| 00:00 | <takkaria> | oh, the first one seems to do something now |
| 00:18 | <Philip`> | takkaria: You have to wait for it to load a zillion script files |
| 00:18 | <Philip`> | (Presumably it'd be possible to pack them all into a single .js file if you cared about performance) |
| 00:19 | Philip` | hopes he is actually testing it correctly, and didn't forget to change some of the prefixes or something |
| 00:40 | <LeifHS> | annevk5: Yes, <embed> is like <source> - kind of. However, <source> is part of the <video> element - it is an extension of <video>. What I meant was that WebKit is treating <embed> as if is part of <object> (the same way that <source> is part of <video>). |
| 00:40 | karlcow | always wonders why talented people becomes suddenly blind when they are passionate about their tech |
| 00:50 | <LeifHS> | zcorpan: Is it possible for you to create an arch typical exsample of what you mean? |
| 08:31 | Hixie | now has a basical proof of concept of his new <datagrid> API design |
| 08:32 | <Hixie> | now i just have to check how plausible it is |
| 08:32 | <Hixie> | from the authoring side... |
| 08:37 | <Hixie> | this is one of the first times that i've designed something for which i think a structure a bit like a B-tree would actually be a pretty good fit |
| 08:45 | <MikeSmith> | wow |
| 08:45 | <MikeSmith> | B-tree |
| 08:45 | <MikeSmith> | that's a blast from the past |
| 08:46 | <MikeSmith> | Hixie: seems like datagrid is ultimately going to sink or swim based on how much implementor commitment there is |
| 08:47 | <MikeSmith> | and given that there's not be much implementor commitment forthcoming so far, I wonder where that's going to leave it |
| 08:48 | <MikeSmith> | if we really want to get to LC by this Fall |
| 08:48 | <Hixie> | my impression is that browser vendors seem in agreement that it would be a useful feature, but that it isn't a priority |
| 08:51 | <MikeSmith> | Hixie: I think it's something more than just a useful feature |
| 08:52 | <Hixie> | oh? |
| 08:52 | <MikeSmith> | to me, it's something that would appeal quite a bit to Web developers |
| 08:52 | <Hixie> | that's what i mean by "useful feature" :-) |
| 08:53 | <Hixie> | note that by LC we don't have to have commitements to implement, only agreement that the features should be in the language |
| 08:53 | <MikeSmith> | Hixie: it's the "isn't a priority" to browser developers that's the big stumbling block |
| 08:53 | <Hixie> | in fact it's only when entering CR that we have to list features that might be at risk |
| 08:53 | <Hixie> | and even then we have until REC to see them implemented |
| 08:54 | <Hixie> | so lack of implementation commitements is not a big deal so long as implementors don't disagree that it would be good to implement eventually |
| 08:54 | <Hixie> | commitments |
| 08:54 | <MikeSmith> | I guess |
| 08:55 | <MikeSmith> | I would personally rather not see us take a particular feature into CR without a clear commitment from multiple browser vendors to implement it |
| 08:55 | <Hixie> | sure |
| 08:56 | <MikeSmith> | maybe we need to light a fire under some asses as far as datagrid |
| 08:56 | <hsivonen> | myvidoop.com no longer shows an EV cert. Am I being MITMed? |
| 08:56 | <MikeSmith> | e.g., threat "this is going to be removed from HTML5 unless we get a clear indication of vendor support" |
| 08:56 | <MikeSmith> | hsivonen: MITMed |
| 08:57 | <MikeSmith> | ? |
| 08:57 | <hsivonen> | MikeSmith: Man In The Middle |
| 08:57 | Hixie | watches the EV cert security model fall apart |
| 08:58 | <zcorpan> | hsivonen: do what normal people do and discard any dialogs and don't notice the lack of dialogs |
| 08:58 | <MikeSmith> | EV cert, despite whatever faults it has, is approximately one gazillion times better than non-EV certs |
| 08:59 | <Hixie> | for cert vendors, sure |
| 08:59 | <hsivonen> | MikeSmith: how so. It's causing me distress now. Probably for no good reason. |
| 08:59 | <hsivonen> | If I'm being MITMed, how do I email vidoop, when I should assume that by now the MX record in my DNS cache is poisoned as well? |
| 09:00 | <Hixie> | certs in general can answer the question "is this who i think it is", which is never a question users ask. They ask (implicitly) the question "am I being attacked", which is not possible to answer using any kind of SSL cert that I know of, EV or otherwise. |
| 09:01 | <Hixie> | (and EV certs can answer the first question with $450 more confidence than non-EV certs, that's about it.) |
| 09:01 | <MikeSmith> | Hixie: EV certs have documented rules for identity |
| 09:01 | <Hixie> | right. "$450 more confidence". |
| 09:01 | <MikeSmith> | when you pay (whatever dollar amount) for a non-EV cert, you are a sucker |
| 09:02 | <MikeSmith> | because you are paying for nothing |
| 09:02 | <Hixie> | you are paying for someone to sign it |
| 09:02 | <MikeSmith> | CAs are bound to do zero vetting otherwise |
| 09:02 | <Philip`> | You're paying for something that prevents passive network attackers from reading all the data send to/from your web site, which seems quite useful |
| 09:03 | <Philip`> | *sent |
| 09:05 | <Hixie> | Philip`: you can get that with a self-signed cert |
| 09:05 | <Philip`> | Hixie: You can't since it won't work in Firefox 3 |
| 09:05 | <MikeSmith> | the EV-cert work is similar to a lot of things in that it's really easy to make jackass comments about it from the outside after the fact, throwing rocks, without any understanding of the difficulties involved in ever having tried to do something like it all |
| 09:05 | <Hixie> | Philip`: that's a bug (it should encrypt but not show any ui about it) |
| 09:05 | <Hixie> | imho, anyway |
| 09:06 | <Hixie> | MikeSmith: as someone who is involved in the security work of three browsers, i feel pretty qualified in the subject. I'm fully aware that there has not been anything better proposed. That doesn't mean it's good. |
| 09:06 | <hsivonen> | I emailed vidoop. I wonder if they'll answer and with what answer. |
| 09:07 | <hsivonen> | Hixie: hey, that's the argument against the Public Suffic List |
| 09:07 | <Hixie> | i'm not saying we shouldn't have EVs |
| 09:07 | <Hixie> | and the public suffix list sucks too, yes |
| 09:07 | <Hixie> | :-) |
| 09:07 | zcorpan | thinks they'll answer "What is MITM?" |
| 09:09 | <Hixie> | (and before someone thinks i'm only picking on technologies i'm not working on, html sucks in many ways too. It, like the other two, just happens to be the best we have in the real world.) |
| 09:09 | <hsivonen> | zcorpan: their core business is making people feel more secure, so one would hope they'd know what MITM is |
| 09:10 | <MikeSmith> | Hixie: yeah, the point is that there was not anything better proposed, and that coming up with anything at all involved burning up a lot of time negotiating things with the CAs, and not all browser projects cared to actually show up to bother to take the time to actually involve themselves actively in the discussions |
| 09:10 | <Hixie> | oh? |
| 09:10 | <Hixie> | who wasn't involved? |
| 09:11 | <MikeSmith> | Opera was involved, Microsoft was involved, George Staikos was involved |
| 09:14 | <Hixie> | interesting |
| 09:14 | <Hixie> | anyway, bed time |
| 09:15 | <Hixie> | nn |
| 09:23 | <MikeSmith> | hsivonen: about http://bugzilla.validator.nu/show_bug.cgi?id=437 |
| 09:23 | <MikeSmith> | I'm trying to figure out which part of the code it is that adds the hyperlinking back to the spec |
| 09:24 | <hsivonen> | MikeSmith: nu.validator.spec.html5.Html5SpecBuilder |
| 09:24 | MikeSmith | looks now |
| 09:28 | <MikeSmith> | hsivonen: also, seems to be some brokenness in currently linking |
| 09:29 | <MikeSmith> | test, with, e.g., <!DOCTYPE html><title>foo</title><img> |
| 09:29 | <MikeSmith> | links to <img> doc end up as: |
| 09:29 | <MikeSmith> | http://www.whatwg.org/specs/web-apps/current-work/#null |
| 09:29 | <hsivonen> | MikeSmith: that code is *very* brittle :-( |
| 09:30 | <MikeSmith> | hsivonen: i see |
| 09:31 | <hsivonen> | I really should get to fixing outstanding V.nu bugs, but I'm still blocked on eliminating a key memory management issue from the C++ version of the parser. |
| 09:32 | <MikeSmith> | hsivonen: yeah, understood |
| 09:33 | <MikeSmith> | I'm happy to help with, at the very least, dealing with some of the low-hanging fruit |
| 09:33 | <MikeSmith> | as far a v.nu bugs |
| 09:33 | zcorpan | wonders how to ensure that text doesn't use non-XML 1.0 Char characters in a text-based CMS |
| 09:33 | <hsivonen> | MikeSmith: I very much appreciate your help. The recent fixes have been great. |
| 09:34 | <hsivonen> | zcorpan: run everything through a magic regexp? |
| 09:35 | <zcorpan> | i guess |
| 09:37 | <Philip`> | zcorpan: Remove all non-ASCII characters, and all below 0x20 |
| 09:38 | <Philip`> | Alternatively: Don't try to stop people inputting non-XML 1.0 Char characters |
| 09:38 | <jgraham> | zcorpan: html5lib has a big regexp somewhere |
| 09:38 | <Philip`> | and just make sure you deal with the issue when outputting to XML |
| 09:38 | <zcorpan> | Philip`: i need non-ascii characters |
| 09:38 | <jgraham> | zcorpan: Why are you trying to produce AML with a text-based CMS? |
| 09:38 | <jgraham> | *XML |
| 09:38 | <Philip`> | (because otherwise you'll forget to validate some of the input, and get invalid text in your database, and then it'll be a pain to get rid of) |
| 09:39 | <zcorpan> | jgraham: i want to have an Atom feed and the CMS i have is text-based |
| 09:39 | <jgraham> | s/somewhere/in ihatexml.py/ |
| 09:39 | <zcorpan> | lol |
| 09:39 | <Philip`> | Escape on output, it's the only way to be sure :-) |
| 09:40 | <jgraham> | zcorpan: I feel there must be a "now I have two problems" type comment in here somewhere |
| 09:41 | <MikeSmith> | heh |
| 10:19 | <MikeSmith> | http://www.whatwg.org/specs/web-apps/current-work/multipage/editing.html#spelling-and-grammar-checking |
| 10:20 | <MikeSmith> | "The spellcheck attribute is an enumerated attribute whose keywords are the empty string, true and false." |
| 10:21 | <MikeSmith> | so <p spellcheck>foo</p> should be invalid, right? |
| 10:21 | <Philip`> | That looks like an empty string to me |
| 10:21 | <MikeSmith> | but <p spellcheck="">foo</p> is valid |
| 10:21 | <MikeSmith> | Philip`: <p spellcheck> is the same as empty string? |
| 10:21 | <Philip`> | The tokeniser doesn't make any distinction between those two cases, unless I'm horribly mistaken |
| 10:21 | <MikeSmith> | OK |
| 10:21 | <Philip`> | and I'm pretty sure I'm not horribly mistaken |
| 10:21 | <Philip`> | though I could be wrong about that |
| 10:21 | <jfkthame> | Philip`: are you the one behind http://fonts.philip.html5.org/ by any chance? |
| 10:22 | <Philip`> | jfkthame: I am |
| 10:22 | <MikeSmith> | hsivonen: <p spellcheck> vs. <p spellcheck=""> ? |
| 10:22 | <jfkthame> | Philip`: cool page! however, there's a problem with at least one of the fonts |
| 10:22 | <jfkthame> | please see https://bugzilla.mozilla.org/show_bug.cgi?id=487549 |
| 10:23 | <Philip`> | jfkthame: "Not authorized" |
| 10:23 | <jfkthame> | aaarrrgghhh, sorry.... |
| 10:23 | <Philip`> | jfkthame: You could CC me on the bug if you want :-) |
| 10:24 | <jfkthame> | bugzilla id? |
| 10:24 | <Philip`> | philip.taylor⊙ccau |
| 10:24 | <jgraham> | Has Philip` inadvertently found yet another browser security bug |
| 10:24 | <Philip`> | If I had to guess wildly, it would be that it crashes on OS X |
| 10:24 | <jfkthame> | right - due to an invalid kern table in one of the subsetted fonts |
| 10:25 | <jfkthame> | you're cc'd |
| 10:27 | <jfkthame> | we could potentially do some more validation, but it's not feasible to exhaustively test fonts before handing them over to the OS or font rendering library, so if that isn't completely robust, bad fonts will always be a risk |
| 10:29 | Philip` | wonders why his copy of the font doesn't look anything like a font, and then realises after several minutes that it's been gzipped |
| 10:31 | <jfkthame> | looking at the original PixAntiqua.ttf, i see that it has the apple-format kern table, and offhand it looks valid |
| 10:31 | <jfkthame> | i guess the subsetting code is at fault, then, when it strips out all the irrelevant kern pairs from the subset font |
| 10:36 | <jfkthame> | ah, looks like the problem is Kern.pm (from Font::TTF).... it always packs the kern table header in the old format, without checking the version number to see whether that's correct |
| 10:37 | Philip` | was coming to the same conclusion :-) |
| 10:37 | <Philip`> | Is kern table version 1 documented somewhere? |
| 10:38 | <jfkthame> | http://developer.apple.com/textfonts/TTRefMan/RM06/Chap6kern.html |
| 10:38 | jgraham | notes there seems very little point in the bug being marked security sensitive now since most of the information needed to reproduce it seems to be documented here :) |
| 10:38 | <Philip`> | I suppose I'd really have to support it properly, to fix up all the glyph IDs, and can't just pass it through :-( |
| 10:39 | <Philip`> | but I'm too lazy to do that |
| 10:39 | <Philip`> | but I could make it just drop the entire kern table |
| 10:40 | <Philip`> | jgraham: It's not a clearly exploitable bug, and there's probably hundreds of other ways you could make font renderers crash because they weren't designed for untrusted input, so I suppose the relative risk is quite low :-) |
| 10:40 | <jfkthame> | jgraham: yup - 'fraid so - it's basically just another case of "feed random data to the font system and you'll probably crash something" |
| 10:41 | <Philip`> | Doesn't fill me with great confidence in the security of browsers if they're dependent on such things :-/ |
| 10:42 | <takkaria> | font renderers obviously need to stop trusting input |
| 10:42 | <jfkthame> | you mean "the security of operating systems", don't you? the browser is just a way to get a font onto the user's system |
| 10:43 | <Philip`> | I mean the security of the process of using my web browser to view a (untrusted) page |
| 10:43 | <jfkthame> | takkaria: right - it's no different from jpeg or png decoders or whatever - just that we've been attacking those for longer and so they've had a lot more hardening |
| 10:44 | <Philip`> | Were JPEG and PNG decoders as crash-prone as TTF decoders are, when they were first used in browsers? |
| 10:44 | Philip` | supposes it's quite possible they were |
| 10:44 | <Philip`> | (Has anyone tried making Theora files that crash Firefox?) |
| 10:44 | <jfkthame> | ISTR plenty of security alerts in that area over the years |
| 10:47 | jfkthame | needs to go for now - Philip`, have fun with the Theora idea! |
| 10:48 | <Philip`> | jfkthame: Thanks for pointing me at the bug :-) |
| 10:50 | <gsnedders> | Oh, yeah, it's Thursday today. |
| 10:50 | gsnedders | forgot |
| 10:55 | Philip` | fixes his code to drop the kern table |
| 10:55 | <Philip`> | I wonder how many OS X users' browsers my page crashed before that fix |
| 10:58 | <jgraham> | gsnedders: You never could get the hang of Thursdays? |
| 10:59 | <gsnedders> | jgraham: Nah, it's just it's school holidays so I have no concept of time anymore. |
| 10:59 | Philip` | is not on holiday but has no concept of time anyway |
| 11:44 | <eighty4> | gsnedders: so no concept of time == holiday? |
| 12:02 | <gsnedders> | eighty4: No, if holiday then no concept of time. The inverse does not hold true. |
| 12:03 | <eighty4> | :) |
| 12:03 | <eighty4> | I'm having 2 days completly alone tomorrow... during that time I'll probably have no concept of time |
| 12:15 | <remysharp> | Is this a sensible place to ask html5 questions - in particular about using it for new sites/pages? |
| 12:16 | <gsnedders> | remysharp: Yes |
| 12:16 | <gsnedders> | remysharp: But I would say that, wouldn't I? :) |
| 12:16 | <jgraham> | gsnedders: The right answer was no. We fail on the first clause (Is this a sensible place) |
| 12:16 | <gsnedders> | jgraham: True. |
| 12:17 | <remysharp> | Hmm, okay - how about senseless questions? |
| 12:17 | <jgraham> | remysharp: However it is still a good place to ask HTML 5 questions |
| 12:17 | <gsnedders> | jgraham: But giving the right answer would be sensible. |
| 12:17 | <jgraham> | Perhaps the best |
| 12:17 | <jgraham> | Just note the /topic |
| 12:18 | <remysharp> | So, in the <footer> element, if I have a list of elements - |
| 12:18 | <remysharp> | As per this example: http://dev.w3.org/html5/spec/Overview.html#the-nav-element |
| 12:18 | <remysharp> | shouldn't the list be in a nav element? |
| 12:18 | remysharp | hmm - was that the right link... |
| 12:19 | <gsnedders> | remysharp: "Not all groups of links on a page need to be in a nav element — only sections that consist of primary navigation blocks are appropriate for the nav element." |
| 12:19 | <gsnedders> | remysharp: They aren't primary navigation links |
| 12:19 | <remysharp> | gsnedders: ah, right, so only "primary" - great - thanks. |
| 12:20 | <remysharp> | So I've been using a few live examples around the web to help me code my html5 page - |
| 12:20 | <remysharp> | and I've been looking at the uxlondon.com site - |
| 12:20 | <remysharp> | which uses the <div class="section"> method to get around having to use JS to trigger IE to see html |
| 12:20 | <remysharp> | so - my question is - |
| 12:20 | <jgraham> | (On the opther hand it does make some sense that the primary navigation might be partially in the footer so the content model restriction maybe doesn't make sense) |
| 12:21 | <remysharp> | is there a real reason why all their 'section's contain 'article's? |
| 12:21 | <remysharp> | I wouldn't have always nested an article in a section |
| 12:21 | <remysharp> | and the spec description of an article is user generate content or article blog entry, etc. |
| 12:22 | <remysharp> | example: http://uxlondon.com/speakers/ |
| 12:22 | <jgraham> | remysharp: At first glance, no |
| 12:22 | <remysharp> | any thoughts on that at all? or perhaps just a design choice they went with |
| 12:23 | <gsnedders> | At a further glance, no. |
| 12:23 | <jgraham> | remysharp: I think their markup is unnecessarily redundant |
| 12:23 | <remysharp> | okay, good - that's what I was thinking too - so at least I'm partially following it. |
| 12:23 | <remysharp> | Using the same page as an example (the output) |
| 12:24 | <remysharp> | I am marking up a speakers page for my own project: |
| 12:24 | <remysharp> | and I've got a list of speakers - which normally I'd put in a <ul> |
| 12:24 | <remysharp> | however - I kind of want to put each one in a <section> element with the whole thing nested inside an <article> element |
| 12:25 | <remysharp> | would that make sense? (or would you like a quick example of what I mean?) |
| 12:25 | <gsnedders> | You're asking about sense again… |
| 12:25 | <remysharp> | !! :-D |
| 12:25 | <remysharp> | yeah, sorry! |
| 12:25 | gsnedders | reads the actual question |
| 12:26 | <jgraham> | remysharp: Something like <article><h2>Speakers</h2><section><h3>J. Smith</h3>[...] |
| 12:26 | <remysharp> | I suspect a quick mock of the markup I'm suggesting might help |
| 12:26 | <remysharp> | yeah - jgraham that looks like what I was thinking |
| 12:26 | <remysharp> | so no use of <ul> at all |
| 12:26 | <jgraham> | It's not really clear why the outside bit would be <article> rather than <section> but I guess it barely makes any difference anyway |
| 12:26 | <remysharp> | but my normal html4 approach would be to use a list element, but with html5, I kinda don't want to |
| 12:26 | <gsnedders> | remysharp: I'd go for <article><h2>Speakers</h2><dl><dt>Mr John Smith<dd><p>Mr John Smith is awesome.<p>He even wrote <cite>My Magical Wonderland</cite></dl></article> |
| 12:27 | <remysharp> | but isn't Mr John Smith a header within a section? |
| 12:27 | <jgraham> | gsnedders: Doesn't work so well if you ant to generate a toc that has the speakers listed |
| 12:27 | <jgraham> | *want |
| 12:27 | gsnedders | shrugs |
| 12:27 | <beowulf> | i'd go with gsnedders |
| 12:28 | <beowulf> | in terms of that question :) |
| 12:28 | <jgraham> | remysharp: I think, assuming each spaeker will have a little description, that using section+headers is fine |
| 12:28 | <beowulf> | though i'd probably s/article/section |
| 12:28 | <gsnedders> | jgraham: I do have a vague clue about what the outlining algorithm says :P |
| 12:29 | <remysharp> | damn - is there a prefered pastebin? |
| 12:30 | <gsnedders> | The one that Google takes you to when you click on "I'm feeling lucky!" because it takes the least amount of effort to find. |
| 12:30 | <jgraham> | remysharp: Try your markup in http://gsnedders.html5.org/outliner/ |
| 12:30 | <remysharp> | jgraham: ta |
| 12:30 | <gsnedders> | (OMG! I WROTE THAT1) |
| 12:30 | <jgraham> | remysharp: (That is not a pastebin) |
| 12:30 | <remysharp> | yeah, sure - oh |
| 12:30 | <remysharp> | there's not going to be url is there? |
| 12:30 | <remysharp> | hixie has one I believe - that saves the url...trying that |
| 12:30 | <jgraham> | remysharp: Ask gsnedders :) |
| 12:31 | <jgraham> | Oh, yeah you could use the LDV |
| 12:31 | <gsnedders> | remysharp: Just allowing a textarea? |
| 12:31 | <gsnedders> | Yeah, that's on my to-do list. |
| 12:32 | <remysharp> | right - there you: |
| 12:32 | <remysharp> | http://tr.im/iv56 |
| 12:34 | <beowulf> | i'd say the first header is redundant unless it wraps the <p> and that the outer article is a section |
| 12:34 | <jgraham> | remysharp: No need for the <header> element unless you plan to make a subheading |
| 12:34 | <gsnedders> | The header elements are both needless |
| 12:34 | <beowulf> | but i'm easily the dumbest person in the room, just to clarify |
| 12:35 | <remysharp> | There would obviously be a header element at the top of the page |
| 12:35 | <remysharp> | but you're saying the name shouldn't be in a header |
| 12:35 | <jgraham> | remysharp: No, no no :) |
| 12:36 | <gsnedders> | I'm saying you gain nothing by having it in a header element |
| 12:36 | <jgraham> | <header><h2>Foo</h2></header> === <h2>Foo</h2> |
| 12:36 | <gsnedders> | A header element only is of use when you having a heading and subheading and you only want the heading to appear in the TOC |
| 12:36 | <remysharp> | sorry, yeah, doesn't gain anything |
| 12:36 | <jgraham> | The extra <header> is redundant |
| 12:36 | <gsnedders> | <header><h2>Foo</h2><h3>Bar</h3></header> === <h2>Foo</h2> in terms of TOC too |
| 12:37 | <remysharp> | gsnedders: why does the h3 text get lost then? |
| 12:37 | <remysharp> | (in the TOC) |
| 12:37 | <remysharp> | or does it read the highest level heading and use that? |
| 12:37 | <gsnedders> | jgraham: Explain! |
| 12:37 | <gsnedders> | :P |
| 12:37 | <gsnedders> | remysharp: The highest level heading is all that's used |
| 12:37 | <remysharp> | cool - that makes sense. |
| 12:38 | <remysharp> | so if a TOC was generated from that page - if I omitted the <header> on the names of speakers, their names wouldn't appear in the TOC - is that right? |
| 12:38 | <beowulf> | gsnedders: what does <header><h1>Foo</h1><h1>Bar</h1></header> come out as in the TOC? |
| 12:38 | <gsnedders> | beowulf: Foo |
| 12:39 | <gsnedders> | remysharp: They would. The h3 element would make them. |
| 12:39 | <remysharp> | right - got you. |
| 12:39 | <remysharp> | so header is only if there's mixed content and you want a specific (or the highest) to be used. |
| 12:39 | <remysharp> | that makes sense as to why it's utterly redundant in my example. |
| 12:39 | <gsnedders> | beowulf: (It's the first highest order header) |
| 12:40 | <gsnedders> | (There are, however, bugs in what the spec currently says.) |
| 12:40 | <remysharp> | So, on that same topic, would it be fair to say that in this example, the <header> is redundant: |
| 12:40 | <remysharp> | <header><h1>My site</h1><p>Tag line</p></header> |
| 12:41 | <gsnedders> | remysharp: No |
| 12:41 | <jgraham> | remysharp: Technically, no |
| 12:41 | <gsnedders> | (the p element will not be associated with any section, IIRC) |
| 12:42 | <jgraham> | It won't have any observable effect on the outline, but semantically it is right |
| 12:42 | <gsnedders> | http://www.w3.org/Bugs/Public/show_bug.cgi?id=6750 |
| 12:42 | <jgraham> | Oh, maybe it does have some observable effect |
| 12:42 | <jgraham> | then |
| 12:42 | <gsnedders> | jgraham: Not really |
| 12:42 | <gsnedders> | jgraham: Unless you have a UI which shows what element is linked to what section, it doesn't. |
| 12:42 | <jgraham> | gsnedders: That is an observable effect |
| 12:43 | <gsnedders> | Well, not in any current implementation of the algorithm seeming both of our implementations just build a TOC. |
| 12:43 | <jgraham> | gsnedders: In principle though |
| 12:43 | <remysharp> | just backpeddling to the question about using sections within an article instead of a <ul><li> collection - does that look right in essence? |
| 12:44 | <jgraham> | remysharp: Yes |
| 12:44 | <jgraham> | It is better than <ul> for sure |
| 12:44 | <remysharp> | awesome. total head f**k based on getting used to using lists for everything, but feel right. |
| 12:44 | <remysharp> | *feels right |
| 12:45 | <beowulf> | jgraham: why is it better than a list? |
| 12:46 | <jgraham> | beowulf: Even extant AT will allow you to nagivate easilly by header elements, for example |
| 12:58 | <Philip`> | If you want to generate a TOC of all the speakers, use <span class="author"> and then write a script that extracts all the author names and sticks them into a TOC list |
| 12:58 | <Philip`> | which is, like, two lines of code |
| 12:59 | <hsivonen> | Hmm. Is keygen really meant to differ from input 'in select'? |
| 13:16 | <MikeSmith> | hsivonen: I wondered about that too, when looking at your treebuilder code |
| 13:16 | <MikeSmith> | maybe worth a bugzilla to clarify |
| 13:20 | <hsivonen> | http://software.hixie.ch/utilities/js/live-dom-viewer/saved/73 |
| 13:21 | <hsivonen> | like <input> in Gecko. Not like <input> in Opera and Safari. |
| 13:24 | <hsivonen> | MikeSmith: bug filed just in case |
| 13:30 | <gsnedders> | MikeSmith: Which Irish author is it you want me to read, again? |
| 13:43 | <hsivonen> | could someone please point out to me why the following XSD regexp: "\s*(none|xMinYMin|xMidYMin|xMaxYMin|xMinYMid|xMidYMid|xMaxYMid|xMinYMax|xMidYMax|xMaxYMax)\s+(meet|slice)?\s*" |
| 13:43 | <hsivonen> | does not match the word 'none'? |
| 13:44 | <hsivonen> | ooh |
| 13:44 | <hsivonen> | now I see it |
| 13:44 | <hsivonen> | \s+ |
| 13:45 | <hsivonen> | that's a classing mis-use of XSD \s, BTW |
| 13:45 | <hsivonen> | XSD \s does *not* equal XML whitespace |
| 13:46 | <gsnedders> | WHAT!? |
| 13:46 | <hsivonen> | gsnedders: welcome to the world of i18n political correctness. Zs FTW! |
| 13:46 | <gsnedders> | Not White_Space and not Zs? |
| 13:46 | <hsivonen> | gsnedders: IIRC, Zs |
| 13:47 | <gsnedders> | That doesn't even include U+000A IIRC! |
| 13:47 | <hsivonen> | s/classing/classical/ |
| 13:48 | <hsivonen> | s/al// I suppose |
| 13:48 | <gsnedders> | Yeah |
| 13:48 | <gsnedders> | It's now right :) |
| 13:48 | Philip` | would be unable to resist the temptation to write that regexp with (none|x(Min|Mid|Max)Y(Min|Mid|Max)) |
| 13:49 | <Philip`> | (Fortunately I would be able to resist (none|xM(in|id|ax)YM(in|id|ax)) because that's just crazy) |
| 13:51 | <Philip`> | (Shouldn't the regexp start with "(defer\s+)?"?) |
| 13:53 | <hsivonen> | defer? |
| 13:54 | <Philip`> | That's what http://www.w3.org/TR/SVG/coords.html#PreserveAspectRatioAttribute says |
| 13:55 | <hsivonen> | Philip`: good point. |
| 13:56 | <hsivonen> | is defer only allowed on <image>? |
| 13:56 | <Philip`> | Sounds like it's allowed everywhere, but ignored except on <image> |
| 13:57 | <hsivonen> | heycam: ^ |
| 13:57 | <hsivonen> | heycam: the delta of the V.nu copy of the SVG 1.1 schema and the W3C copy is growing |
| 13:59 | <hsivonen> | MikeSmith: I deployed your recent checkins. |
| 14:05 | <MikeSmith> | hsivonen: thanks |
| 14:06 | <MikeSmith> | gsnedders: Flann O'Brien |
| 14:10 | <heycam> | hsivonen, spec problem? |
| 14:11 | <heycam> | or just a problem with the SVG 1.1 DTD not being as restrictive as it could be? |
| 14:37 | <hsivonen> | heycam: schema bug |
| 14:37 | <hsivonen> | heycam: schema bug in the RELAX NG schema |
| 14:55 | <heycam> | hsivonen, ok |
| 14:55 | <heycam> | so that rng isn't really official |
| 14:56 | <heycam> | we're going to be making a new one soon, for 1.1, but starting from the 1.2T rng |
| 14:58 | <heycam> | ah i see that regex you quote is from that unofficial rng |
| 14:58 | <gsnedders> | MikeSmith: Well, it's my birthday in just over a week ;P |
| 14:59 | <jgraham> | gsnedders: How old are ypu going to be 5? 6? I lose track |
| 14:59 | <gsnedders> | jgraham: 7 |
| 14:59 | <gsnedders> | Sorry, I lie. 10, in an as of yet undecided base. |
| 15:00 | <heycam> | hsivonen, erk, seems like the regex in the 1.2T relaxng is horribly wrong! |
| 15:00 | <heycam> | \s*(none|xMidYMid)\s*(meet)?\s* |
| 15:00 | heycam | raises an issue |
| 15:02 | <krikey> | I was sat like billy no mates in #html5 |
| 15:02 | <krikey> | someone could have coma and got me :) |
| 15:02 | <heycam> | http://www.w3.org/Graphics/SVG/WG/track/issues/2257 |
| 15:02 | <jgraham> | Hard to get someone if you're in a coma |
| 15:03 | <krikey> | ye |
| 15:03 | heycam | goes to watch some more daily show |
| 16:26 | <Philip`> | heycam: That issue fails to mention that it accepts strings like "nonemeet" too |
| 16:26 | <jgraham> | http://www.ecma-international.org/news/PressReleases/PR_Ecma_finalises_major_revision_of_ECMAScript.htm |
| 16:27 | <cryzed> | Hey :) |
| 16:27 | <cryzed> | Is someone from the html5lib for Python here? |
| 16:27 | <jgraham> | cryzed: Yes |
| 16:28 | <cryzed> | great :) |
| 16:28 | <jgraham> | "TC39 members will create and test implementations of the candidate specification to verify its correctness and the feasibility of creating interoperable implementations". I wonder if they mean interoperable implemetations that can ship on the web |
| 16:28 | <cryzed> | So basically I want to know |
| 16:29 | <cryzed> | Do I get the lxml.html parser with this : html5lib.HTMLParser(tree=treebuilders.getTreeBuilder("lxml"))? |
| 16:29 | <Philip`> | jgraham: Try reading the next sentence |
| 16:29 | <Philip`> | "The test implementations will also be used for web compatibility testing to ensure that the revised specification remains compatible with existing web applications." |
| 16:29 | <jgraham> | Philip`: Oh. |
| 16:30 | <Philip`> | cryzed: No - that uses html5lib's HTML5 parser (and constructs an lxml document from it), not lxml's non-standard HTML parser |
| 16:30 | <jgraham> | cryzed: Yes, unless you use the latest svn in which case the best option is to use html5lib.parse(input, tree="lxml") |
| 16:31 | <cryzed> | Well |
| 16:31 | <cryzed> | what now? |
| 16:31 | <jgraham> | Oh, sorry, I misunderstood the question |
| 16:31 | <jgraham> | cryzed: What do you actually want to do |
| 16:31 | <cryzed> | wait |
| 16:31 | <cryzed> | http://paste.pocoo.org/show/ad9DwVRGDKIhXNgqSk8j/ |
| 16:31 | jgraham | is not doing very well at reading at the moment |
| 16:31 | <cryzed> | I want to parse my blog with the html5lib |
| 16:32 | <cryzed> | and then scrape it with the resulting elementtree |
| 16:32 | <cryzed> | Unfortunately I don't really find any documentation for the ElementTree except http://effbot.org/zone/pythondoc-elementtree-ElementTree.htm#elementtree.ElementTree.ElementTree-class |
| 16:32 | <jgraham> | cryzed: That looks vaugely sensible. What is the problem? |
| 16:33 | <cryzed> | http://paste.pocoo.org/show/KHOsRJHJNnHNvS9ec0Bg/ that should work |
| 16:33 | <cryzed> | doesn't though |
| 16:33 | <jgraham> | http://codespeak.net/lxml/tutorial.html |
| 16:33 | <cryzed> | http://paste.pocoo.org/show/ld7f4BUjcUZbnyFG4vGd/ |
| 16:33 | <cryzed> | ah... |
| 16:33 | <cryzed> | the case.. |
| 16:34 | <jgraham> | findall()? |
| 16:34 | <cryzed> | yes |
| 16:34 | <cryzed> | and I DO need to supply |
| 16:34 | <cryzed> | an argument |
| 16:34 | <cryzed> | which argument do I need to supply to find all tags? |
| 16:34 | <cryzed> | . |
| 16:34 | <cryzed> | ? |
| 16:35 | <jgraham> | if you just want all the child nodes you can just do "for item in element:" |
| 16:35 | <cryzed> | etree.findall(".//*"): that works aswell |
| 16:35 | <cryzed> | thanks though jgraham :) |
| 16:35 | <cryzed> | Is the argument which I pass to findall called a "xpath"? |
| 16:35 | <cryzed> | ElementPath |
| 16:36 | <cryzed> | found it.. |
| 16:36 | <cryzed> | sorry I seem to ask only stupid questions |
| 16:36 | <jgraham> | cryzed: If you know xpath and are using lxml you can do element.xpath(xpath_expression) |
| 16:37 | <cryzed> | Or I can just use the ElementPath? |
| 16:37 | <cryzed> | http://effbot.org/zone/element-xpath.htm |
| 16:37 | <jgraham> | e.g. element.xpath(".//a") finds all a descendants |
| 16:37 | <jgraham> | cryzed: ElementPaths are like a subset of XPath 1.0 |
| 16:38 | <cryzed> | Hrmm, I don't think I do need full xpath support, thanks for the tip though |
| 16:38 | <cryzed> | jgraham, I read on the lxml.html documentation |
| 16:38 | <cryzed> | about the following function: |
| 16:38 | <cryzed> | *method |
| 16:38 | <cryzed> | .text_content(): |
| 16:38 | <jgraham> | Also if you use a really up to date lxml you can probably get CSS Selectors |
| 16:39 | <cryzed> | this isn't available in the lxml etree, right? |
| 16:39 | <jgraham> | cryzed: html5lib just generates an lxml tree. It has all the features of whichever lxml you have installed |
| 16:39 | <cryzed> | well, yes |
| 16:39 | <cryzed> | the problem is |
| 16:40 | <cryzed> | lxml.html |
| 16:40 | <cryzed> | the html Etree is a special tree |
| 16:40 | <cryzed> | How do I tell html5lib |
| 16:40 | <cryzed> | to use the lxml.html tree? |
| 16:40 | <jgraham> | Oh, yeah |
| 16:40 | <smedero> | jgraham: I don't think the .text_content() method exist in lxml.etree |
| 16:40 | <cryzed> | Is there any way to use the lxml.html tree? |
| 16:40 | <jgraham> | So, I don't think you can at the moment because of some weirdness in the way that lxml is set up |
| 16:40 | <cryzed> | I think this would be really comfortable for webscraping |
| 16:41 | <jgraham> | At least that is my recollection from when I implemented this stuff a while ago |
| 16:41 | <Philip`> | etree.tostring(node, method='text') |
| 16:41 | <Philip`> | might be similar to node.text_content() |
| 16:41 | <smedero> | yeah, that should be in the ballpark |
| 16:41 | <jgraham> | It's something like you can't create comments in lxml.html or... |
| 16:41 | <jgraham> | .xpath(".//text()) works |
| 16:42 | <jgraham> | .xpath(".//text()") |
| 16:42 | <cryzed> | thanks |
| 16:42 | <cryzed> | I found in the lxml.html implementation the following |
| 16:42 | <cryzed> | _collect_string_content(self) |
| 16:42 | <cryzed> | should work if there is no other way |
| 16:59 | <cryzed> | the .text attribute |
| 16:59 | <cryzed> | works beautifully |
| 17:01 | <cryzed> | ..not |
| 17:02 | <jgraham> | cryzed: for <a><b>foo</b>bar</a> a.text == None |
| 17:02 | <jgraham> | b.text == "foo" |
| 17:02 | <jgraham> | b.tail =="bar" |
| 17:03 | <cryzed> | is there any way to get the whole text and protect the formatting? |
| 17:03 | <cryzed> | for example replace <br> |
| 17:03 | <cryzed> | with \n |
| 17:03 | <jgraham> | cryzed: Not an easy way that I know of |
| 17:03 | <cryzed> | hrm |
| 17:03 | <jgraham> | You would need to walk the tree, normalize whitespace and make whatever replacements you want |
| 17:05 | <cryzed> | http://paste.pocoo.org/show/5gbDet52tiXSMRWkLZ4y/ ? |
| 17:06 | <jgraham> | That will only do children of the td |
| 17:07 | <cryzed> | yes |
| 17:07 | <cryzed> | that's what I want actually |
| 17:08 | <cryzed> | sorry if I start to get annoying |
| 17:08 | <cryzed> | but why doesn't that work: http://paste.pocoo.org/show/cfmGTUkDaWSCr1ehDH3X/ ? |
| 17:13 | <jgraham> | Trying to get the "blockquote" attribute of a "blockquote" element? |
| 17:14 | <cryzed> | well.. kinda |
| 17:14 | <cryzed> | ^^ |
| 17:14 | <cryzed> | It works in BeautifulSoup :D... |
| 17:14 | <jgraham> | Er, what does it do? |
| 17:14 | <jgraham> | I mean if you really have <blockquote blockquote=something> I guess it should work |
| 17:15 | <cryzed> | it should get me the text WITHOUT markup WITH formatting out of the blockquote tags |
| 17:15 | <cryzed> | >IN BETWEEN HERE< |
| 17:16 | <gsnedders> | with formatting without markup? how? |
| 17:16 | <cryzed> | well |
| 17:16 | <cryzed> | for example |
| 17:16 | <jgraham> | cryzed: If you just want the test you can do element.xpath(".//text()") |
| 17:16 | <gsnedders> | jgraham: You don't need XPath for that! Peh! |
| 17:17 | <cryzed> | <pre>That's some fancy text <br>comment'</pre> |
| 17:17 | <cryzed> | Should result to |
| 17:17 | <jgraham> | If you want to do some formatting on the text you need to decide what formatting you want |
| 17:17 | <cryzed> | That's some fancy text |
| 17:17 | <cryzed> | comment |
| 17:17 | <cryzed> | gsnedders, what should I use? |
| 17:17 | <gsnedders> | jgraham: return etree.tostring(element, encoding=unicode, method='text', with_tail=False) is better than that |
| 17:18 | <cryzed> | I can't acces this function |
| 17:18 | <jgraham> | And implement that by e.g. walking the tree replacing <br> with "\n" and adding the .tail of the br to the right place |
| 17:18 | <jgraham> | gsnedders: Define "better" |
| 17:18 | <gsnedders> | jgraham: Quicker |
| 17:18 | <gsnedders> | jgraham: The result is identical :P |
| 17:19 | <jgraham> | gsnedders: Seems unlikely to be a problem in this case |
| 17:19 | <jgraham> | It is much longer to type and easier to get wrong (maybe) |
| 17:19 | gsnedders | just wraps it in a function :P |
| 17:19 | <cryzed> | etree doesn't got the attribute .tostring |
| 17:19 | <gsnedders> | cryzed: from lxml import etree |
| 17:20 | <cryzed> | I originally only wanted to import html5lib :|.. |
| 18:01 | <cryzed> | Is it a good idea to use the latest svn build? |
| 18:01 | <gsnedders> | cryzed: Absolutely. |
| 18:02 | <gsnedders> | cryzed: It's a better idea than using the latest release |
| 18:02 | <cryzed> | Okay |
| 18:02 | <cryzed> | btw |
| 18:02 | <cryzed> | gsnedders, if I had an custom BeautifulSoup.py |
| 18:02 | <cryzed> | *a |
| 18:02 | <cryzed> | Could I somehow tell the treebuilder |
| 18:02 | <cryzed> | to use this this BeautifulSoup? |
| 18:03 | <cryzed> | http://furyu-tei.sakura.ne.jp/archives/BSXPath.zip I found this |
| 18:03 | <cryzed> | it has xpath support as it seems |
| 18:03 | <Philip`> | html5lib's BeautifulSoup support is not particularly reliable |
| 18:03 | <cryzed> | Yeah, but it works |
| 18:03 | <cryzed> | I did some things with it |
| 18:03 | <cryzed> | And I somehow think that using BeautifulSoup is easier than lxml |
| 18:04 | <Philip`> | It works as long as you don't do one of the things that doesn't work :-) |
| 18:04 | <cryzed> | e.g? |
| 18:05 | <Philip`> | e.g. http://code.google.com/p/html5lib/issues/detail?id=80 |
| 18:05 | <cryzed> | oh.. |
| 18:06 | <cryzed> | lxml |
| 18:06 | <cryzed> | doesn't work any better |
| 18:06 | <cryzed> | lol |
| 18:06 | <cryzed> | html5lib.HTMLParser(tree=html5lib.treebuilders.getTreeBuilder("lxml")).parse("<a><div><div><a>") |
| 18:06 | <cryzed> | >>> etree.tostring(e) |
| 18:06 | <cryzed> | '<html><head/><body><a/><div><a></a><div><a></a><a/></div></div></body></html>' |
| 18:06 | <Philip`> | That works much better, since it gives the right output and doesn't throw exceptions :-p |
| 18:06 | <cryzed> | right? |
| 18:06 | <cryzed> | oh.. you are right |
| 18:06 | gsnedders | guesses you want the html5lib serializer |
| 18:07 | <Philip`> | cryzed: "Right" according to the HTML5 spec |
| 18:07 | <cryzed> | Oke |
| 18:07 | <cryzed> | Guys let me tell what I want to do |
| 18:07 | <Philip`> | Bit peculiar how it mixes "<a></a>" and "<a/>"... |
| 18:07 | <cryzed> | Oke |
| 18:07 | <cryzed> | I'm sure some of you know 4chan |
| 18:08 | <cryzed> | I want to write a Python "API" for it |
| 18:08 | <cryzed> | take a look at this page for example: |
| 18:08 | <cryzed> | http://zip.4chan.org/x/res/1648607.html |
| 18:08 | <cryzed> | So |
| 18:08 | <cryzed> | each reply has got an id attribute and a class |
| 18:08 | <cryzed> | Well, basically I want to use lxml |
| 18:08 | <cryzed> | to scrape this site |
| 18:08 | <cryzed> | get the text between the blockquotes |
| 18:09 | <cryzed> | remove the tags |
| 18:09 | <cryzed> | and replace the <br> tags with \n |
| 18:09 | <cryzed> | and save it into a variable |
| 18:09 | <cryzed> | at the this is all getting wrapped in a class called Reply |
| 18:09 | <cryzed> | *at the end |
| 18:10 | <cryzed> | http://iohosaf.pastebin.com/d5b10f992 |
| 18:10 | <cryzed> | This is what I've got so far |
| 18:10 | <cryzed> | in the implementation with BeautifulSoup |
| 18:10 | <cryzed> | works, but is ugly imho |
| 18:13 | <Philip`> | (On your original question about using a different BeautifulSoup.py: It may be sufficient to just put it in Python's search path, like by using PYTHON_PATH=some-directory-which-contains-that-file) |
| 18:13 | <cryzed> | Philip`, okay |
| 18:13 | <cryzed> | I could just rename it to BeautifulSoup.py |
| 18:13 | <cryzed> | and place it locally |
| 18:13 | <cryzed> | next to my script |
| 18:13 | <cryzed> | probably |
| 18:13 | <Philip`> | I'm not certain but I think that ought to get picked up when html5lib tries loading it |
| 18:14 | Philip` | has to go away |
| 18:15 | <cryzed> | Well oke |
| 18:16 | <cryzed> | thanks Philip` |
| 18:16 | <cryzed> | I'll get myself the newest html5lib |
| 18:16 | <cryzed> | and just tune my old lib |
| 18:16 | <cryzed> | with BeautifulSoup a bit |
| 18:16 | <cryzed> | and hope that it works |
| 19:42 | <hsivonen> | "I don't think anyone wants to break plugins" - iPhone, anyone? |
| 19:42 | <hsivonen> | (quote from HTML WG telecon minutes) |
| 19:43 | <smedero> | it was a surreal a couple of moments... for sure. |
| 19:43 | <gsnedders> | hsivonen: Well, arguably it never broke |
| 20:45 | <tantek> | hsivonen - neither iPhone nor BlackBerry browsers support Flash / plugins |
| 21:07 | gsnedders | feels bad… |
| 21:07 | <gsnedders> | Appealing to authority :( |
| 21:08 | <takkaria> | some authority is good |
| 21:08 | <gsnedders> | annevk? |
| 21:16 | <gsnedders> | ARGH! |
| 21:16 | gsnedders | gets annoyed at Google again |
| 21:16 | <gsnedders> | I can't simply look for anything relating to Lolita without getting a ton of results of porn |
| 21:26 | <Philip`> | gsnedders: Maybe you shouldn't be using the image search with SafeSearch off |
| 21:27 | <gsnedders> | Philip`: I'm not using image search. That doesn't help me write an English dissertation. |
| 21:40 | <gsnedders> | Hmm… Nabokov almost always uses to sob and rarely to cry… |
| 22:42 | <Hixie> | heycam: yt? |
| 22:43 | <Hixie> | i have a method take takes as its argument an array of values |
| 22:43 | <Hixie> | the values are typed |
| 22:44 | <Hixie> | er, i mean, an array of arrays of values, which are typed |
| 22:44 | <Hixie> | let's say, it's an array of arrays of DOMString, long pairs |
| 22:44 | <Hixie> | e.g. |
| 22:44 | <Hixie> | foo([['a', 1], ['b', 2], ['c', 3]]); |
| 22:44 | <Hixie> | is there a sane way to describe that in WebIDL? |