| 00:00 | <ronaldmansveld> | JonathanNeal: ARIA and microformats both use their own attributes. aria is prepended with 'aria-', microformats depend on the format your using |
| 00:00 | <ronaldmansveld> | JonathanNeal: Schema.org for instance uses itemscope, itemtype and itemprop (apart from meta-tags) |
| 00:00 | <tantek> | microformats pretty consistently use the class attribute |
| 00:01 | <tantek> | sometimes rel for page to page relationship things |
| 00:01 | <ronaldmansveld> | microformats that use the class-attribute? o_0 |
| 00:01 | <tantek> | always have |
| 00:01 | <Hixie> | ronaldmansveld: you may be confusing microformats with microdata |
| 00:01 | <ronaldmansveld> | Hixie: thanks! Indeed |
| 00:02 | <tantek> | http://microformats.org/wiki/ for more |
| 00:02 | <Hixie> | microdata is an HTML feature that uses item* attributes, microformats long predates microdata and is a way of using other HTML features for semantics to mark up similar things |
| 00:02 | <Hixie> | previously microformats was less generic than microdata, though more recently that's changed (right tantek? i'm not up to date on this) |
| 00:03 | <ronaldmansveld> | tantek: thanks! I'm gonna read up on it :) |
| 00:09 | <tantek> | Hixie, yes, microformats2 has syntax independent of vocabulary |
| 00:09 | <MikeSmith> | jpwhiting: here now |
| 00:09 | <tantek> | I'm not sure about all the weird RDFa-y features that got shoehorned into microdata. |
| 00:09 | <tantek> | in terms of being "generic" |
| 00:09 | <ronaldmansveld> | Will microformats and microdata both be here to stay? |
| 00:10 | <jpwhiting> | MikeSmith: any idea how I can make my local validator.nu always use my schema? |
| 00:10 | <ronaldmansveld> | cause a quick read/glance at the wiki gives quite the feeling that they're both trying to achieve the same |
| 00:10 | <jpwhiting> | or at least make it the default if no schema is specified |
| 00:10 | <tantek> | ronaldmansveld - given that web developers use the class attribute to markup semantics before microformats, and it's on billions of pages already, it's likely to be around a while. |
| 00:10 | <tantek> | with search engines supporting both, they're likely to be around a while |
| 00:10 | <ronaldmansveld> | tantek: fair enough, but this feels like we have two standards coming up, where only one could (and should) be sufficient |
| 00:11 | <tantek> | ronaldmansveld - actually, three standards, when you include RDFa, which predates microdata. |
| 00:12 | <ronaldmansveld> | tantek: I say overkill... it's like having 3 different specs for HTML, all trying to do the same in a slightly different way |
| 00:12 | <tantek> | ronaldmansveld, it's a long story, but in short, microformats solved a bunch of 80/20 use-cases, but was limited. RDFa was totally generic, but much harder to use. microdata reimplemented a simplified version of some 90%+ of the features of RDFa. microformats evolved to become generic. |
| 00:13 | <MikeSmith> | jpwhiting: that logic is in VerifierServletTransaction.java in the presets file |
| 00:13 | MikeSmith | takes a look |
| 00:14 | <yroc> | Hixie: What does it mean to say that an <article> is a composition on a site (not just a document)? I'm interpreting "site" as multiple pages, but I don't see how that would work. |
| 00:14 | <tantek> | ronaldmansveld, I'd say try out both microformats and microdata and see what it feels like to write in your HTML. whichever one seems to work better for you, go ahead and use it and feel free to provide feedback. |
| 00:14 | <tantek> | also - if you have any microformats specific questions, feel free to /join #microformats |
| 00:15 | <tantek> | a lot of this stuff is still evolving, and we're still figuring out better ways of doing things. |
| 00:15 | <ronaldmansveld> | tantek: I will :) So far my experience with microdata is pretty good, but I'm sure to give microformats a try as well, and yes, feedback (if any) will be provided |
| 00:15 | <Hixie> | yroc: i don't understand the question |
| 00:15 | <tantek> | thanks ronaldmansveld - appreciated! |
| 00:16 | <Hixie> | re microformats and microdata, the most important thing to worry about when picking one or the other is to pick the one that your processing software will parse :-) |
| 00:16 | <yroc> | Hixie: Referring to the defn 4.4.2, "The article element represents a complete, or self-contained, composition in a document, page, application, or site..." |
| 00:16 | <Hixie> | using microdata when your software only reads microformats will just waste your time, e.g. |
| 00:16 | <MikeSmith> | jpwhiting: in the presets.txt file, you'll notice that there are integers before the name of the presets. Give your schema the number 3 if it's a HTML schema, and 7 if it's an XHTML schema |
| 00:16 | <tantek> | Hixie - and that your teams can maintain over time (#1 cause of data failure overtime in these things) |
| 00:16 | <MikeSmith> | jpwhiting: and just remove the rest of the presets |
| 00:17 | <jpwhiting> | MikeSmith: ah, magic numbers? |
| 00:17 | <Hixie> | tantek: i mean, if your processing software supports both, then sure, that's a fine thing to pick as a tie breaker, but if your processing sotware only supports microformats, doesn't make much sense to use something else |
| 00:17 | <MikeSmith> | yeah |
| 00:17 | <jpwhiting> | then if I just submit a file to the service it still does autodetect, though, no? |
| 00:17 | <Hixie> | yroc: yes? |
| 00:17 | <Hixie> | yroc: sorry, i don't mean to be obtuse, i just don't understand your question |
| 00:17 | <jpwhiting> | though I set mine to 1 and removed the others, didn't realize the significance of those numbers |
| 00:18 | <yroc> | Hixie: LOL. What does it mean to say that the composition is on a "site" as opposed to a page? |
| 00:18 | <tantek> | Hixie, I guess. Though it used to be that search engines supported meta keywords, and nothing supported microformats - yet that changed once publishers started publishing meta keywords. |
| 00:18 | <jpwhiting> | MikeSmith: /me changes to 3 and tries |
| 00:18 | <yroc> | you're not obtuse. Compared to me you're a genius! |
| 00:19 | <Hixie> | yroc: well a "site", generally speaking, is a set of pages, right? |
| 00:19 | <Hixie> | like, cnn.com is a site with many pages |
| 00:19 | <Hixie> | but you can also make a "site" that is actually just one HTML file |
| 00:19 | <yroc> | Hixie: Right. But how can an article be on multiple pages? Or am I reading it wrong? |
| 00:19 | <Hixie> | with many "pages" |
| 00:19 | <yroc> | Hixie: Pls explain. |
| 00:20 | <Hixie> | like, twitter.com |
| 00:20 | <Hixie> | or plus.google.com |
| 00:20 | <Hixie> | it's basically one HTML file; as you navigate the "site", it loads differnet "pages" dynamically, but there's no actual navigation going on typically |
| 00:21 | <Hixie> | the URL changes only because of pushState() or using fragment identifiers |
| 00:21 | <Hixie> | so the spec is just saying that it's fine to use <article> for e.g. the posts on plus.google.com |
| 00:21 | <jpwhiting> | MikeSmith: awesome, works like a charm |
| 00:21 | <jpwhiting> | thanks a lot |
| 00:21 | <Hixie> | even though it's a "site" rather than a "page" |
| 00:21 | <Hixie> | basically it's trying to be exhaustively inclusive in its definition of what <article> can be used in |
| 00:21 | <MikeSmith> | jpwhiting: np |
| 00:22 | <ronaldmansveld> | yroc: if an article is spread among multiple pages, it isn't complete anymore (on a given page). So you can't use <article> then. The spec clearly states it has to be "complete or self-contained", which it doesn't in that case |
| 00:22 | <yroc> | I see. |
| 00:23 | <ronaldmansveld> | yroc: I'm assuming you refer to a 'page' as something with it's own URL |
| 00:23 | <yroc> | ronaldmansveld: Yes, a discrete file. |
| 00:24 | <yroc> | So when I saw site, I thought somehow it was possible to spread <article> across multiple files |
| 00:24 | <yroc> | Though I had no idea how that would work... |
| 00:25 | <Hixie> | yeah, dunno how to get around that |
| 00:25 | <Hixie> | the problem is "file" is so vague these days |
| 00:25 | <Hixie> | e.g. plus.google.com returns unique content anytime you visit it, with the "current page" prerendered |
| 00:25 | <Hixie> | and it then downloads JS to turn the page into dynamic magic |
| 00:25 | <ronaldmansveld> | Hixie: I think the 'complete and self-contained' solves the problem |
| 00:25 | <Hixie> | that's why e.g. if you try to comment on a post in g+ just after the page is up it doesn't work |
| 00:26 | <MikeSmith> | jpwhiting: yeah if you look at the logic in VerifierServletTransaction.java, there's a constant defined there called HTML5_SCHEMA, with a value of 3. and you'll see further down there's switch statement there, "switch (parser)", and for "case HTML" it does schemaId = HTML5_SCHEMA. That's the logic that handles the defaulting. |
| 00:26 | <Hixie> | ronaldmansveld: clearly not since it still confused at least one person :-) |
| 00:26 | <jpwhiting> | MikeSmith: ah, gotcha |
| 00:26 | <jpwhiting> | very nice |
| 00:26 | <ronaldmansveld> | Hixie: don't be too hard on yourself ;) |
| 00:26 | <Hixie> | it's my job :-) |
| 00:26 | <yroc> | Hixie: Well, I think a lot of the problem is that people who are reading the spec who don't have the JS background |
| 00:27 | <yroc> | DOM background |
| 00:27 | <yroc> | etc. |
| 00:27 | <ronaldmansveld> | that's a good spirit ;) |
| 00:27 | <Hixie> | yroc: yeah |
| 00:27 | <yroc> | All these dynamic possibilities that are totally off the radar for me. |
| 00:29 | <ronaldmansveld> | yroc: even though it's not complete yet, http://docs.webplatform.org/wiki/html/elements/article might be helpful :) |
| 00:30 | <Hixie> | i like how it says "W3C Working Draft" rather than "WHATWG Living Standard" |
| 00:30 | <Hixie> | "first introduced in HTML5" hah |
| 00:31 | <jamesr> | Hixie, it's a wiki, should i go fix it and get into an edit war? |
| 00:31 | <tantek> | Hixie, it is cc-by - you can fork webplatform.org and fix it ;) |
| 00:32 | <zewt> | Hixie: i think w3c spec wrecks need a laugh track |
| 00:33 | <Hixie> | turns out it's a structured database, not a wiki |
| 00:33 | <Hixie> | in particular, "W3C Working Draft" is from a list of fixed options |
| 00:33 | <Hixie> | none of which are the actual status |
| 00:34 | <Hixie> | hm, it's some sort of mediawiki frankenmonster? |
| 00:34 | <ronaldmansveld> | Hixie: I'm sure shepazu can help you with complaints about that ;) |
| 00:34 | <yroc> | ronaldmansveld: Thanks :) |
| 00:35 | <Hixie> | if you want to ignore the spec, https://developer.mozilla.org/en-US/docs/HTML/Element/article is a way better page than the webplatform.org page |
| 00:35 | <ronaldmansveld> | (or at least redirect you to one of the other people that can change those things) |
| 00:36 | <ronaldmansveld> | Hixie: that's interesting, since MDN was imported to webplatform to form the base-content |
| 00:36 | <yroc> | I went to "DOM learning materials". It's blank! :( |
| 00:36 | <zewt> | dom learning? you really don't want the DOM becoming self-aware |
| 00:37 | <Hixie> | ronaldmansveld: mdn was? or msdn? |
| 00:37 | <astearns> | ronaldmansveld: MDN was not mass-imported, there was some licensing issue that prevented it. MDN content is being moved over bit by bit |
| 00:37 | <ronaldmansveld> | zewt: thank you for making me laugh at this time of the night |
| 00:37 | <yroc> | zewt: Why not? What will happen? |
| 00:37 | <yroc> | Are we talking judgment day? |
| 00:38 | <Hixie> | so wait, webplatform.org is licensed CC-BY. Does that mean that anyone coping the article page now has to cite me? |
| 00:38 | <zewt> | window.navigator.skynet |
| 00:38 | <ronaldmansveld> | astearns: Ah, I was under the impression all of MDN was imported, sorry 'bout that |
| 00:39 | <zewt> | occasionally wishing for native vectors in js |
| 00:39 | <zewt> | (but i guess i wish for that any time i need to do vector math in any language except shading languages, which have them) |
| 00:42 | <jamesr> | zewt, assuming you mean the geometry type? do you know of many languages that have that sort of stuff built in? |
| 00:43 | <zewt> | glsl, etc |
| 00:43 | <jamesr> | what i've heard from people who work on these things a lot is nobody can ever agree on what the right APIs for vectors/matrices/etc |
| 00:43 | <zewt> | they're optimized for that sort of thing though |
| 00:43 | <zewt> | eg. Vec3(1,2,3) * 2 = Vec3(2,4,6), Vec3(1,2,3) * Vec3(.5) = Vec3(0.5, 1, 1.5) |
| 00:43 | <jamesr> | yeah, but everybody ends up with their own version |
| 00:44 | <zewt> | well, that's not really a good reason for not picking a competently-designed one and using it |
| 00:44 | <zewt> | (not to say there may not be other reasons to not have them in a language like JS, of course) |
| 00:47 | <ronaldmansveld> | anyhow, it's nearing 2 o'clock here, so I better be off to bed |
| 00:47 | <ronaldmansveld> | goodnight all! |
| 01:35 | <ojan> | TabAtkins: What happened with removing the min-width: min-content behavior from the flexbox spec? I'd really like to change this ASAP. People keep running into it. |
| 01:35 | <ojan> | TabAtkins: there didn't seem to be any opposition on the list |
| 01:35 | <ojan> | TabAtkins: are we just waiting for a F2F? |
| 02:20 | <TabAtkins> | ojan_away: Got fantasai with me today. I'll see if I can squeeze it in before we're done for the day. |
| 04:43 | <JonathanNeal> | is rel="self" not allowed? |
| 04:48 | <MikeSmith> | JonathanNeal: not in HTML |
| 04:48 | <MikeSmith> | it's an atom thing |
| 04:50 | <MikeSmith> | as in I don't think there's any spec anywhere that defines what it's supposed to mean if you use it in an HTML document |
| 04:50 | <JonathanNeal> | Yea, the validators were throwing me an error, but it was recommended to me in another channel. I am trying to markup a simple site navigation that, with currently implemented CSS, can be transformed into a variety of the different navigation styles we see online. |
| 04:50 | <MikeSmith> | OK |
| 04:50 | <JonathanNeal> | And someone said use rel="self" instead of, say, class="current" on the anchor that links to the current page. |
| 04:51 | <JonathanNeal> | made sense to me, and I was a little surprised to see it wasn't a rel. |
| 04:52 | <MikeSmith> | yeah I guess it makes some sense but the problem is there's no spec that says what that sense is |
| 05:01 | <JonathanNeal> | I hope in the future we have something like :has |
| 05:01 | <JonathanNeal> | In the meantime, basing it off the imaginary future of :has and rel="self", https://gist.github.com/jonathantneal/5096851 |
| 06:51 | <SimonSapin> | Is there any way to know which email address I used on bugzilla? |
| 08:04 | <paul_irish> | This may come across as naïve, so apologies. In the wake of Opera adopting WebKit. one of the posts that stuck out to me was David Storey's, where he indicated that the now reduction to WebKit, Gecko, and IE introduces an odd constraint on standardization |
| 08:04 | <paul_irish> | (his post: http://generatedcontent.org/post/43036827576/hey-o-lets-go ) |
| 08:05 | <paul_irish> | In both whatwg and w3c land, the buy-in from two vendors is critical. either to land something into a spec or ship a recommendation. |
| 08:06 | <paul_irish> | does the new playing field of only 3 engines introduce a big challenge in this regard? |
| 08:07 | <paul_irish> | or is this mostly moot as it's vendors rather than engines that matter? |
| 08:11 | <MikeSmith> | paul_irish: to me it's just less competition, less diversity, one fewer separate entity trying something different |
| 08:11 | <MikeSmith> | Opera was different |
| 08:11 | <MikeSmith> | well it still is |
| 08:11 | <paul_irish> | Yeah certainly. |
| 08:11 | <MikeSmith> | but Presto was different, the Presto core team was different |
| 08:12 | <paul_irish> | But you don't see this as a threat to standardization progress when it comes to multiple vendor buy-in, right? |
| 08:12 | <MikeSmith> | every browser team has its own culture that brings something valuable |
| 08:12 | <paul_irish> | perhaps it could accelerate it because eng resources are partially pooled |
| 08:12 | <MikeSmith> | ah specificallly around standardization I do see it as a bit of threat |
| 08:13 | <MikeSmith> | in that Opera has a lot of resources -- a lot of people who have their heads in the right place about standards, and Opera had a committment to investing heavily in standards development |
| 08:14 | <MikeSmith> | way out of proportion to the size of the company or how much money they were making |
| 08:15 | <MikeSmith> | hopefully a lot of those same people are going to be putting that energy into WebKit+Chromium+V8 standards features now |
| 08:15 | <MikeSmith> | and testing |
| 08:15 | <MikeSmith> | I think Opera has the best testing organization among any of the browser projects |
| 08:15 | <paul_irish> | Yeah, confirm. |
| 08:16 | <tobie> | yay testing. |
| 08:16 | <SimonSapin> | paul_irish: AFAIK in W3C-land this is more about two interoperable *implementations* than buy-in from two vendors. So two WebKit vendors still only have one implementation |
| 08:17 | <MikeSmith> | paul_irish: so would be good to try to ensure we have those people still working on testing and working on Web-platform stuff like they have been for the last many years now -- 6,7,8,9 years |
| 08:17 | <paul_irish> | SimonSapin: is that true? is implementation defined as "you did the work"? e.g. 3D transforms have multiple implementations in webkit ports, but flexbox does not? |
| 08:19 | <jgraham> | paul_irish: Yes |
| 08:19 | <SimonSapin> | paul_irish: I’m not sure but I guess so: http://dev.w3.org/csswg/css3-flexbox/#cr-exit-criteria "each implementation must be developed by a different party and cannot share, reuse, or derive from code used by another qualifying implementation" |
| 08:19 | <MikeSmith> | yeah the multiple-implementations criterion needs to be evaluated feature-by-feature, with an understanding of the multiple-ports reality and also separation of the platform parts |
| 08:19 | <paul_irish> | Ah. it gets specific. Thanks SimonSapin |
| 08:19 | <jgraham> | That was a problem for WebSQL, for example (SQLite was the only implementation) |
| 08:19 | <SimonSapin> | I think that was the issue with WebSQL |
| 08:19 | <paul_irish> | :) |
| 08:19 | <jgraham> | It seems like it will be a problem for WebRTC |
| 08:19 | <tobie> | I think it depends where in the stack the implementation is happening. The goal of this "rule" is to prove independent implementations can be built from the spec. |
| 08:20 | <SimonSapin> | there is also "Sections of code that have no bearing on the implementation of this specification are exempt from this requirement." so eg. using the same C++ compiler is fine |
| 08:20 | <zcorpan> | MikeSmith: what's the procedure for restarting the websocket server on w3c-test.org? i've changed a handler under Opera/ |
| 08:21 | <jgraham> | zcorpan: Isn't it MikeSmith-over-irc protocol? |
| 08:21 | <MikeSmith> | SimonSapin: I would not ust CSS WG CR exit criteria as a sterling model of how well things should be scoped. That language there seems overly restrictive to me, and not aligned well with how things actually work |
| 08:21 | <MikeSmith> | jgraham: heh |
| 08:22 | <paul_irish> | MikeSmith: i hope so. that exit criteria seems like it just got dramatically more challenging. :/ |
| 08:22 | <MikeSmith> | zcorpan: it involves me ssh'ing into the test machine and running /etc/init.d/apache2 restart |
| 08:23 | <MikeSmith> | which I'll doo right now |
| 08:23 | <jgraham> | +Scooby |
| 08:23 | <SimonSapin> | MikeSmith: Well, that exit criteria is there, and 3 engines instead of 4 is relevant to it. I don’t know if others WGs have other criteria. And what the CSSWG criteria *should* be is yet another issue |
| 08:24 | <SimonSapin> | MikeSmith: I think CSS WG would be open to rewording the criteria |
| 08:24 | <tobie> | SimonSapin: exit criteria are WG-specific. |
| 08:25 | <jgraham> | But generally there is a standard of two interoperable implementations |
| 08:25 | <jgraham> | And all the wiggle is in "implementation" "two" and "interopable" |
| 08:25 | <tobie> | jgraham: agreed. |
| 08:25 | <jgraham> | *interoperable |
| 08:26 | <MikeSmith> | SimonSapin: I'm saying they're free to do what they want but that places no prohibition on the rest of us from constraining exit criteria more thoughtfully. It's not a binding precedent. |
| 08:26 | <zcorpan> | MikeSmith: excellent |
| 08:26 | <SimonSapin> | MikeSmith: sure. |
| 08:27 | <tobie> | jgraham: interoperability is too long a word. Let's be hip and call it i14y. |
| 08:27 | <MikeSmith> | zcorpan: restarted now |
| 08:27 | <JonathanNeal> | tobie: hi5 / facepalm |
| 08:27 | <jgraham> | tobie: Everyone knows that the \w\d\d\w things are things wew can't get right |
| 08:28 | <MikeSmith> | tobie: I vote for icKy instead |
| 08:28 | <jgraham> | a11y - sucks. i18n - sure, if you live in the USA. So I guess i14y is a bad idea (but perhaps fitting) |
| 08:29 | <zcorpan> | MikeSmith: thanks. hmm. http://w3c-test.org/webapps/WebSockets/tests/submissions/Opera/cookies/002.html still fails for me. is something more needed than restarting to make changes to http://w3c-test.org/webapps/WebSockets/tests/submissions/Opera/handlers/ take effect? |
| 08:29 | <tobie> | jgraham it was a joke. |
| 08:29 | <jgraham> | tobie: So was my response :) |
| 08:29 | <jgraham> | Maybe I just failed at "funny" |
| 08:30 | <MikeSmith> | zcorpan: nothing more should be needed as far as I know |
| 08:31 | <MikeSmith> | lemme check |
| 08:31 | <JonathanNeal> | jgraham: you don't like a11y? |
| 08:31 | <zcorpan> | ok. then maybe it's something else being wrong with the test. damn |
| 08:31 | <JonathanNeal> | and by that I mean the abbreviation |
| 08:32 | <jgraham> | JonathanNeal: I meant that we are bad at it, not that the name sucks |
| 08:32 | jgraham | should be quiet in the future |
| 08:32 | <tobie> | paul_irish: great post on the subject of WebKit architecture/multiple WebKits, btw. |
| 08:32 | <paul_irish> | :) thanks! |
| 08:33 | <SimonSapin> | paul_irish: for your original question: apparently the details depend on what group a spec is in, but most often we seem to be counting implementations rather than vendors. So yes, I think it matters. |
| 08:33 | <MikeSmith> | zcorpan: hmm I'm seeing this error: "mod_pywebsocket: web_socket_transfer_data raised exception for /ws/set-cookie: Connection read error" |
| 08:34 | <MikeSmith> | but I'm not sure that's indicates an actual problem but I think there might be some retry that happends |
| 08:34 | <zcorpan> | MikeSmith: hmmm |
| 08:34 | <JonathanNeal> | SimonSapin: implementations = engine, or, features? |
| 08:35 | <zcorpan> | MikeSmith: the indentation looks weird in the handler, wonder if that causes ill effects |
| 08:35 | <JonathanNeal> | I guess it's implied that it would be engines, since the vendors are likely the ones cutting the features. |
| 08:35 | <MikeSmith> | zcorpan: well, it's python.. |
| 08:36 | MikeSmith | looks |
| 08:36 | <SimonSapin> | JonathanNeal: I’m not sure what you mean. At least in CSS WG, each feature needs to have at least two independent implementations, where "independent" means not sharing too much relevant code |
| 08:36 | <paul_irish> | SimonSapin: Yeah agreed. Seems like something to bring up as a topic of conversation within the WG sometime. there were 6 two-implementation possibilities before. now there are 3. |
| 08:36 | <JonathanNeal> | jgraham: we are bad at it because the platform or tools are bad, or that we just don't use them enough or effectively? |
| 08:37 | <MikeSmith> | zcorpan: that indentation doesn't look to me like it would cause a problem in python |
| 08:37 | <SimonSapin> | paul_irish: Do you mean we should change the criteria to something less restrictive? |
| 08:38 | <MikeSmith> | .win 20 |
| 08:38 | <MikeSmith> | oops |
| 08:38 | <paul_irish> | SimonSapin: yes |
| 08:38 | <zcorpan> | MikeSmith: the last "return" statement has a tab but the line above has 5 spaces |
| 08:38 | <SimonSapin> | paul_irish: I’m not sure how to do that and still keep the criteria useful :/ |
| 08:38 | <paul_irish> | SimonSapin: unless the WG is comfortable with Exit being ~100% harder than before? |
| 08:38 | <SimonSapin> | multiple independent implementations are still a Good Thing® |
| 08:38 | <zcorpan> | now i don't remember the rules exactly for python, but i'm not sure a tab is greater than 5 spaces |
| 08:39 | <MikeSmith> | zcorpan: ah I need to do ":set list" in vim (which I normally got set in my own .vimrc but am su'ed to root) |
| 08:39 | <jgraham> | 8 |
| 08:39 | <zcorpan> | jgraham: ok |
| 08:40 | <jgraham> | But the real rule is that you burn in hell for mixing tabs and spaces |
| 08:40 | <SimonSapin> | paul_irish: it’s a real issue, but I don’t have a good answer. Please bring it up on www-style |
| 08:41 | <zcorpan> | changed to 4 spaces |
| 08:41 | <tobie> | paul_irish: Here's what the W3C process doc says: http://www.w3.org/2005/10/Process-20051014/tr#cfr |
| 08:41 | <tobie> | "Shown that each feature of the technical report has been implemented. Preferably, the Working Group should be able to demonstrate two interoperable implementations of each feature." |
| 08:42 | <MikeSmith> | zcorpan: but the "while True:" above it is only indented 4 spaces, right? |
| 08:42 | <tobie> | That leaves a lot of leeway. |
| 08:42 | <jgraham> | tobie: Oh, some "preferably" in there too |
| 08:42 | <zcorpan> | MikeSmith: previously it was 5 spaces. now it's 4 and the "return" has 8 |
| 08:42 | <tobie> | yup. |
| 08:43 | <MikeSmith> | zcorpan: ah |
| 08:43 | <tobie> | MikeSmith: don't you just <3 languages where ws is significant? |
| 08:44 | <jgraham> | tobie: Yes, I do :) |
| 08:45 | <Philip`> | They're not quite as bad as languages where whitespace is insignificant and you can write |
| 08:45 | <Philip`> | if (x > 0); |
| 08:45 | <Philip`> | foo(); |
| 08:47 | <tobie> | Lets have a language flamewar in #whatwg. That sounds like a productive use of everyone's time. Ideally, we should ping people, so everyone can have a say. |
| 08:47 | <Philip`> | Clearly the best solution is to write all your programs in a single line |
| 08:47 | MikeSmith | dials up all his lisp buddies |
| 08:48 | <jgraham> | Philip`: SO, you been learning APL? |
| 08:48 | <JonathanNeal> | WEN WILL SASS BE IN SPEC? LINE RETURNS MATTER TOO |
| 08:48 | <MikeSmith> | Tastes great! vs Less filling! |
| 09:01 | <JonathanNeal> | Is it likely that more link types will be added to the spec, like rel="self"? |
| 09:06 | <MikeSmith> | JonathanNeal: some of them may graduate into the HTML spec, but rel="self" isn't going anywhere til somebody actually writes a formal definition of what it means in an HTML doc (as opposed to an atom doc), and probably even then not until there is evidence of wide use of it by applications that consume and actually do something with it |
| 09:12 | <zcorpan> | MikeSmith: should the ws handler be working now? |
| 09:12 | <annevk> | Oh fricking lovely |
| 09:13 | <annevk> | Because Blob does not check its type parameter everyone else will have to https://bugs.webkit.org/show_bug.cgi?id=111380 |
| 09:13 | <annevk> | Blob should check its type parameter imo |
| 09:13 | <marcosc_> | I really don't f'ing understand why the W3C refuses to enable CORS on TR |
| 09:15 | <marcosc> | MikeSmith, darobin, any chance you guys could talk to the w3c sys people to enable CORS |
| 09:15 | <darobin> | marcosc: enable it where? |
| 09:15 | <marcosc> | on /TR/ |
| 09:16 | <MikeSmith> | marcosc: have been talking already |
| 09:16 | <darobin> | h http://www.w3.org/TR/html5/Overview.html | grep Access |
| 09:16 | <darobin> | 12:Access-Control-Allow-Origin: * |
| 09:21 | <marcosc> | darobin: hmm... was not seeing that response header in the browser |
| 09:22 | <marcosc> | darobin: I can see it in curl |
| 09:22 | <darobin> | you mean you weren't seeing it in XHR? |
| 09:22 | <marcosc> | testing it now |
| 09:23 | <darobin> | I don't think XHR surfaces CORS headers, but anyway.... |
| 09:23 | <darobin> | are you telling me that you made a cross-request domain to /TR/ |
| 09:23 | <darobin> | that you got something back |
| 09:23 | <darobin> | and that you thought it didn't work because you couldn't see the header? |
| 09:23 | <darobin> | ... |
| 09:23 | <darobin> | s/cross-request domain/cross-domain request/ |
| 09:24 | <annevk> | s/cross-domain request/cross-origin request/ |
| 09:24 | <darobin> | yeah that |
| 09:24 | <marcosc> | no, I went to a random spec on TR and was checking if Access-Control-Allow-Origin was sent |
| 09:24 | darobin | heads for coffeee |
| 09:27 | <marcosc> | darobin: |
| 09:27 | <marcosc> | http://jsfiddle.net/B7AZu/ |
| 09:28 | <marcosc> | XMLHttpRequest cannot load http://www.w3.org/TR/REC-xml/. Origin http://fiddle.jshell.net is not allowed by Access-Control-Allow-Origin. |
| 09:28 | <hsivonen> | jpwhiting: you may be interested in looking at the changesets that added ITS validation |
| 09:29 | <darobin> | marcosc: I don't know man, it works on the CLI :) |
| 09:29 | <marcosc> | CORS is kinda pointless in the CLI :) |
| 09:30 | <marcosc> | Quick, W3C meme this: "Hey guys! We enabled CORS in the CLI. Now you can curl without problems!" |
| 09:32 | <darobin> | marcosc: your fiddle works perfectly fine for me... |
| 09:32 | <marcosc> | really? |
| 09:32 | <marcosc> | what browser? |
| 09:32 | <hsivonen> | annevk: FYI: https://www.w3.org/Bugs/Public/show_bug.cgi?id=21203 (CORS and EME) |
| 09:33 | <darobin> | wanna W3C "I don't always use jsfiddle, but when I do I can't figure out how to press the Run button"? |
| 09:33 | <darobin> | marcosc: Fx18 |
| 09:33 | <darobin> | anyway, I really need that coffee |
| 09:33 | <hsivonen> | darobin: you need a Firefox upgrade |
| 09:33 | <darobin> | I need it to make cross brain-blood barrier requests |
| 09:33 | <darobin> | hsivonen: I always wait a little bit |
| 09:35 | <darobin> | and besides, I don't really restart my browser every six weeks, that's a tad too often |
| 09:37 | <hsivonen> | Is the picture in http://w3cmemes.tumblr.com/post/43403131361/and-he-thought-that-special-protector-of-drm a common meme picture one is supposed to recognize? |
| 09:38 | <MikeSmith> | hsivonen: I think it is not a common one, no |
| 09:40 | <hsivonen> | MikeSmith: ok. do you happen to know the origin of the picture? |
| 09:41 | <annevk> | hsivonen: added myself to the cc list |
| 09:42 | <annevk> | marcosc: does TR/ send multiple CORS headers? |
| 09:42 | <marcosc> | I don't think so. Looks like this is just a bug in Chrome 25. Chrome 27 is fine, as is Safari. |
| 09:43 | <hsivonen> | annevk: I tend to think that allowing only one origin or * was a mistake. |
| 09:43 | <MikeSmith> | hsivonen: I promised my girlfriend the other day that I would quite lying. So because of that promise, you must permit me to be excused from answering your question directly. |
| 09:43 | <hsivonen> | MikeSmith: ok |
| 09:44 | <hsivonen> | annevk: that is, some big sites might not want to put a huge ACL in there, but small sites might still benefit from being able to put their full (small but larger than one item) ACL in there |
| 09:45 | <annevk> | hsivonen: allowing more can be done, but I suppose the fallback could have been better |
| 09:45 | <annevk> | hsivonen: maybe if we want to allow more we should introduce a shorter header too... |
| 09:46 | <MikeSmith> | hsivonen: I did find http://i.imgur.com/nMTEg.jpg from google image search |
| 09:46 | <annevk> | then Access-Control-Allow-Origin is for fallback |
| 09:50 | <hsivonen> | MikeSmith: hmm. so there is some meme pattern there |
| 09:52 | <hsivonen> | I wonder how much Stack Overflow is fueled by 386. http://stackoverflow.com/questions/5175840/is-html-a-context-free-language needs a better answer. |
| 09:54 | <MikeSmith> | hsivonen: I wouldn't make too much of those two instances as evidence of a pattern. My guess is that one of them just copied the meme connection from the other. :) |
| 09:54 | <MikeSmith> | hsivonen: oh boy "The answer is yes. SGML is a well defined Context Free Language, and HTML defined on top of it is also a CFL." |
| 09:54 | <MikeSmith> | definitely 386 |
| 09:55 | <hsivonen> | it's ridiculous that hg strip --no-backup 'roots(outgoing())' takes hours longer than just recloning the full repo |
| 09:55 | <MikeSmith> | hsivonen: the NO answer there is pretty good though |
| 09:57 | <hsivonen> | today I learned: Even Kingston sells ridiculously slow USB sticks |
| 09:57 | <hsivonen> | and here I thought Kingston was an OK brand for that |
| 10:03 | <jgraham> | hsivonen: What is that doing? Removing all unpushed code? |
| 10:03 | <zcorpan> | MikeSmith: a tcp trace suggests it's still the old handler |
| 10:04 | <MikeSmith> | oh |
| 10:04 | <MikeSmith> | zcorpan: yeah |
| 10:04 | <MikeSmith> | I remember now |
| 10:04 | <MikeSmith> | I need to manually copy them |
| 10:04 | <zcorpan> | aha |
| 10:05 | <MikeSmith> | sorry it just came back to me, from when we talked at TPAC. You remember? |
| 10:05 | <MikeSmith> | anyway, fixing it now |
| 10:05 | <zcorpan> | hmm. yeah now you mention it |
| 10:06 | <zcorpan> | just installed packet peeper, which seems to be pretty neat |
| 10:07 | <MikeSmith> | zcorpan: ok the updated file is in the right place now |
| 10:08 | <MikeSmith> | sorry for making you burn up time man :( |
| 10:08 | <zcorpan> | Pass WebSockets: Set-Cookie in response |
| 10:08 | <zcorpan> | yay |
| 10:08 | <MikeSmith> | cool |
| 10:08 | <zcorpan> | MikeSmith: it's ok, i was mostly reading email anyway |
| 10:09 | <zcorpan> | hey i just joined 3 WGs |
| 10:09 | <zcorpan> | webapps, css, webrtc |
| 10:09 | <MikeSmith> | nice |
| 10:10 | <zcorpan> | i wonder which specs are in greatest need of an editor |
| 10:10 | <MikeSmith> | zcorpan: I just use tcpdump. packet peeper is a gui I guess? |
| 10:10 | <zcorpan> | MikeSmith: yeah |
| 10:10 | <zcorpan> | MikeSmith: i didn't work out how to read the output of tcpdump |
| 10:10 | <zcorpan> | and since the websocket is on port 80 i didn't figure out a good filter |
| 10:11 | <hsivonen> | jgraham: it removes all changesets that are not in the current upstream after you've edited .hg/hgrc to give the repo a different upstream than it had originally |
| 10:11 | <zcorpan> | in packet peeper i could just scroll through the list and go "trace tcp" when i saw something involving "websocket" |
| 10:11 | <zcorpan> | which gave a nice request+response for the websocket connection |
| 10:12 | <MikeSmith> | ok |
| 10:12 | <zcorpan> | which is similar to the experience of wireshark, but wireshark's UI is horrible |
| 10:13 | <MikeSmith> | zcorpan: yeah wireshark is painful |
| 10:14 | <jgraham> | hsivonen: I have to ask why? I mean I can imagine wanting to reset the current branch to the new upstream head, but why remove commits? Or is this some hg-ism of not allowing commits that aren't reachable from a head? |
| 10:15 | <zcorpan> | the installation was similarly seamless with peeper but less so with wireshark |
| 10:28 | <hsivonen> | jgraham: it saves storage and is supposed to be faster than a full clone from the network |
| 10:28 | <hsivonen> | jgraham: except that it really is slower |
| 10:28 | <hsivonen> | jgraham: the storage saving is probably real, though |
| 10:28 | <hsivonen> | I succumbed to 386: http://stackoverflow.com/questions/5175840/is-html-a-context-free-language/15244772#15244772 |
| 10:46 | <hsivonen> | hmm. maybe the Kingston USB stick is OK but a couple of USB ports on my computer are just terrible... |
| 10:47 | <gsnedders> | hsivonen: Is HTML without parse-errors (ignoring validity for now) a CFG? |
| 10:48 | gsnedders | thinks not |
| 10:54 | <zcorpan> | hsivonen: " and those rules are complex enough that they can be fully defined using tree automata" s/can/can't/ ? |
| 11:04 | <hsivonen> | zcorpan: thanks. fixed |
| 11:05 | <hsivonen> | gsnedders: Pretty sure not due to feedback from the tree builder into the tokenizer, but I didn't try to actually develop a proof |
| 11:09 | <annevk> | Nice demonstration on stackoverflow.com how you're only as smart as the people that pay attention. |
| 11:17 | <hsivonen> | gsnedders: I edited the answer to ack interesting questions in the no parse errors but ignoring validity scenario |
| 11:21 | <SimonSapin> | zcorpan: do you really want to spec table layout? |
| 11:28 | <annevk> | well, someone has to |
| 11:28 | <annevk> | but defining the box model in a different way first might be a good idea |
| 11:35 | <zcorpan> | SimonSapin: maybe |
| 11:41 | <SimonSapin> | annevk: box model yay https://twitter.com/SimonSapin/status/263190763851116544/photo/1 |
| 11:41 | <annevk> | SimonSapin: uhuh |
| 11:42 | <SimonSapin> | more seriously, I’d help with a better definition |
| 11:42 | <SimonSapin> | it’d also need hooks for other CSS modules to influence box generation |
| 11:44 | <annevk> | CSS needs a little less of "this is how we organize specs" and therefore "this is how to write them" and more "this is how the code works" |
| 11:44 | <SimonSapin> | for table layout, hopefully we can push dbaron to help update http://dbaron.org/css/intrinsic/ and put it in css3-sizing |
| 11:53 | <annevk> | Hixie: I thought that you said Origin could not always be included in the request. But it seems per the Origin specification it's always included, just sometimes it has the value "null"... |
| 12:12 | <annevk> | cross-origin = no-and-fail, no-and-taint, yes-and-fail seems to be pretty much it |
| 12:13 | <annevk> | no-and-taint is a typical fetch today |
| 12:13 | <annevk> | no-and-fail is workers |
| 12:14 | <annevk> | yes-and-fail is XHR, stuff with crossorigin set |
| 12:17 | <SimonSapin> | anyway zcorpan, good to have you in the group |
| 12:17 | <zcorpan> | SimonSapin: thanks |
| 12:19 | <zcorpan> | annevk: what does "yes" and "no" refer to here? |
| 12:19 | <annevk> | zcorpan: whether you want to go there at all |
| 12:20 | <zcorpan> | ah |
| 12:21 | <zcorpan> | so <track> is yes-and-fail, <video> is yes-and-taint (i think)? |
| 12:21 | <annevk> | <video> is yes-and-taint? I thought that didn't exist |
| 12:22 | <zcorpan> | <video> has a crossorigin attribute, though we didn't implement it for <video> in presto (only for track) |
| 12:22 | <annevk> | <track> depends on crossorigin="" |
| 12:22 | <zcorpan> | yeah |
| 12:22 | <annevk> | if <track> has no crossorigin="" attribute it would be no-and-fail |
| 12:22 | <annevk> | otherwise yes-and-fail |
| 12:22 | <annevk> | afaict |
| 12:23 | <zcorpan> | ah yeah, indeed |
| 12:23 | <annevk> | whereas most such as <img> are no-and-taint and yes-and-fail |
| 12:23 | <annevk> | I think <video> would just be no-and-taint then |
| 12:23 | <zcorpan> | isn't <img> yes-and-taint without crossorigin set? |
| 12:23 | <annevk> | no |
| 12:24 | <annevk> | yes-and-taint does not exist |
| 12:24 | <annevk> | afaik |
| 12:24 | <annevk> | (although that was my original design) |
| 12:24 | <zcorpan> | but <img src="//cross-origin.com"> loads the resource. so i guess i don't understand your "yes" and "no" still |
| 12:26 | <annevk> | yeah, should name it differently |
| 12:26 | <annevk> | mode = same-origin, CORS-cross-origin, CORS-same-origin |
| 12:26 | <annevk> | well no that's wrong |
| 12:27 | <annevk> | same-origin, tainted cross-origin, CORS |
| 12:27 | <jgraham> | Hmm, it seems presto and WebKit both set an event property on window during execution of a event handler, at least one set via a content attribute (dunno how it works with nested events) |
| 12:27 | <jgraham> | Is that in the spec somewhere I'm not looking |
| 12:27 | <jgraham> | ? |
| 12:27 | <annevk> | it's an IE thing |
| 12:28 | <annevk> | which we've wanted to kill for a long time now |
| 12:29 | <jgraham> | Some variant of "we" that doesn't include WebKit, it seems |
| 12:30 | <annevk> | well yeah, other than arv_ there's not much communication from WebKit about these things :-( |
| 12:33 | <hsivonen> | Interesting that my Voddler/Skyfall tweet is not getting retweets. Too shocking to be believed? |
| 12:34 | <annevk> | hsivonen: I don't know about Voddler and a single incident is unlikely to convince anyone I think |
| 12:34 | <zcorpan> | MikeSmith: i've updated the handlers again. take it when you have time, i'll be travelling anyway |
| 12:35 | <hsivonen> | annevk: depends on whether one wants to be convinced of existence proof or about generalization |
| 12:35 | <hsivonen> | annevk: it's existence proof but doesn't generalize ATM |
| 12:36 | <annevk> | hsivonen: I guess it's kind of noteworthy that you can stream Skyfall to a browser, hadn't really considered it from that perspective |
| 12:36 | <annevk> | hsivonen: on the other hand, I don't really follow what's going on in that space |
| 12:36 | <hsivonen> | annevk: to a browser without DRM on Linux |
| 12:37 | <hsivonen> | that should be off-the-charts notable compared to the other stuff I tweet that gets retweeted all the time |
| 12:47 | <annevk> | Dunno how it works. Maybe "HTML5 DRM-free version of Skyfall available for rent. Why do we need EME again?" would work better? |
| 13:12 | <gsnedders> | hsivonen: I'm not sure that that cannot be reduced to a single stack in the valid case. |
| 13:13 | <gsnedders> | hsivonen: I tried to construct a semi-formal proof before, but got bored and gave up. |
| 13:56 | <annevk> | Okay, so the Origin header situation seems really messed up... |
| 13:57 | <annevk> | For "potentially CORS-enabled fetch" it's basically only included if a) mode is "Anonymous" or "Use Credentials" and b) the request is cross-origin |
| 13:57 | <annevk> | For "fetch" it's only included if "from" is present in the invocation of the algorithm |
| 13:57 | <annevk> | The intersection of "potentially CORS-enabled fetch" and "fetch" is "No CORS" taint |
| 13:58 | <annevk> | I think to properly support the "potentially CORS-enabled fetch" I'd need to introduce some kind of suppress Origin header flag |
| 14:00 | <annevk> | which when set "suppresses" the header for same-origin requests and tainted cross-origin requests while still allowing the response to the request to be marked as CORS-same-origin |
| 14:01 | <annevk> | :-( |
| 14:06 | <SimonSapin> | getting 500 errors on https://dvcs.w3.org/hg/csswg/ … who should I ping? |
| 14:08 | <jgraham> | MikeSmith or darobin_ |
| 14:08 | <jgraham> | Who might proxy the request |
| 14:08 | <darobin> | oh for fuck's sake |
| 14:08 | <darobin> | that thing's down all the time |
| 14:08 | <jgraham> | Or just cuss |
| 14:08 | <darobin> | it beats me why hg was even brought into this world |
| 14:09 | <darobin> | SimonSapin: should work now |
| 14:11 | <darobin> | for reference: https://twitter.com/robinberjon/status/309305376329830400 |
| 14:13 | <SimonSapin> | darobin: great! Now testing https://github.com/msysgit/msysgit/wiki/Guide-to-git-remote-hg … |
| 14:15 | <darobin> | SimonSapin: good luck |
| 14:15 | <darobin> | at some level I think that persuading the CSS WG to switch to git would actually prove *easier*, but I know that's setting the bar pretty high |
| 14:16 | <SimonSapin> | easier than using something like git-remote-hg? |
| 14:16 | <jgraham> | Easier than what? |
| 14:31 | <MikeSmith> | darobin: did you restart the server yet or should I? |
| 14:32 | <darobin> | MikeSmith: I did it |
| 14:32 | <darobin> | SimonSapin: well, tell me how it goes |
| 14:32 | <darobin> | but I suspect that there are enough incompatibilities between git and hg that you might end up with weird issues |
| 14:33 | <SimonSapin> | darobin: git and hg are actually very close, afaik hg-git works well enough |
| 14:33 | <SimonSapin> | https://github.com/dubiousjim/yagh documents various ways of how to do the reverse |
| 14:34 | <darobin> | I couldn't get hg-git to work reliably actually |
| 14:34 | <jgraham> | SimonSapin: Whenever I have tried using git with a hg remote "well enough" was a ephemism for "with extreme pain" |
| 14:34 | <SimonSapin> | git-hg-again worked fine, but was slow. I suspect it to rewrite the whole commit history … sometimes |
| 14:34 | <darobin> | and neither could Aryeh |
| 14:34 | <jgraham> | I mean it did sort of work |
| 14:35 | <darobin> | for the very basics, most of the time |
| 14:35 | <jgraham> | But it was hard to recommend as a solution |
| 14:35 | <darobin> | but anything slightly exotic... |
| 14:35 | <SimonSapin> | with msysgit’s git-remote-hg, this Just Worked®: git clone hg::http://dvcs.w3.org/hg/csswg/ |
| 14:35 | <SimonSapin> | git gc --aggressive to make the repo smaller |
| 14:36 | <SimonSapin> | we don’t have a fancy branch setup in CSSWG |
| 14:37 | <SimonSapin> | this implementation looks more solid than various patterns for "hg-git in reverse" |
| 14:37 | <jgraham> | Well, if it still works once you are actually using the local repo and want to pull in updates that will be more impressive |
| 14:38 | <jgraham> | Since just cloning is quite straightforward, but dealing with merges and so on is less so |
| 14:39 | <SimonSapin> | well, merges are done locally by git |
| 14:39 | <jgraham> | (I assume the reason that git-svn works OKish is because no one over does anything remotely complex with svn) |
| 15:00 | <SimonSapin> | jgraham: https://dvcs.w3.org/hg/csswg/rev/ad000eb1bb5d was just pushed from git |
| 15:21 | <zewt> | very first impression of ie10: i load it, click the address bar, start typing a url, and midway through typing it erases the url i'm typing and goes to some "welcome to IE10!" thing |
| 15:24 | <darobin> | I've had similar issues with it insisting that I really meant my URLs to be Bing searches |
| 15:25 | <jgraham> | IE has terrible UI? Say it isn't so! |
| 15:35 | <hsivonen> | I'm eager to see charset menu telemetry hit the release channel. Can't really conclude anything about Nightly, since the usage may be devs debugging rather than real normal users seeing broken pages. |
| 15:51 | <annevk> | So lets see. Send Origin when passed for mode tainted cross-origin. Otherwise only send for requests with mode CORS |
| 15:56 | <zewt> | all w3 lists are subscriber-only-posting, right? |
| 15:56 | <Ms2ger> | I don't think so |
| 15:56 | <Ms2ger> | I'm pretty sure anyone can post to www-archive |
| 15:56 | <zewt> | (not sure what I should trim from this huge crosspost on the IDB thread) |
| 15:57 | <zewt> | probably going to trim to just webapps (has -tag and a zillion random people on i) |
| 16:18 | <zewt> | ugh, somebody at firefox needs to be shot |
| 16:18 | <zewt> | as of firefox 17 and current webkit, both browsers agreed: headers in XHR are encoded/decoded as UTF-8 and exposed as normal strings |
| 16:19 | <zewt> | only IE had the insane, broken and completely unusable behavior of exposing the raw data |
| 16:19 | <zewt> | ... but firefox 19 is now doing what IE does |
| 16:20 | <jgraham> | zewt: What does the spec say? |
| 16:20 | <zewt> | it's depressing when we have a majority agreement on good, sane behavior, then we regress to insanity |
| 16:20 | <jgraham> | And is there a test case? |
| 16:20 | <zewt> | jgraham: the spec says IE's behavior (which is why I was looking into this, I wanted to tell anne to get rid of that crap) |
| 16:21 | <zewt> | http://zewt.org/~glenn/blar.html, check the console |
| 16:22 | <zewt> | logs filename=漢字 in FF17 and current Chrome, mojibake in FF19 and IE |
| 16:25 | <zewt> | xhr.setRequestHeader also takes a regular UTF-16 string and sends UTF-8 in the header in FF17/WebKit (it doesn't throw); haven't tested to see if FF19 changed anything there |
| 16:32 | <SimonSapin> | pango-view --font 'Fontin 30' <(echo cffi) |
| 16:32 | <SimonSapin> | sorry, wrong tab |
| 16:44 | <asmodai> | Anyone else notice embedding issues with Chrome 26 and the latest Flash for, say, YouTube videos on pages? It shows the Flash embed, but nothing's clickable. |
| 16:51 | <dglazkov> | good morning, Whatwg! |
| 16:52 | <wilhelm> | "Therefore HTML5 is now permitted in W3C Recommendations." |
| 16:55 | <darobin> | wilhelm: that's actually not true |
| 16:55 | <darobin> | (or soon not true, I forget) |
| 16:57 | <Ms2ger> | Or relevant, for that matter |
| 16:58 | <darobin> | oh, wilhelm said "now" :) |
| 16:58 | <darobin> | I had read not |
| 16:58 | <darobin> | lots of information riding on that letter |
| 17:07 | <bholley> | Hixie: yt? |
| 17:27 | <asmodai> | Mmm, interesting. Enable the builtin pepperflash of Chrome, the embed works. Use the official flash from the Adobe site and the embed is unclickable. |
| 17:45 | <Hixie> | bholley: briefly, if it's quick, or i'll be online again in a bit |
| 17:46 | <bholley> | Hixie: naw, it's ok. I commented in bug 19662 |
| 17:46 | <Hixie> | k |
| 17:53 | <JonathanNeal> | Does rel="index" point to a reference list for the site (like the index in a book), or does it point to the main/central document of the site (like how the index file commonly works on the web). |
| 17:53 | <Hixie> | what does the definition say? |
| 17:54 | <JonathanNeal> | "Refers to a document providing an index for the current document." |
| 17:57 | <JonathanNeal> | And I'm unsure if "index" means "top level resource" or not. |
| 17:57 | <JonathanNeal> | since "index" uses "index" to define itself. |
| 17:58 | <Ms2ger> | Index in a book, I guess |
| 17:58 | <Ms2ger> | You refer to index.html? |
| 17:58 | <JonathanNeal> | When I think of an index in a book, I think of the table of contents. |
| 17:58 | <JonathanNeal> | When I think of an index on the web, I think of a top level resource. |
| 17:59 | <JonathanNeal> | like a "home page" |
| 18:00 | <JonathanNeal> | Perhaps an index on the web can be thought of similarly as a table of contents, just also with independent content and/or other aggregated content. |
| 18:00 | <Ms2ger> | When I think of an index, it's a list of terms at the end of a book |
| 18:01 | <JonathanNeal> | Ms2ger: right you are, I was a little turned around. |
| 18:01 | <Ms2ger> | I think that's the meaning that's used in the spec too :) |
| 18:02 | <JonathanNeal> | Originally, I thought that too, but the spec doesn't clarify, and other resources take the web definition, for instance, https://developer.mozilla.org/en-US/docs/HTML/Link_types |
| 18:04 | <JonathanNeal> | Also, I don't see many traditional indexes on the web, except in the form of search results. |
| 18:05 | <JonathanNeal> | I'm at a draw, leaning to the web definition, but I think you prove that there's considerable reason to doubt that it means top level resource, unless there's another part of spec that clarifies. |
| 18:16 | <jgraham> | tobie: "desired tests" scares me |
| 18:16 | <jgraham> | Since there is no mechanism to prove that tests are covering different parts of the specification it's a rather meaningless number |
| 18:17 | <jgraham> | Certainly it would be really easy to stuff the repo. with many tests for a feature without testing it deeply |
| 18:17 | <jgraham> | This is not in any way a theoretical concern |
| 18:18 | <tobie> | That's why we have a review process. |
| 18:19 | <tobie> | And we'll have ways to manually override this "desired test" data |
| 18:19 | <jgraham> | Well the review process is only likely to catch "this test is wrong" (and even then) |
| 18:20 | <jgraham> | It's hard to say to someone "no we don't need tests for 'outline' with 20 different colours" |
| 18:20 | <tobie> | how so? |
| 18:20 | <tobie> | It's the whole point of this exercise. |
| 18:20 | <Ms2ger> | Because Microsoft will push those hard... |
| 18:20 | <jgraham> | Because it will cause conflict |
| 18:21 | <tobie> | Really? |
| 18:21 | <jgraham> | yes |
| 18:21 | <tobie> | The whole point of this exercise is to identify areas where writing tests is particularly inmpactful |
| 18:21 | <jgraham> | Because if you complain about it the complaint amounts to "I think you are acting in bad faith" |
| 18:21 | <jgraham> | Sure, I agree that it is highly useful to identify undertested areas |
| 18:22 | <tobie> | well there you go. |
| 18:22 | <jgraham> | I don't think the inverse is true or possible |
| 18:22 | <jgraham> | "it" - a general method |
| 18:22 | <jgraham> | Not this specific presentation |
| 18:22 | <tobie> | We can have reviewers override coverage for a section. |
| 18:22 | <tobie> | Or test facilitators do the same. |
| 18:23 | <tobie> | So you add a hundred useless test and the reviewer looks at it and says: oh well, that's does;t move the needle one bit. |
| 18:23 | <tobie> | Actual coverage is still 50%/ |
| 18:23 | <jgraham> | (I guess the inverse might be possible if we used code coverage data from the implementations contributing to the conformance report... Might be hard to identify the "relevant" code for each part though) |
| 18:23 | <tobie> | That's an interesting strategy. |
| 18:25 | <tobie> | That said, running these scripts on the HTML5 spec has already uncovered some very valuable info: |
| 18:25 | <jgraham> | I guess for Gecko and WebKit implementing new features it might actually work |
| 18:25 | <tobie> | we're missing test reviewers more than tests writers. |
| 18:25 | <jgraham> | But not for Trident, or features that already exist |
| 18:26 | <tobie> | jgraham: are you interested in looking more deeply at this and trying to create similar report using the strategy you just described? |
| 18:27 | <jgraham> | tobie: Interested, yes. But not prepared to make promises :) |
| 18:28 | <Ms2ger> | jgraham, actually, I'd love to see code coverage, can I task you to do that? :) |
| 18:28 | <jgraham> | Ms2ger: How much do you pay? :p |
| 18:28 | <Ms2ger> | jgraham, on Opera's time ;) |
| 19:00 | <zewt> | wow, gross |
| 19:01 | <zewt> | in mobile safari, clicking the document while the onscreen keyboard is open appears to close the keyboard ... if a click event handler is registered |
| 19:09 | <astearns> | jgraham: tobie: Ms2ger: If there's a spec section with a questionable set of tests, one possible solution is to go more granular |
| 19:09 | <Ms2ger> | Hmm? |
| 19:09 | <astearns> | map those tests to subsections, paragraphs or spans (the way WOFF does) to show what's still not covered |
| 19:09 | <Ms2ger> | The solution is better tests |
| 19:10 | <astearns> | agreed. I'm just looking for a way to settle an argument if one crops up |
| 19:11 | <tobie> | what about we wait for the argument to crop up and then act accordingly. |
| 19:11 | <tobie> | the data we have is far from being a panacea. |
| 19:12 | <tobie> | Yet we can improve it over time, either through better heuristics or by overriding parts of it manually. |
| 19:12 | Ms2ger | grinds his axe |
| 19:13 | <tobie> | This data is better than just pulling numbers out of thin air like we've been doing so far. |
| 19:14 | <astearns> | yep yep |
| 19:14 | <tobie> | During the same meeting, I heard HTML5 needed 1 million tests or 10'000. |
| 19:14 | <tobie> | That's two orders of magnitude. And that's ridiculous. |
| 19:14 | <Ms2ger> | Both are too low |
| 19:15 | <jgraham> | For comparison we had like 1/3 of a million tests for Presto |
| 19:15 | <jgraham> | And that was not enough |
| 19:15 | <jgraham> | (that's the whole platform of course) |
| 19:15 | <tobie> | So one thing we should start doing is count in number of assertions. |
| 19:16 | <tobie> | a test means little, really. |
| 19:16 | <tobie> | unless we have a shared definition of what a test is. |
| 19:16 | <tobie> | ? |
| 19:16 | <tobie> | jgraham: how much of those can we bring to the W3c repository |
| 19:18 | <jgraham> | tobie: Lots of that was stuff that we imported from outside |
| 19:18 | <jgraham> | But we are, and have been, working on releasing tests we wrote ourselves |
| 19:19 | <jgraham> | There are lots of tests that will likely not get released though just because it would be so much effort |
| 19:19 | <tobie> | yeah. git push w3c is painful to type. |
| 19:19 | <tobie> | :P |
| 19:20 | <jgraham> | You are focussing on the wrong part :) |
| 19:20 | <tobie> | :) |
| 19:21 | <jgraham> | e.g. for the websockets tests which were written in a slightly pre-testharness.js framework and were already organised well, it took me maybe 2 days of effort to convert the tests and zcorpan is still fixing some edge cases |
| 19:21 | <jgraham> | In fact, including the time at TPAC, I guess 2 is an underestimate |
| 19:22 | <Ms2ger> | I wish we'd use testharness.js more :/ |
| 19:25 | <jgraham> | (kind of hard to estimate since it was evenings + etc. so I wasn't really counting) |
| 19:26 | <Ms2ger> | I wonder if Bj�rns research on quotes in email could be integrated in thunderbird... |
| 19:38 | <Hixie> | JonathanNeal: [define:index] on google says "An alphabetical list of names, subjects, etc., with references to the places where they occur, typically found at the end of a book. |
| 19:38 | <Hixie> | " |
| 19:41 | <JonathanNeal> | Hixie: exactly, and, as of this moment, I have adopted that definition, since microformats also specifies as "home" type. |
| 19:41 | <JonathanNeal> | http://microformats.org/wiki/existing-rel-values#HTML5_link_type_extensions |
| 19:43 | <JonathanNeal> | *specifies a "home" type. |
| 19:44 | <JonathanNeal> | I also edited the document to prevent the keywords from being used to define themselves. |
| 19:45 | <JonathanNeal> | I think TabAtkins has a name for when you do that, something like a game of intellectual taboo. |
| 19:46 | <Hixie> | the rel=index and the english word "index" have nothing to do with each other |
| 19:46 | <Hixie> | it's not like they're the same word |
| 19:47 | <Hixie> | one is an opaque string, the other is a human language word defined in dictionaries and by usage |
| 19:47 | <Hixie> | there's no difference between "the 'index' keyword declares an index" and "the 'zoopie' keyword declares an index" |
| 19:47 | <Hixie> | if the second of those is well-defined, then so is the first |
| 19:48 | <Ms2ger> | Well-defined as long as you have a shared definition of the English word "index" :) |
| 19:49 | <Hixie> | that's what i mean |
| 19:49 | <Hixie> | but that's got nothing to do with whether you're defining the _keyword_ "index" or "zoopie". |
| 19:49 | <Hixie> | (JonathanNeal: ^) |
| 19:51 | <JonathanNeal> | I follow, but there were two conflicting meanings for "index". Again, MDN had it the other way https://developer.mozilla.org/en-US/docs/HTML/Link_types |
| 19:52 | <Hixie> | two meanings for the keyword or for the word? |
| 19:52 | <JonathanNeal> | for the keyword. |
| 19:58 | <JonathanNeal> | That's why I linked to both the microformats and mdn wikis, Hixie. I follow what you mean about keyword vs word, and I am talking about the keyword. |
| 20:00 | <Hixie> | well the mdn wiki isn't normative, so it's pretty easy to decide which is the right definition :-) |
| 20:01 | <Hixie> | if anyone has an opinion on what "the initial about:blank" is, please let https://www.w3.org/Bugs/Public/show_bug.cgi?id=17093 know |
| 22:22 | <ronaldmansveld> | tantek, Hixie either of you here? |
| 22:24 | <tantek> | sometimes ;) |
| 22:25 | <ronaldmansveld> | haha |
| 22:25 | <ronaldmansveld> | I have a quick question about the :after pseudo |
| 22:25 | <tantek> | which we now prefer ::after for - since it's a pseudo-element (rather than pseudo-class) |
| 22:25 | <ronaldmansveld> | (and since you're both editors, I guess you know the answer) |
| 22:26 | <ronaldmansveld> | You're correct, thank you |
| 22:26 | tantek | wonders if this is a references to CSS 2.1 or Selectors ;) |
| 22:26 | <ronaldmansveld> | 2.1 |
| 22:26 | <ronaldmansveld> | if I do a::after {content: "something";}, I guess the "something" should be clickable as well right? Since it's part of the anchor |
| 22:27 | <ronaldmansveld> | (which is implemented that way in IE10, FF and Opera, but not in Chrome) |
| 22:27 | JonathanNeal | looks among his protest paraphernalia and grabs the ::outside advocacy sign |
| 22:27 | <tantek> | ronaldmansveld - yes that sounds correct |
| 22:28 | <ronaldmansveld> | OK, that makes case for a bugreport :) |
| 22:28 | <JonathanNeal> | ronaldmansveld: woohoo |
| 22:28 | <tantek> | from a "tree" perspective, ::before and ::after are both *inside* their element, inherit its inheritable CSS properties etc. |
| 22:29 | <tantek> | yes that sounds like a webkit bug |
| 22:29 | <ronaldmansveld> | I thought so |
| 22:31 | <JonathanNeal> | Then, with your clout from this bug, report the missing ::outside |
| 22:31 | <JonathanNeal> | Get it landed in canary :) |
| 22:37 | <ronaldmansveld> | well, it isn't quite behaving as ::outside either |
| 22:37 | <ronaldmansveld> | it does get the styling of the anchor, it just doesn't register the click |
| 22:37 | <ronaldmansveld> | but I'll make mention of ::outside in the report :P |
| 22:39 | <JonathanNeal> | Oh, I was just being fun. I wouldn't want to discredit your report with something as trivial as ::outside. |
| 22:39 | <JonathanNeal> | Your bug report is totally valid and the expectations are clear. |
| 22:39 | <JonathanNeal> | My thing was just … me wishing. |
| 22:40 | <JonathanNeal> | So, by all means, advocate my wish, but don't let it hold you up. |
| 22:41 | <ronaldmansveld> | Ah, no need for a report |
| 22:41 | <ronaldmansveld> | had been fixed in canary |
| 22:41 | <ronaldmansveld> | *has |
| 22:41 | <JonathanNeal> | nice |
| 22:42 | <ronaldmansveld> | Now I have to wait even longer for my first bug-report :( |
| 22:45 | <bholley> | Hixie: yt? |
| 22:54 | <Hixie> | bholley: yo |
| 22:54 | <bholley> | Hixie: just commented |
| 22:54 | <bholley> | Hixie: that make sense? |
| 22:54 | <Hixie> | bug #? |
| 22:54 | <bholley> | Hixie: bug 19662 |
| 22:55 | <bholley> | Hixie: if you prefer to do this more async, that's totally fine |
| 22:55 | <bholley> | Hixie: I'm not sure how many pots you have boiling |
| 22:55 | bholley | is just in the habit of pinging people to sort out confusion |
| 22:56 | <Hixie> | too many to count but sync is fine too :-) |
| 22:57 | <Hixie> | bholley: i don't understand what you mean by "implementing the navigation restrictions as restrictions at property-access time" |
| 22:57 | <Hixie> | bholley: that doesn't seem to match the spec at all |
| 22:57 | <bholley> | Hixie: well, not as it stands now, no |
| 22:57 | <bholley> | Hixie: but I'm wondering if there's a qualitative difference |
| 22:58 | <bholley> | Hixie: other than the testcase in comment 6 |
| 22:58 | <Hixie> | well that's the one i care about here :-) |
| 22:58 | <Hixie> | why would you do it with this complicated property access thing rather than the simple model in the spec? :-) |
| 22:59 | <bholley> | Hixie: because it's faster |
| 22:59 | <bholley> | Hixie: and I don't think it's really any more complicated |
| 23:00 | <bholley> | Hixie: but I may not understand all the intricacies |
| 23:01 | <bholley> | Hixie: is there a reason the spec does things the way it does, as opposed to just defining sandbox as a restriction on the cross-origin accessible properties? |
| 23:01 | <Hixie> | i don't understand how a few pointer tests at navigation time can possibly be slow enough to warrant an entirely different security model |
| 23:01 | <Hixie> | well, to start with, these aren't cross-origin |
| 23:02 | <bholley> | Hixie: they're not? I thought the whole point was that they get a "unique origin"? |
| 23:02 | <bholley> | Hixie: if they're same-origin as the stuff they're navigating then it seems pretty pointless to try to prevent them from doing that |
| 23:03 | <bholley> | Hixie: as for the pointer tests - the expensive part is computing whether the navigatee is descendant from the frame doing the navigation or not |
| 23:03 | <bholley> | Hixie: it's not really that bad, and we only need to do it for sandboxed stuff. But this seemed like any easy optimization |
| 23:06 | <bholley> | Hixie: to be clear, I'm totally willing to converge around the spec here, especially if there are substantive differences and the property-access thing is harder for others to implement. But I'd like to work through this disconnect here first :-) |
| 23:15 | <Hixie> | bholley: sandbox="allow-same-origin" means the origin is left as is |
| 23:16 | <Hixie> | bholley: you can still sandbox same-origin stuff, just turn off scripting, e.g. |
| 23:16 | <bholley> | Hixie: sure. But allow-same-origin effectively means that the navigation restrictions go away, right? |
| 23:16 | <Hixie> | no |
| 23:16 | <Hixie> | doesn't affect them |
| 23:17 | <Hixie> | you can have a same-origin blog comment, and still prevent it from navigating the other blog comment iframes using target="" |
| 23:17 | <Hixie> | you need it same-origin so you can do seamless="" |
| 23:17 | <Hixie> | but you need it sandboxed to kill scripting |
| 23:17 | <Hixie> | and so on |
| 23:17 | <bholley> | Hixie: ok. Are they useless? |
| 23:17 | <bholley> | Hixie: (the navigation restrictions in the same-origin case) |
| 23:17 | <Hixie> | no |
| 23:17 | <Hixie> | especially not if JS is disabled |
| 23:18 | <Hixie> | (obviously same-origin and js-enabled is only useful in some pretty limited "good practice" cases, not for real security) |
| 23:18 | <bholley> | Hixie: if js is disabled, what is the difference between allow-same-origin and lack thereof? How does the same-origin policy affect things other than JS? |
| 23:20 | <Hixie> | bholley: if they're not same-origin, you can't style the page using seamless="" |
| 23:20 | <Hixie> | bholley: e.g. |
| 23:21 | <bholley> | Hixie: but the navigation restrictions are irrelevant if JS is disabled, right? |
| 23:21 | <Hixie> | no, i just gave some examples where they are still relevant |
| 23:21 | <Hixie> | <Hixie> you can have a same-origin blog comment, and still prevent it from navigating the other blog comment iframes using target="" |
| 23:21 | <bholley> | Hixie: oh, I missed the target="" part |
| 23:22 | <Hixie> | the way sandbox="" is specified right now is the result of over a year of careful design and deliberation, if we want to change it we need to do it really carefully and bring in the other implementors and make sure everyone's on board again with the new model |
| 23:22 | <Hixie> | i'm really skeptical that this is a good use of our time :-) |
| 23:22 | <Hixie> | security feature design is _hard_ |
| 23:23 | <bholley> | Hixie: sure. The target="" thing is important |
| 23:23 | <bholley> | Hixie: and it means that this can't be done the way imelven was doing it |
| 23:24 | <bholley> | Hixie: this is more or less what I was asking in the bug |
| 23:26 | <bholley> | Hixie: but in general, it's worth remembering that the spec describes security in a way that matches WebKit's implementation much more than Gecko's. So it's not always simple for us to "just" implement it the way it's written in the spec |
| 23:27 | bholley | goes to comment on the bug |
| 23:29 | <Hixie> | bholley: it matches the security model that firefox used to have, that opera used to have, that IE has, and that webkit has |
| 23:29 | <Hixie> | bholley: not my fault y'all moved away from that model for what are imho non-web reasons :-P |
| 23:30 | <bholley> | Hixie: are you sure it's the model that opera used to have? They did revocation for document.domain, which suggest that they do something more similar to Gecko in terms of cross-origin security checks |
| 23:34 | <bholley> | Hixie: and it seems kind of unproductive to call Gecko's security constraints "non-web" when "web" is really just defined by the behavior that browsers happen to have been implementing :P |
| 23:37 | <bholley> | (and are capable of implementing) |
| 23:47 | <yroc> | Hixie, you there? |
| 23:47 | <TabAtkins> | yroc: He just left where he was, so he's presumably currently in transit to his next location. |
| 23:48 | <yroc> | TabAtkins: I see. Can I ask you a question re: the article element? |
| 23:48 | <TabAtkins> | You can ask me a question about anything, though whether I can answer it is unclear. |
| 23:49 | <yroc> | OK, I'll chance it! |
| 23:49 | <yroc> | Hixie said yesterday that he can't envision a nested <article> being anything other than a comment. |
| 23:50 | <yroc> | This is part of the reason why no need for a <comment> element. |
| 23:50 | <yroc> | I suggested, how about a document (single file) that contains a book... |
| 23:51 | <yroc> | ... where individual chapters might be independently distributable. |
| 23:51 | <yroc> | Hixie said why would you wrap the book in an <article> element, because it's the whole page. |
| 23:52 | <TabAtkins> | I disagree with Hixie that nested articles must be comments. Your example seems reasonable (for example, maybe they're short stories!). |
| 23:52 | <TabAtkins> | Unless, of course, it's *not* the whole page. |
| 23:52 | <yroc> | Yes, that's a good example. |
| 23:54 | <yroc> | If the whole page is the book, Hixie said you shouldn't wrap it in an article? Why not I wonder (he scooted for lunch before I had a chance to ask him) |
| 23:54 | <yroc> | Does he mean the book should just be the <body>? |
| 23:57 | <astearns> | I'm currently reading a book that has chapters, and each chapter consists of several short stories. Seems like at least one level of <article> nesting there |
| 23:58 | <yroc> | astearns: That sounds like it could be *two* nesting levels. |
| 23:59 | <yroc> | astearns: Would you say the book as a whole is an <article>? |
