS insecure but the <img> is not? weird |
| 18:16 | <jamesr_> | it's mixed content in all browsers, i think chrome may be the only one that straight up blocks that |
| 18:16 | <annevk> | it doesn't block all of it |
| 18:17 | <annevk> | dglazkov: never mind about the example in the end |
| 18:25 | <TabAtkins> | annevk: Let me try explaining in more detail. Say you're inventing the <details> element with shadow dom. The shadow of it will look like: <shadow><content select=summary:first-of-type><div><content></div></shadow>. |
| 18:25 | <csuwldcat> | dglazkov: tags in the style of <foo/bar> aren't able to be doc.createElement generated in Firefox |
| 18:25 | <csuwldcat> | document.createElement('foo/bar'); |
| 18:25 | <TabAtkins> | (For the sake of this discussion, I'm going to use soemthing silly.) |
| 18:25 | <csuwldcat> | invalid char errpr |
| 18:25 | <csuwldcat> | meaing all of x-tag and any polylib that provides backwards compat is impossible |
| 18:26 | <TabAtkins> | Now, say you wanted a details element that only showed the even children, and there was already a component for that - <x-even-children-only>. That element's shadow looks like <shadow><content select=:nth-child(even)></shadow>. |
| 18:26 | <csuwldcat> | fails in Chrome too |
| 18:26 | <csuwldcat> | Error: INVALID_CHARACTER_ERR: DOM Exception 5 |
| 18:27 | <TabAtkins> | So, you rewrite your <x-details> shadow to be: <shadow><content select=summary:first-of-type><x-even-children-only><content></x-even-children-only></shadow> |
| 18:27 | <csuwldcat> | Hixie: <foo/bar> tags can't be polyfilled |
| 18:27 | <TabAtkins> | Naively, though (and this is how the spec worked at first), the <x-even-children-only> element only has a single child - the <content> element! |
| 18:27 | <csuwldcat> | that's a show-stopper for our apps initiative at Mozilla |
| 18:28 | <TabAtkins> | In other words, components aren't composible if they actually want to filter their contents. |
| 18:28 | <TabAtkins> | The reprojection makes it so that, when the <x-even-children-only> element looks at its children, it sees the *results* of the <content> element, rather than the <content> element itself. |
| 18:28 | <TabAtkins> | Thus, composition works. |
| 18:39 | <annevk> | csuwldcat: <foo/bar> can as it would be a <foo> element with a bar attribute |
| 18:41 | <annevk> | TabAtkins: ah okay, got it |
| 18:41 | <annevk> | TabAtkins: fwiw, using <shadow> is actually confusing (I later learned) as there's an element called that way |
| 18:44 | <csuwldcat> | annevk: what? |
| 18:44 | <csuwldcat> | are you saying that it is a boolean attribute? |
| 18:44 | <csuwldcat> | so it actuall is <foo bar> |
| 18:45 | <Stevef_> | mikesmith: didn't know you were a roky erickson fan |
| 18:45 | <csuwldcat> | annevk: ^ |
| 18:45 | <csuwldcat> | annevk: how does that escape the metric ton of problems of an is="" attribute? |
| 18:45 | <csuwldcat> | the problem set is the same |
| 18:46 | <csuwldcat> | you are now back to false inferrences of proto inheritance and mutability |
| 18:48 | <Stevef_> | MikeSmith: bull of the woods is my fave elevators album |
| 18:49 | <volkmar> | http://www.whatwg.org/specs/web-apps/current-work/multipage/common-microsyntaxes.html#valid-time-string <- does someone understand the logic of the point 4 here? |
| 18:49 | <volkmar> | having an optional part needed or not depending on what the optional part contains seems at best odd |
| 18:51 | <annevk_> | I guess it means that if second is zero you can write either :00 or nothing |
| 18:54 | <volkmar> | this is oddly expressed |
| 18:54 | <volkmar> | this should be marked as optional |
| 18:54 | <volkmar> | and if not present, seconds should be considered as = 0 |
| 18:58 | <TabAtkins> | annevk_: Yeah, I know it's confusing, sorry. I think we actually tend to use <shadow-root> to indicate it in bug conversations. |
| 19:23 | <csuwldcat> | annevk_: from now on, if people mean to describe an attribute, let's not use <foo/bar> |
| 19:23 | <csuwldcat> | it's foo[bar] |
| 19:23 | <csuwldcat> | no need to confuse ;) |
| 19:26 | <Ms2ger`> | foo@bar, duh |
| 19:36 | <annevk_> | csuwldcat: I meant when used in markup you end up with <foo bar> so you could use that for polyfill |
| 19:36 | <annevk_> | csuwldcat: if you want to use it through document.createElement you'd have to patch that method |
| 19:37 | <gsnedders> | Ms2ger`: I do like how this is the only context in which XPath is used. |
| 19:37 | <Ms2ger`> | What's XPath? :) |
| 19:38 | <csuwldcat> | [bar] has all the same ridiculous issues is="bar" has |
| 19:41 | <annevk_> | csuwldcat: I don't think you'll be winning anyone over if you keep using words like "ridiculous" |
| 19:45 | <gsnedders> | annevk: Now don't be ridiculous. |
| 19:52 | <jgraham> | gsnedders: I care not one jot about Python < 2.6 |
| 19:52 | <jgraham> | MikeSmith: Secret testing meeting :( |
| 19:53 | <jgraham> | +? |
| 19:54 | <gsnedders> | jgraham: The encoding tests are a mess. I'm not sure whether they're testing the prescan or the changing encoding. |
| 19:56 | <jgraham> | dglazkov: I suppose one option would be <input +component> |
| 19:56 | <jgraham> | Needing the space is a bit unfortunate |
| 19:57 | <annevk> | the discussion on es-discuss reveals quite a few problems btw with the notion of making the entire platform hackable |
| 19:58 | <annevk> | e.g. https://mail.mozilla.org/pipermail/es-discuss/2013-January/027952.html |
| 19:59 | <jgraham> | " |
| 19:59 | <jgraham> | That's a _really_ interesting definition of "can't do anything". |
| 19:59 | <jgraham> | " |
| 20:00 | <jgraham> | Hah |
| 20:00 | <Ms2ger`> | What he's missing is that the guy has a craving in search of a problem, I guess |
| 20:01 | <annevk> | Ms2ger`: well given that he's essentially arguing the implement the DOM-in-JS party line the exchange seems relevant |
| 20:02 | <Ms2ger`> | s/the guy/the DOM-in-JS party/ |
| 20:02 | <jgraham> | Yeah well |
| 20:02 | <jgraham> | The idea that browsers can toss out two decades of invariants isn't going to fly |
| 20:03 | <jgraham> | Regardless of how eloquently someone argues for it |
| 20:04 | <jgraham> | Doesn't mean that there aren't improvments we can make in terms of the flexbility of the platform of course |
| 20:06 | <gsnedders> | jgraham: So it appears a lot of html5lib has got broken due to tests getting broken. I'm blaming you. |
| 20:07 | <gsnedders> | And I seem to be fixing it when I try to test Py3 changes. |
| 20:07 | <gsnedders> | So oh well. |
| 20:07 | <jgraham> | gsnedders: All I heard there was "I'm fixing it" |
| 20:07 | <gsnedders> | Just making it a lot more work than I expected, given the Py2 code is broken too. :( |
| 20:08 | <annevk> | We'll get you some cake |
| 20:09 | <annevk> | gsnedders: btw, for running tests, there's no option to just ignore a file like the one for <main>? |
| 20:10 | <gsnedders> | annevk: I just did if file == "main-element": continue |
| 20:17 | <Hixie> | hsivonen: you seem significantly more conservative about the parser than i do. Is there some reason I'm missing? It seems to me that if anything, we should be less conservative about it than we've ever been before, given how good a position we're in right now with respect to parsing HTML. |
| 20:20 | <Hixie> | abarth: the security problem in the thread you're in with bz is that you can get the _getter_ for Window.document from a Window prototype, and then apply it to a Window you have that's cross-origin |
| 20:20 | <Hixie> | abarth: so the security check for _getting "document"_ is bypassed, and you need a security check in the execution of the getter itself. |
| 20:21 | <Hixie> | abarth: doesn't require assymetric checks though, that's a separate issue |
| 20:43 | <gsnedders> | jgraham: I've just pushed a ton of stuff, leading up to the new tip which regenerated the Py2 code from Py3. |
| 20:44 | <gsnedders> | So Py3 should now be the authorative copy. |
| 20:55 | <jgraham> | gsnedders: Nice |
| 21:00 | <gsnedders> | Interesting side-note about the new html5lib revision: on the speed.pypy.org benchmark data, it's 8% quicker than before with PyPy (mainly down to 2to3 and then 3to2 sorting out all the implicit str/unicode conversions) and 25% quicker than the old html5lib revision tested on speed.pypy.org; CPython shows a 14% regression over the same set of changes despite having optimization work specifically |
| 21:00 | <gsnedders> | targetting it. |
| 21:15 | <annevk_> | gsnedders: okidoki |
| 21:26 | <jgraham> | gsnedders: More optimisation from whom? |
| 21:27 | <jgraham> | annevk: Producing multipart responses in HTTP isn't exactly well supported |
| 21:27 | <jgraham> | +frameworks |
| 21:27 | <Hixie> | yeah really |
| 21:27 | <jgraham> | and maybe s/HTTP/server-side/ |
| 21:28 | <jgraham> | (I don't really like magic comments either, but I think "this is a sunk cost on the producer side" is a weak argument) |
| 21:28 | <annevk> | jgraham: given it's in <iframe> and such I'd think it is |
| 21:29 | <jgraham> | annevk: I have never used a web framework that made it easy to produce a multipart response |
| 21:29 | <jgraham> | I'm not aware of a way to configure apache to serve a resource that is several files stuffed into a multipart response |
| 21:29 | <annevk> | it was part of Comet |
| 21:29 | <hober> | MikeSmith: irc.w3.org has been really flaky for me lately. was it something i said? :) |
| 21:31 | <jgraham> | annevk: I'm not saying it's a dealbreaker, or that it's impossible to generate multipart responses. But I don't think it's really a built-in feature anywhere at present. So not really a sunk cost. |
| 21:36 | <gsnedders> | jgraham: From myself/Philip` at various points. |
| 21:45 | <annevk> | jgraham: couple of searches on Google shows that there's plenty of documentation on how to produce multipart responses in various popular languages |
| 22:01 | <jgraham> | annevk: All the pages I can see say "this isn't built in" (if you just look at things serving normal GET requests, not at comet servers) |
| 22:02 | <jgraham> | Of course I'm not disputing that it's *possible*, just that it is well supported |
