| 00:20 | <jacobolus> | ooh, this one is fun: http://ejohn.org/blog/meta-madness/ |
| 00:21 | <hober> | yup |
| 00:22 | <jacobolus> | it's a fun day to be Chris Wilson |
| 00:22 | <hober> | fsvo fun |
| 00:46 | <zcorpan_> | Hixie: have you confirmed your theory about http://ln.hixie.ch/?start=1190803943&count=1 ? :) |
| 01:31 | <Hixie> | zcorpan_: yeah, most people seem to use the tip line |
| 01:35 | <Philip`> | I really should get a blog, just so I can put invalid UTF-8 in its title and try trackbacking people who use XHTML |
| 01:37 | <Hixie> | heh |
| 01:41 | <jacobolus> | haha |
| 01:43 | <othermaciej> | Philip`: you're mean |
| 01:44 | <othermaciej> | jacobolus: Microsoft is getting more support for their move than I would have expected |
| 01:45 | <jacobolus> | othermaciej: the comments everywhere seem overwhelmingly negative to me |
| 01:45 | <othermaciej> | jacobolus: they got some positive remarks from big shot web standards people like Eric Meyer |
| 01:45 | <othermaciej> | and some supportive comments from web developers |
| 01:45 | <othermaciej> | I do see some negativity as well |
| 01:46 | <othermaciej> | though much of it from Mozilla-affiliated bloggers |
| 01:46 | <jacobolus> | yeah, and Zeldman |
| 01:47 | <jacobolus> | I really don't think Meyer et al. have done a great job explaining their 180 on the question though |
| 01:47 | <Hixie> | i didn't understand eric's article |
| 01:47 | <Hixie> | it seemed uncharacteristically rambly |
| 01:47 | <othermaciej> | I was unable to follow Eric's reasoning |
| 01:47 | <othermaciej> | I do think arguments could be made for Microsoft's choice |
| 01:47 | <othermaciej> | I don't know if they have been expounded that well |
| 01:48 | <jacobolus> | I think mostly that those arguments underestimate the benefits of getting browsers on the same page |
| 01:48 | hdh | doesn't get the post's graphics |
| 01:49 | <Hixie> | the strange thing about this move is that it would make a lot of sense if they had 95% of the market and wanted to block competition |
| 01:49 | <jacobolus> | which is odd because the same people have spent years explaining the other side |
| 01:49 | <Hixie> | but in their position, bleeding market share, it actually hurts them |
| 01:49 | <othermaciej> | Hixie: so it's 5 years late? |
| 01:50 | <Hixie> | and it could be near-fatal if they ever get to low enough numbers that they are no longer the leading browser |
| 01:50 | <Hixie> | othermaciej: yes |
| 01:50 | <jacobolus> | othermaciej: I guess the main thing is I don't see the arguments working out at all for public-facing websites |
| 01:50 | <hdh> | does he mean to say a) it's like cc, numbered, documented and b) doesn't affect other browsers like "like Gecko"? |
| 01:50 | <jacobolus> | where any kind of x-browser compat is desired whatsoever |
| 01:51 | <othermaciej> | jacobolus: for public-facing web sites that are actively maintained, they'd probably want to either upgrade the switch for every IE release, or never upgrade it and be stuck in IE7 mode forever |
| 01:51 | <othermaciej> | or something |
| 01:51 | <jacobolus> | the idea that this new switch will allow microsoft to do a better job innovating seems completely ludicrous |
| 01:52 | <othermaciej> | I don't think they are interested in innovating on top of standards-based web technologies |
| 01:52 | <othermaciej> | so it doesn't much matter |
| 01:53 | <jacobolus> | sorry, i mean bug fixing: “This actually makes browser vendors more susceptible to pressure to fix their bugs, and less fearful of doing so.” |
| 01:53 | <jacobolus> | I just don't see that increased pressure coming out of this |
| 01:54 | <othermaciej> | I don't think it increases pressure |
| 01:54 | <othermaciej> | it somewhat decreases reasons *not* to fix bugs |
| 01:55 | <othermaciej> | but also decreases the web-wide benefit of fixing bugs |
| 01:55 | <jacobolus> | someone should start some sort of betting pool about percentage of pages which use the new meta tag within 6 months of IE 8 being out |
| 01:55 | <jacobolus> | assuming they don't alter their plans between now and then |
| 01:56 | <Philip`> | Percentage of which finite set of pages? |
| 01:56 | <othermaciej> | probably fewer than the percentage of users using it |
| 01:56 | <othermaciej> | (using IE8 that is) |
| 01:56 | <jacobolus> | Philip`: I dunno… I'm mostly joking :) |
| 01:57 | <Philip`> | Only "mostly"? :-) |
| 01:57 | <jacobolus> | well, in the sense that I don't actually plan to make any bets, but would be interested in seeing the studies about adoption |
| 01:58 | <Philip`> | I'd just make a CGI script which produces an infinite number of pages using the meta tag |
| 01:59 | <jacobolus> | and given that there are already infinitely many without, it becomes impossible to compare? |
| 02:01 | <Philip`> | If someone else has an infinite set of pages without the meta tag, where each is identified by some URL µ, I'll make a CGI script which you call like 'exciting.cgi?id=n;w=µ' for any positive integer n, and therefore there'll be infinitely many more pages with my meta tag than there are others without |
| 02:02 | <Philip`> | and you can't argue with maths! |
| 02:03 | <roc> | sorry |
| 02:03 | <roc> | those sets are the same size |
| 02:04 | <Philip`> | Hmph |
| 02:05 | <Philip`> | How about I have 'exciting2.cgi?ids=a,b,c,...' where you pass a comma-separated set of URLs, for each member of the power set of URLs |
| 02:06 | <Philip`> | (Is there any RFC that requires URLs to be of finite length?) |
| 02:09 | <roc> | as long as URLs are of finite length, you lose |
| 02:10 | <Hixie> | not really, a large enough number is all you really need |
| 02:10 | <Hixie> | assuming you're trying to bias a result |
| 02:10 | <Hixie> | bloo2, meet Philip` |
| 02:10 | <Hixie> | Philip`, meet bloo2 |
| 02:10 | <Hixie> | you have both worked on studies of the Web |
| 02:11 | <Hixie> | markup and that kind of thing |
| 02:11 | <bloo2> | hi philip 8-} |
| 02:11 | <jacobolus> | Philip`: wait a second there, both are countably infinite |
| 02:11 | <jacobolus> | Philip`: if you can make an uncountable set of pages, you can certainly win the bet though |
| 02:11 | <Hixie> | aren't URIs countable by definition? |
| 02:11 | <Hixie> | since they're bit strings |
| 02:12 | <jacobolus> | Hixie: that's why making an uncountable set of them is damn impressive |
| 02:12 | <roc> | yes |
| 02:12 | <Hixie> | ok, just checking |
| 02:12 | <roc> | Philip` is going to argue that infinite URIs are allowed by the spec |
| 02:12 | <Philip`> | bloo2: Good morning |
| 02:12 | <jacobolus> | lol |
| 02:12 | <Hixie> | i stick by my original point that if you're just trying to bias a sample, though, you just need a large number of uris, not an infinite number |
| 02:12 | <Hixie> | but i wasn't here for the start of the discussion, so... |
| 02:13 | <Philip`> | If you're trying to bias a sample, the best thing to do is be the person who picks the sample |
| 02:13 | <Hixie> | indeed |
| 02:14 | <Hixie> | my prices start at $1000. |
| 02:14 | <Hixie> | depending on how much of a skew you want. |
| 02:14 | <bloo2> | philip`: Hixie had mentioned you'd done some work in the area recently, but I hadn't followed up. Pleased to meet you (hrm. Must register to send a prv msg...have to look in to that) |
| 02:15 | <jacobolus> | Hixie: the start of it was my curious musing about the adoption of IE's new meta tag, after IE8 has been on the market a few months, plus an apparently ill-advised reference to betting :) |
| 02:15 | <Hixie> | hah |
| 02:16 | <jacobolus> | Philip`: a better bet might be about how whether microsoft.com uses their new tag ;) |
| 02:16 | <Philip`> | bloo2: Indeed - I did some stuff like http://canvex.lazyilluminati.com/survey/2007-07-17/analyse.cgi/index a while ago, and looked at more/different things recently |
| 02:17 | <Philip`> | jacobolus: Which part of microsoft.com? :-) |
| 02:17 | <jacobolus> | but I still wouldn't put real money on it :) |
| 02:17 | <jacobolus> | the front page? |
| 02:17 | <jacobolus> | any part of it whatsoever? |
| 02:17 | <Philip`> | (It was msdn.microsoft.com that originally made me switch from IE to Firefox, because I wanted to look up some documentation but it kept freezing and crashing in IE) |
| 02:18 | <Philip`> | jacobolus: I bet at least one part will have it, since they'll give examples in the documentation |
| 02:19 | <jacobolus> | heh |
| 02:19 | <jacobolus> | do they give examples as complete pages? |
| 02:20 | <Philip`> | Er, I'd guess so, but I don't actually know |
| 02:20 | <bloo2> | philip`: (question about the research) what is "Duplicate attribute names"? |
| 02:20 | <jacobolus> | anyway, gotta run. have fun with the rest of standards flamewar day, everyone |
| 02:21 | bloo2 | considered doing a 6 month metric test of IE's new META syntax to see if there is any uptake. |
| 02:21 | <Philip`> | bloo2: That's when an element has >= 2 attributes with the same name |
| 02:21 | <bloo2> | philip`: interesting. I hadn't thought of that one before. 8-} |
| 02:22 | <Philip`> | (Since it's only counting the number of pages with duplicates, it doesn't matter if it's 2 or 3 etc) |
| 02:22 | <Philip`> | bloo2: I was mostly looking for duplicate style attributes, since that potentially has to be handled specially for compatibility with existing content |
| 02:23 | <jwalden> | https://bugzilla.mozilla.org/show_bug.cgi?id=postMessage |
| 02:23 | <Philip`> | (but it looks like it's not a huge problem in practice) |
| 02:23 | <jwalden> | ^ this has a reasonable chance of being in Firefox 3 |
| 02:23 | <jwalden> | amazingly enough |
| 02:23 | gavin | would be interested in seeing the uptake of the new meta tag before IE8 is released |
| 02:24 | <Philip`> | (That data came from an ugly slow C++ / Perl mixture with various problems that cause things like "u000D" to appear in the output - I've switched to a nicer faster Java approach now) |
| 02:25 | bloo | works on trying to find an open nick |
| 02:27 | <Philip`> | gavin: http://www.cl.cam.ac.uk/~pjt47/misc/attributes.html and scroll down to "meta http-equiv": uptake was 0% about a month ago, so there's one data point already |
| 02:27 | <gavin_> | heh |
| 02:28 | <blooberry> | philip`: I found the same thing today from my research of the full dmoz set. 8-} We can consider any search before today a good "starting point" I think |
| 02:49 | jruderman | wonders whether hixie plans to put "X-UA-Compatible" into acid2 or acid3 |
| 02:49 | <Hixie> | hahahaha no. |
| 02:50 | <eseidel_> | Hixie: I assume you saw the security bug... |
| 02:50 | <eseidel_> | about data: |
| 02:51 | <eseidel_> | and Acid3 |
| 02:51 | <Hixie> | haven't looked at bugmail yet |
| 02:53 | <jruderman> | is IE8's X-UA-Compatible really "render this exactly as IE7 would" or is it a limited number of quirks? in particular, say IE8 adds support for text-shadow; will you have to use X-UA-Compatible for text-shadow to work? |
| 02:55 | <Hixie> | IE7's quirks mode is "render this exactly as IE5.5 would" |
| 02:55 | <Hixie> | i would expect the same to apply |
| 02:55 | <Hixie> | in particular, i'd expect new features not to be supported |
| 02:55 | <Hixie> | just like they aren't in the current fewer-quirks quirks mode |
| 02:58 | <jruderman> | ok |
| 03:25 | <othermaciej> | Hixie: is there any info technically on whether IE7 actually has two separate layout engines, or just one with a bunch of conditional quirks? |
| 05:10 | <jruderman> | othermaciej: http://ejohn.org/blog/meta-madness/#comment-296861 ? |
| 05:11 | <jruderman> | "I'll detail more about the Javascript solution in the future - the short answer is that it's not two engines, it's two modes of a single engine; we can, in fact, work across boundaries like that." |
| 05:28 | <othermaciej> | jruderman: can't tell if that's referring to the JavaScript engine or the whole rendering engine |
| 05:36 | <roc> | me neither but it sounds more like the JS engine |
| 05:36 | <gavin_> | when I read it I assumed that his "it" refers to the overall solution |
| 05:37 | <roc> | we'll see |
| 05:38 | <othermaciej> | I assume some chunk of code has to be duplicated to totally freeze all bugs |
| 05:38 | <othermaciej> | unless they are super anal about conditionalizing every single code change |
| 05:38 | <othermaciej> | in which case I feel even more sorry for them |
| 05:52 | <othermaciej> | wow, there are a lot of blog posts about the IE version thing |
| 06:01 | <roc> | othermaciej: yeah |
| 06:01 | <roc> | othermaciej: shipping multiple engines sucks, but weaving them together in your source code must suck even more |
| 06:03 | <roc> | and the first time you want to rearchitect your code in a substantial way, you have to fork anyway |
| 07:12 | <MikeSmith> | kind of amazed to see so much attention about a change related to a product that hasn't even shipped yet |
| 07:13 | <MikeSmith> | and that nobody knows when it will actually ship |
| 07:14 | <MikeSmith> | the option remains of them not actually implementing this |
| 07:14 | <MikeSmith> | it's not like just because they have announced it preemptively, it's now written in stone |
| 07:15 | <MikeSmith> | I'm a bit confused about why they did go ahead and announce it already |
| 07:15 | <MikeSmith> | the timing -- what benefit it was to them to do so now |
| 07:16 | <MikeSmith> | worries that word about it would get leaked early? |
| 07:23 | <othermaciej> | information is good |
| 07:23 | <othermaciej> | I would not want to discourage them |
| 07:28 | <Hixie> | othermaciej: i don't know exactly what IE7 is like inside, but it certainly seems to have two engines, as they don't even, e.g., support any new features in quirks mode |
| 07:29 | <othermaciej> | it could just mean every change is bracketed with an if statement |
| 07:29 | othermaciej | shrugs |
| 07:29 | <othermaciej> | Hyatt seems to think IE7's quirks mode is based on conditionals, not multiple engines, but I'm not sure how he would know |
| 07:33 | <Hixie> | yeah but god, can you imagine that? |
| 07:33 | <Hixie> | the code would suck so bad |
| 07:34 | <Hixie> | and as roc said, what if you rearchitect something? like their removal of hasLayout |
| 07:40 | <Hixie> | i mean, if you want to change something like your class hierarchy, which they well have to for what they've done in IE8... |
| 07:46 | <othermaciej> | at some point the whole engine has to be an if statement |
| 07:46 | <othermaciej> | roc, annevk: http://webkit.org/blog/155/versioning-compatibility-and-standards/ |
| 07:46 | <othermaciej> | (since y'all suggested I should post something) |
| 07:57 | <hsivonen> | Aaron Gustafson contradicts what Andy Clarke said on Anne's site: http://www.webstandards.org/2008/01/22/ie8-will-see-the-smile/ |
| 08:06 | <othermaciej> | hsivonen: I'm slow - what's the contradiction? |
| 08:07 | <hsivonen> | othermaciej: whether or not the people involved were acting as individuals or as WaSP |
| 08:08 | <hsivonen> | annevk: aside: http://annevankesteren.nl/2008/02/ is 500 |
| 08:09 | <Hixie> | http://www.w3.org/mid/1A373905-DEAF-429E-B358-C8A49CAD550F⊙yc is funny, because Mark basically says that HTTP headers are "whole new way to associate policy with resources" |
| 08:09 | <Hixie> | which is news to me! |
| 08:11 | <othermaciej> | hsivonen: I don't see him claiming that this was an official WaSP project, just that Microsoft approached WaSP |
| 08:12 | <hsivonen> | othermaciej: oh. good point. so the exact wordings don't contradict each other, but at least I get the impression from the WaSP post that WaSP was consulted as WaSP even if it doesn't strictly say that |
| 08:17 | <othermaciej> | this post (also on the WaSP site) is pretty unambiguous: http://www.webstandards.org/2008/01/22/microsofts-version-targeting-proposal/ |
| 08:19 | <hsivonen> | yeah |
| 09:14 | <jacobolus> | othermaciej: excellent statement :) |
| 09:16 | <annevk> | hsivonen, hmm |
| 09:17 | annevk | has an idea as to what the issue is |
| 09:23 | <krijnh> | http://alistapart.com/comments/beyonddoctype?page=10#97 |
| 09:27 | <takkaria> | he seems to think people don't upgrade their browsers for compatibility reasons |
| 09:27 | <krijnh> | For IE that's probably partly true |
| 09:27 | <krijnh> | And partly in IE terms means a lot of users |
| 09:28 | <takkaria> | I think the vast, vast majority of web browser users don't know or care |
| 09:28 | <Hixie> | it wouldn't be as big an issue if IE didn't have so many bugs |
| 09:28 | <Hixie> | this is only really an issue because IE6 became entrenched after being left unmaintained for over half a decade |
| 09:28 | <Hixie> | nearly one third of the lifetime of the web |
| 09:29 | <Hixie> | it wouldn't be nearly as bad for future releases |
| 09:29 | <Hixie> | at least once microsoft caught up to the other browsers in terms of standards compliance |
| 09:30 | <krijnh> | Yeah, but IE has many bugs |
| 09:30 | <krijnh> | So there is an issue |
| 09:33 | Hixie | blogs |
| 09:36 | <roc> | othermaciej: nice, thanks! |
| 09:49 | jgraham | has only just found out that <!DOCTYPE html> and other unknown doctypes will not require the <meta> madness for the latest rendering mode |
| 09:50 | <annevk> | interesting |
| 09:50 | <annevk> | that might explain Acid2 |
| 09:50 | <Philip`> | What does "unknown" mean? |
| 09:51 | <krijnh> | If "Don't break the web" would be written as "Don't break the intranets which only rely on IE6/IE7 anyway" in the last couple of days and there was no mention of 'and we hope to see it implemented in other browsers as well', I think more people would have been in favor of the meta thing :) |
| 09:51 | <Camaban> | would that suggest that if we wanted to use HTML4.01strict, we could use the appropriate doctype in the code, then use a conditional comment to give IE8 a 'different' doctype to trigger standards mode? |
| 09:53 | <jgraham> | Philip`: Apparently "unknown" means "not widely deployed" |
| 09:53 | <annevk> | Acid3 is testing SVG now |
| 09:53 | <jgraham> | But all I've read is http://blogs.msdn.com/cwilso/archive/2008/01/22/i-feel-happy-too.aspx#7202711 |
| 09:54 | <jgraham> | (it isn't clear what will happen for IE > 8) |
| 09:55 | <jgraham> | given which I shouldn't have said "latest" originally... |
| 09:55 | <Hixie> | hm, i guess we should get major sites to use that doctype asap then |
| 09:56 | <Hixie> | (to force them to make it either trigger ie7 mode or "break the web" in ie8) |
| 09:56 | <Hixie> | (and thus discredit this nonsense some more) |
| 09:56 | <annevk> | Hixie, the second link in your article also points to HTML5 |
| 09:57 | <Hixie> | oops |
| 09:57 | <Philip`> | I suppose they could just say everything containing "DTD HTML 4.0" and "DTD XHTML 1.0" is IE7-mode, and that would cover almost all current standards-mode doctypes |
| 09:58 | <Hixie> | fixed |
| 09:58 | <Hixie> | thanks |
| 09:58 | <annevk> | haven't read it yet: http://norman.walsh.name/2008/01/22/html5 |
| 09:58 | <annevk> | (he works on XML specs and such so it might be interesting) |
| 09:59 | <Hixie> | read it earlier |
| 09:59 | <Hixie> | bed time |
| 09:59 | <Hixie> | nn |
| 10:01 | <annevk> | not so much thoughts in there yet :) |
| 10:08 | <mpt> | "Will they hard-code the URI?" - exactly what I was thinking last week :-) |
| 10:10 | <Philip`> | They wouldn't be the first browser to hard-code the Acid2 URI and do special processing on it :-) |
| 10:10 | <mpt> | Really? |
| 10:11 | <harri> | as everybody went so crazy we once did such a hack for fun in Konqueror. just showed a screenshot. |
| 10:12 | <mpt> | ah |
| 10:12 | <Philip`> | Some version of Opera made the face wink and show a special message if you left it for a couple of minutes |
| 10:12 | <Philip`> | though I can no longer find any references to that... |
| 10:28 | <Lachy> | hmm. Something weird is going on. Firefox 2.0.0.11 (Mac) is failing Acid 1 for me |
| 10:28 | <Lachy> | crap. It seems to be something wrong with my profile, since it works with a new default user profile |
| 10:32 | <annevk> | Philip`, I believe it was a mobile version that had some pretty funny gimmick |
| 10:32 | annevk | made a greasemonkey script that did special processing for the test URI |
| 10:34 | <othermaciej> | roc: if you read *very* between the lines it casts IE in a bit of a negative light |
| 10:34 | <othermaciej> | roc: but I managed to jeep it diplomatic |
| 10:35 | <roc> | yeah, it's good |
| 10:37 | <roc> | I think I did too. At least, Chris Wilson left a favourable comment on my second post :-) |
| 10:37 | <roc> | but I really wanted him to comment on the implementation issues |
| 10:37 | <roc> | oh well |
| 10:38 | <roc> | get some sleep, it's late over there |
| 10:39 | roc | realizes that this time tomorrow he'll be on CA time |
| 10:40 | <roc> | for a loose definition of tomorrow |
| 10:44 | <annevk> | hsivonen, fixed |
| 10:44 | <annevk> | it was some PHP oddity |
| 10:44 | <annevk> | header("Not Found",true,404) versus header("HTTP 1.1 404 Not Found) |
| 10:44 | <annevk> | the latter works |
| 10:45 | <annevk> | well, if you don't omit the final quote |
| 10:49 | <roc> | I feel a bit neglected that there isn't a WaSP-Mozilla task force |
| 10:50 | <annevk> | that's because IE has the most issues |
| 10:50 | <annevk> | i was told |
| 10:59 | <Lachy> | Hixie, I disagree with your recommendation on your blog for authors to continue using IE7 mode. |
| 11:00 | <Lachy> | But everyone using IE=edge is also a bad idea. I think using IE=[large-random-number] is the best solution |
| 11:02 | <Lachy> | although, as I've suggested before, IE should implement it as an opt-out that can be quickly added to existing pages that need it, instead of an opt-in for new pages |
| 11:02 | <krijnh> | That's even more hell for authors |
| 11:03 | <Lachy> | krijnh, why? |
| 11:04 | <krijnh> | Unless they ignore conditional comments and change their UA string |
| 11:05 | <krijnh> | Lachy: Because I seriously don't want to go through all the sites I've built and check them in IE8.. I'd prefer adding a little thingy to _new_ stuff I make, just as I now have to add conditional comments to cater for IE6 and IE7 |
| 11:05 | <annevk> | I think what Hixie suggested is fine |
| 11:06 | <Lachy> | krijnh, if MS did the right thing and *didn't* introduce this evil hack, more complicated changes may have to be made. But, if MS did things right, then you shouldn't need to make any significant changes |
| 11:07 | <krijnh> | Yep |
| 11:07 | <krijnh> | But they didn't |
| 11:07 | <Lachy> | krijnh, so are you saying you support MS's decision for the opt-in? |
| 11:08 | <krijnh> | If only they implement it, I think I can live with it |
| 11:09 | <krijnh> | Err, I think I have to live with it :) |
| 11:09 | <Lachy> | I just hope they make the HTML5 DOCTYPE equivalent to IE=edge, while still allowing the meta as an opt-out if needed |
| 11:10 | <krijnh> | The opt-out ? |
| 11:10 | <krijnh> | Perhaps they need both :) |
| 11:10 | <krijnh> | Ow, wait, they already do, of course ;\ |
| 11:10 | <annevk> | <!doctype html> triggers "edge" |
| 11:10 | <krijnh> | Yeah |
| 11:10 | <krijnh> | And the meta thing with IE=7 untriggers it, right? |
| 11:10 | <krijnh> | Cool stuff :) |
| 11:11 | <Lachy> | yeah, so if you have an HTML5 page that works in IE10 (after IE adds some support for it). Then IE11 comes out and something breaks. You can quickly add the opt-out to your HTML5 page to trigger IE10 mode until a more permanent fix can be deployed |
| 11:11 | <krijnh> | Yeah, it's just that nobody does that |
| 11:11 | <roc> | annevk: do we know that for sure? |
| 11:11 | <krijnh> | There is no 'quickly do something to site x and be done with it' |
| 11:12 | <annevk> | roc, cwilso said so in comments on his blog |
| 11:12 | <annevk> | I'm not sure if it's a good or bad thing though |
| 11:12 | <roc> | it sounded a bit off the cuff |
| 11:12 | <annevk> | it does seem a little bit weird given that normal standards mode is not whitelist based |
| 11:13 | <annevk> | as in, the reason that <!doctype html> triggers standards mode in Presto, Gecko, and WebKit is that it was not part of a blacklist |
| 11:14 | <Lachy> | krijnh, deploying a simple opt-out to fix a bug due to a browser upgrade is faster than working out what the bug is, finding a solution that works in both new and old versions and deploying that. I didn't "ad be done with it" - the opt-out would be a quick and dirty hack that should only be used as a temporary measure |
| 11:15 | <krijnh> | Lachy: but that's possible right? |
| 11:16 | <Lachy> | why wouldn't it be possible? |
| 11:17 | <krijnh> | I thought you meant you wanted to them to only make it possible to opt-out |
| 11:17 | <krijnh> | -to |
| 11:18 | <annevk> | i still don't think they should do this at all |
| 11:19 | <annevk> | especially the infinite amount of quirks modes scares me, but given the current landscape only intranet content might get crippled and I can live with that |
| 11:20 | <krijnh> | As an author, I think I like it |
| 11:21 | <krijnh> | I'm probably the only one here not working for some UA vendor :) |
| 11:21 | <Camaban> | I don't and I don't like it |
| 11:21 | <Camaban> | it's an IE hack for people who use IE hacks |
| 11:21 | <roc> | I like it because I think it'll make IE development implode :-) |
| 11:21 | <krijnh> | Camaban: you don't use conditional comments now? |
| 11:21 | <roc> | (not really) |
| 11:22 | <Camaban> | krijnh: very very very rarely |
| 11:22 | <krijnh> | Yeah, but you use them |
| 11:22 | <Camaban> | maybe a couple of times |
| 11:23 | <krijnh> | If it triggers on the HTML5 doctype, I don't see a problem |
| 11:23 | <Camaban> | I didn't have any problems when IE7 came out, I wasn;t one of the stupid satndards based coders spewing futureproof commentary and then using IE hacks which broke with a new version |
| 11:24 | <annevk> | how they did IE7 was very stupid though, btw |
| 11:24 | <Camaban> | I don't expect to have problems when IE8 comes out, and so I just want my web sites to use the browsers latest standards mode |
| 11:24 | <annevk> | i blogged about that long ago |
| 11:24 | <annevk> | they fixed the hacks but not the underlying issues |
| 11:24 | <annevk> | no wonder stuff breaks |
| 11:25 | <krijnh> | I just want my websites to all be run by people using MOS, but they just don't |
| 11:25 | <krijnh> | Damn users :) |
| 11:26 | <Camaban> | from the moaning I saw, most of it was related to * html, and people using it to hack anything and everything for IE |
| 11:28 | <annevk> | * html, a > b, a + b, etc. |
| 11:28 | <krijnh> | Wrongly written conditional comments were probably messier |
| 11:29 | <Camaban> | I wouldn't place all the blame for that MS, people buying into the standards based coding idea, and then using lots of hacks like that are missing the point |
| 11:29 | <Camaban> | though MS take the original blame for letting things get to that state in the first place |
| 11:29 | <annevk> | MS could have stopped supporting conditional comments, those hacks, etc. |
| 11:30 | <krijnh> | You think they've tried that? |
| 11:30 | <Camaban> | what do you mean by stop supporting? |
| 11:31 | <krijnh> | Just see them as comments, like other browsers do |
| 11:31 | <krijnh> | They must have tried that with IE8, to see what happens |
| 11:31 | <krijnh> | They must have tried removing the UA string or something |
| 11:32 | <annevk> | i don't know |
| 11:32 | <krijnh> | For every thing we come up with, they prolly have 10 outsourced programmers testing it out |
| 11:32 | <krijnh> | And this was probably the best thing they could handle |
| 11:33 | <krijnh> | And now they've wrapped in something 'other UAs could implement as well' |
| 11:33 | <krijnh> | But if that doesn't happen, this probably the best thing for everybody |
| 11:35 | <krijnh> | Do other vendors forsee problems in backwards engineering this IE behavior? |
| 11:36 | <roc> | we don't try hard enough for IE compatibility to care |
| 11:36 | <krijnh> | Well, that's the biggest argument I see people make in comments |
| 11:38 | <krijnh> | annevk: that's the 'infinite amount of quirks modes' (and others have to implement as well) you're talking about, right? |
| 11:39 | <annevk> | the moment we have to start caring is when IE is in a dominant position again |
| 11:40 | <krijnh> | Isn't it still? |
| 11:40 | <annevk> | it's not dominant enough for people to stop care about the other browsers |
| 11:40 | <annevk> | (although sometimes it is, so it may still cause issues) |
| 11:41 | <Camaban> | even then, getting people to care about browsers other than IE and FF can be tricky |
| 11:43 | <roc> | I don't really worry about the scenario where people only care about IE |
| 11:43 | <roc> | we're far away from that in most of the public Web, and getting further away |
| 11:43 | <roc> | and if it happens, the Web is screwed in so many ways |
| 11:44 | <krijnh> | So what's the problem again for authors, users and other vendors? |
| 11:44 | <roc> | I haven't haven't the foggiest idea with IE's mobile strategy is |
| 11:45 | <roc> | this approach to compatibility can only hurt them there |
| 11:45 | <roc> | I don't see them roaring back to unquestioned dominance |
| 11:46 | <roc> | krijnh: it's going to be confusing for authors. Down the road they'll have to know the bugs of several different versions of IE |
| 11:46 | <roc> | if they want to work on someone's existing content |
| 11:47 | <krijnh> | I can only imagine people updating their crappy IE6 quicker now.. Old stuff and intranets still work, and we authors can use shiny new features |
| 11:47 | <Camaban> | nad if several of us start advocating use of 'edge', it'll jsut be an issue that raises it's head everytime they release another version anyway, so doesn't resolve the issue at all |
| 11:48 | <krijnh> | Camaban: not if they fix and release often now |
| 11:48 | <roc> | it reduces the incentive to modernize dusty intranets for standards-based, cross-browser access, that's bad for vendors, users and maybe authors |
| 11:48 | <roc> | krijnh: well no, IE8 doesn't include an IE6 mode AFAIK |
| 11:48 | <Camaban> | krijnh: quite a big IF, an not something they've ever done previously |
| 11:49 | <krijnh> | roc: ow, damn |
| 11:49 | <roc> | AFAIK it's IE7 + a new IE8 mode |
| 11:50 | <roc> | so, IE6 quirks mode, IE7 standards mode, IE8 really-standards mode |
| 11:50 | <krijnh> | Camaban: I think they get it at least a little bit now |
| 11:50 | <Camaban> | and say people do start using IE=8 a lot, when they come to do 9, they could say "oh, there's no demand for various updates, people are using IE8 rendering anyway" |
| 11:51 | <krijnh> | roc: they've put a lot of effort into making sure intranets still work in IE7, no? I think the only issue for not updating a company to IE7 is the OS.. |
| 11:51 | <roc> | it sounds like new features only get added to the very latest mode, so that might inhibit people from adding features to existing content |
| 11:52 | <roc> | krijnh: they did, but they talk a lot about how much pain IE7 caused their customers who had content in IE6 standards mode |
| 11:52 | <roc> | which didn't work in IE7 |
| 11:52 | <roc> | companies are afraid of that |
| 11:53 | <Camaban> | the fact IE6 still has as much market share as it does says something |
| 11:53 | <krijnh> | Perhaps they slap an IE6 standards mode in IE8 as well, and just tell those companies to use IE=6 in their intranet apps, dunno :) |
| 11:53 | <Camaban> | difficult to make a guess as to how much of that is corporates not wanting to upgrade, but I bets there's a good number of them |
| 11:53 | <krijnh> | I'm pretty sure |
| 11:53 | <krijnh> | I've just built a voting site which is mostly used in corporate environments |
| 11:54 | <krijnh> | 38% IE6, 19% IE7, 9% Fx2 |
| 11:54 | <krijnh> | Ow, and 9% IE3 - my statistics suck :p |
| 11:56 | <roc> | I've heard FF2 usage is up around 10% in many corporate samples, which is surprisingly high to me |
| 11:56 | <roc> | since we basically don't target that market at all |
| 11:57 | <Camaban> | roc: any idea of OS' there? could be some of that is due to corporates wiht mac's maybe? |
| 11:58 | <roc> | dunno, but I thought Mac share in corporations was very low |
| 11:58 | <Camaban> | probably |
| 11:59 | <Camaban> | I sued to work for a reasonably big magazine publisher, they ahd a fair few macs around, and FF usage was fairly high, they were a bit split between officially supporting IE and FF, but as IE/Mac was dying out quickly, they could hardly continue to enforce IE usage everywhere |
| 12:06 | <gsnedders> | Philip`: re: invalid UTF-8 for trackbacks: remember trackbacks have no defined character set, so it's sniffed, so it will probably be treated as either ISO-8859-1 or Windows-1851 (forgive me if that number is wrong) |
| 12:08 | <gsnedders> | Philip`: URLs are of infinite length, but IE6 at least had a limit (IIRC IE7 doesn't) |
| 12:10 | <roc> | technically you mean URLs are of unbounded length |
| 12:11 | <roc> | sending an infinite-length URL to the server takes too long :-) |
| 12:12 | <annevk> | :p |
| 12:13 | <annevk> | we need infinite time to go with infinite URIs |
| 12:13 | <gsnedders> | well, there are infinitely long HTML documents on the real world web :P |
| 12:22 | <Dashiva> | But do those documents support offsets when fetching? Otherwise you're stuck when the server goes down after a few years |
| 12:31 | <zcorpan> | oh, completely forgot about <ol start=6 reverse> |
| 12:31 | <zcorpan> | oh well |
| 13:24 | <Philip`> | gsnedders: Re: re: invalid UTF-8 for trackbacks: If it's close-enough-to-valid UTF-8 that's just forbidden by XML, then hopefully it wouldn't get sniffed as something else |
| 13:25 | <Philip`> | Maybe just using � would be enough to break things that do non-strict XML parsing |
| 13:55 | <krijnh> | Can't IE8 just make a guess with the Last-Modified HTTP headers each file has? :] |
| 13:57 | <krijnh> | Everything before X should go in IE7 compatmode/quirksmode, everything after X is in truly standards mode, unless the developer has opted out with a shiny meta |
| 13:57 | <Philip`> | Half of the pages on the web don't have a Last-Modified header at all |
| 13:57 | <krijnh> | Is that the half with no doctype either? |
| 13:58 | <Philip`> | I would assume server configuration is independent of served HTML content |
| 13:58 | <Philip`> | by which I mean, I have no idea whatsoever |
| 13:59 | <krijnh> | Well, other content then |
| 13:59 | <Philip`> | (http://http-parsing.gsnedders.com/ ) |
| 13:59 | <krijnh> | Static files mostly have a last-modified, no? |
| 14:00 | <Philip`> | No idea :-) |
| 14:00 | <krijnh> | Well, PHP generated stuff for example doesn't have a last-modified header by default |
| 14:01 | <Philip`> | That would make sense |
| 14:02 | <krijnh> | But HTML can be ignored |
| 14:02 | <Philip`> | but I have no idea how to determine whether most static files have last-modified |
| 14:02 | <Philip`> | except by ugly things like guessing from the file extension |
| 14:02 | <krijnh> | Uhm |
| 14:03 | <krijnh> | http://http-parsing.gsnedders.com/ - fetch everything where content-type is text/css and has a last-modified |
| 14:04 | <krijnh> | That would be hard with embedded stylesheets though |
| 14:05 | <Philip`> | Not entirely sure what you mean... Follow all the <link rel=stylesheet href=...> links, assume the returned files are static, and then count last-modified? |
| 14:05 | <krijnh> | Yeah |
| 14:05 | <Philip`> | (Could do the same with <img src=...> since most images will be static) |
| 14:05 | <krijnh> | Yeah, but images don't change that much |
| 14:06 | <krijnh> | It's about the development assets |
| 14:07 | <Philip`> | So IE8 should download and parse the page, extract all stylesheet links, download all stylesheets, check their last-modified headers, and if they're past a certain date then switch to the fewer-bugs mode? That doesn't sound totally efficient :-) |
| 14:07 | <krijnh> | If the last change in the newest stylesheet or js file was in 2007 it's safe to assume IE7 compat, or something |
| 14:07 | <krijnh> | Yeah ;p |
| 14:08 | <krijnh> | Glad I'm no browser engineer :D |
| 14:08 | <Philip`> | I'd hate to be the web developer who makes a trivial change to one file, then discovers their whole site has collapsed in a pile of bugs, and reverts their change but still suffers from the same pile of bugs and has no idea what they did wrong :-) |
| 14:09 | <krijnh> | Well then |
| 14:09 | <krijnh> | The simple answer to my first question would be 'No, you idiot' :) |
| 14:10 | <Camaban> | krijnh: No, you idiot |
| 14:10 | <Camaban> | :) |
| 14:11 | <Philip`> | Okay, glad it's resolved so easily ;-) |
| 14:11 | <krijnh> | :P |
| 14:12 | <krijnh> | Meanwhile, in a channel not so far away, people are on mute because his earplug isn't working and his synthesizer tends to bother others |
| 14:21 | <krijnh> | http://krijnhoetmer.nl/stuff/html5/section-elements/ - yay, now it almost works cross browser.. |
| 14:25 | <MikeSmith> | Hixie - for being able to directly reference the list in of doctypes that must trigger quirks mode in HTML5 UAs, it would be useful to have an ID at the beginning of the list |
| 14:26 | <MikeSmith> | maybe at the sentence preceding the list |
| 14:26 | <MikeSmith> | "Then, if the DOCTYPE token matches one of the conditions in the following list, then set the document to quirks mode:" |
| 14:27 | <Philip`> | krijnh: Except for Firefox :-( |
| 14:27 | <krijnh> | Philip`: Yeah |
| 14:28 | <krijnh> | But IE6, IE7, Opera 9 and Safari 3 handle it now |
| 14:28 | <krijnh> | How is Fx3 ? |
| 14:28 | <Philip`> | Still as broken as FF2, as far as I'm aware |
| 14:28 | <krijnh> | Hmm |
| 14:29 | <krijnh> | Hmm |
| 14:29 | <krijnh> | That's 2 times hmm |
| 14:29 | <Philip`> | https://bugzilla.mozilla.org/show_bug.cgi?id=311366 |
| 14:29 | <krijnh> | Would document.createElement('abbr') enable support for <abbr> in IE6? :o |
| 14:30 | <Philip`> | It does |
| 14:30 | <Philip`> | <abbr title> tooltips work too |
| 14:30 | <krijnh> | Yeah, cool |
| 14:31 | <krijnh> | http://krijnhoetmer.nl/zooi/abbr.html |
| 14:31 | <krijnh> | Tee hee |
| 14:34 | <krijnh> | Why did no IE developer ever mention this, when people complained about IE6 not supporting <abbr> ? :| Were all their developers killed after releasing IE6 or something? |
| 14:35 | <Philip`> | krijnh: Maybe no IE developer understands the IE engine well enough to know about such things? :-) |
| 14:35 | <Philip`> | It doesn't seem like an intentional feature, so the thought might have just not occurred to them |
| 14:36 | krijnh | tries document.createElement('canvas') to explore other hidden IE6 features :) |
| 14:36 | <Camaban> | maybe no IE developer understood <abbr> so didnt know what it was meant to do :) |
| 14:37 | <Philip`> | Maybe they've got an entire <canvas> implementation in there and we just have to add a magical script line to enable it! |
| 14:37 | <jwalden> | or a CSS property |
| 14:37 | <Camaban> | document.createElement(godmode') |
| 14:37 | <krijnh> | DNKROZ :P |
| 14:37 | <Philip`> | krijnh: Excanvas does some searching for CANVAS and /CANVAS tags to reconstruct canvases, and that'd be much nicer if they used this method instead |
| 14:38 | <krijnh> | The .createElement() should be called before using the element though |
| 14:39 | <Philip`> | That's not really a problem since everyone puts the <script> in the <head> anyway |
| 14:40 | <krijnh> | Didn't Yahoo teach everybody to put the scripts at the bottom ? |
| 14:41 | <Philip`> | We can just unteach them |
| 14:41 | <krijnh> | Unteach me while you're at it :) |
| 14:41 | <Philip`> | And you should put the IE-parser-fixing scripts in a <![whatever it is, so that other browsers don't have the cost of downloading an extra script |
| 14:42 | <Camaban> | krijnh: then people discovered that if using js based analytics software (like google analytics), having the script at the bottom of the page meant you completely missed some page views, because the user clicked a link before the bottom of the page was loaded and the script was run :) |
| 14:42 | <krijnh> | Statistics are like bikinis anyway |
| 14:43 | <krijnh> | They smell, when used |
| 14:44 | <Camaban> | well, if you're stats aren't right because you included the srcipt at the bottom of the page, quite possibly :) |
| 14:44 | <krijnh> | And when you're not using stats scripts? |
| 14:45 | <Camaban> | well, if you're using web server logs, then you've a whole other raft of issues :) |
| 14:45 | <Camaban> | point being, advice these days is to put js based stats scripts at the top of the code |
| 14:45 | <krijnh> | Yeah, you get 9% IE3 visitors :] |
| 14:46 | <krijnh> | Which means you didn't update your software, even though Fx3 and Safari 3 are out |
| 14:46 | <Camaban> | or means something is faking it's UA :) |
| 14:46 | <krijnh> | Then why did the smart people at Yahoo put a silly warning in YSlow? |
| 14:47 | <Camaban> | probably because laoding a script early is likely to slow down the rest of the page |
| 14:47 | <Camaban> | especially if it's external |
| 14:47 | <Camaban> | swings and roundabouts |
| 14:48 | <Camaban> | people serious about stat sofwtare tend to accep the slow donw a bit |
| 14:48 | <krijnh> | So stats scripts in the head, other stuff at the bottom? |
| 14:48 | <Philip`> | I just write all my JavaScript code inline |
| 14:48 | <Camaban> | probably depends on the script, but potential, yes |
| 14:49 | <Camaban> | have had to put ad code at the bottom of code before because some ad servers were slow responding and casuing long delays in page load |
| 14:49 | <Camaban> | but might be some other cases like stats where getting them fired earlier is desirable |
| 14:50 | <krijnh> | http://krijnhoetmer.nl/irc-logs/xhtml/20080123#l-209 :) |
| 14:51 | <krijnh> | daft - adjective, -er, -est. 1. senseless, stupid, or foolish. 2. insane; crazy. 3. Scot. merry; playful; frolicsome. |
| 14:52 | <Philip`> | HTML5 Editor's Daft |
| 14:52 | <krijnh> | :P |
| 14:56 | <krijnh> | If only we could see the IE7 codebase.. I wonder how they implemented <abbr> |
| 14:59 | <jwalden> | I don't want to know |
| 15:00 | <Philip`> | What does IE7 do about <abbr> in quirks? |
| 15:00 | <krijnh> | Nothing, just supported |
| 15:01 | <krijnh> | Same in IE6 in quirks mode |
| 15:05 | <Philip`> | Hmm, I thought people said IE7 had a frozen copy of an older engine for quirks mode; but IE7 supports quirks <abbr> like standards <abbr>, and unlike IE6 which never supports <abbr> |
| 15:15 | <zcorpan> | Philip`: indeed, ie7's quirks mode is not the same as ie6's quirks mode |
| 15:15 | <zcorpan> | even though they said they wouldn't change anything in quirks mode |
| 15:21 | <Philip`> | Did they intentionally add <abbr> to IE7 quirks mode, or is it just shared code? |
| 15:33 | <zcorpan> | no idea. perhaps they figured that supporting <abbr> in quirks wouldn't break anything. |
| 15:37 | <krijnh> | "I for one am excited about the IE-version Meta tag, as long as I can set it to IE1. That way all of my pages can look good in Safari, Firefox and Opera while looking like badly munged Geocities pages circa 1995. Hoorayl!" |
| 16:06 | jwalden | wonders why freenode doesn't display idle time in /whois |
| 16:09 | <gavin_> | jwalden: it will if you do /whois nick nick rather than /whois nick |
| 16:10 | <jwalden> | ah, of course! how natural! |
| 16:10 | jwalden | wonders why this behavior differs across servers |
| 16:17 | <gsnedders> | jwalden: Because nothing follows the spec, because the spec is totally incompatible with the real world |
| 16:17 | <jwalden> | I'd have expected nominal real-world compatibility |
| 16:17 | <jwalden> | at least on common things like /nick |
| 16:17 | <jwalden> | er |
| 16:18 | <jwalden> | /whois |
| 16:18 | <gsnedders> | oh, hell no. :) |
| 16:18 | <annevk> | "you must be new here" |
| 16:20 | <gsnedders> | "Please leave your sense of logic at the door, thanks!" |
| 16:21 | <gsnedders> | on an unrelated note, can we have a spec for feed autodiscovery that actually is usable? |
| 16:21 | <Ketsuban> | /whois is remarkably uncommon. Most IRCites use IRC for two things only - speaking and actions. :P |
| 16:27 | <SadEagle> | Philip`: are non-JS comments in e.g. 2d.imageData.put.round intentional? |
| 16:29 | <Philip`> | SadEagle: Urgh, no |
| 16:29 | <Philip`> | That's what happens when I mix JS and Python and YAML in a single source file... |
| 16:30 | <SadEagle> | BTW, for some reason your script doesn't work with 'real' yaml implementation.. |
| 16:30 | Philip` | unindents the comments so they become YAML instead of JS |
| 16:30 | <Philip`> | Hmm, what's a "'real' yaml implementation"? |
| 16:32 | <SadEagle> | as in not 'syck' or whatever is supposed to be faster. whines about double arcto annotation or something(?) |
| 16:32 | <Philip`> | SadEagle: Fixed and uploaded (for .round and .clamp) - thanks |
| 16:32 | <SadEagle> | thanks for the testcases |
| 16:33 | <SadEagle> | are you planning on security testcases for getImageData? |
| 16:33 | <Philip`> | Oh, that's probably because I've got two &arcto |
| 16:34 | <SadEagle> | heh, I fail .round pretty miserably :-) |
| 16:34 | <Philip`> | Fixed local copy of YAML file |
| 16:35 | <SadEagle> | thanks. |
| 16:35 | <Philip`> | I should probably just duplicate all the toDataURL cases for getImageData, since I can't think of any differences |
| 16:36 | <Philip`> | (If you want the YAML thing to work, just change the second "&arcto ..." line to say "*arcto" instead) |
| 16:37 | <Hixie> | Lachy: using IE=large-number or IE=edge are both trivial for microsoft to route around |
| 16:37 | <SadEagle> | BTW, do i understand correctly that the origin of image data in a canvas element should be transitive? |
| 16:38 | <annevk> | Philip`, in http://philip.html5.org/data/doctypes.txt what's the difference between #2, #4, and #5? |
| 16:38 | <Philip`> | Not sure what you mean by transitive here |
| 16:39 | <Philip`> | annevk: Whitespace (which was all collapsed to spaces before outputting to the .txt file) |
| 16:40 | <Lachy> | Hixie, but MS have to realise that if someone has gone to all the trouble of using IE=edge or large-number then they should accept that the author has taken responsibility for themselves and not screw them over |
| 16:40 | <annevk> | thx |
| 16:40 | <jwalden> | Hixie: if I have pages A and B on the same domain, and A calls a function in B which calls someWindow.postMessage, the uri of the event is B, correct? |
| 16:40 | <Hixie> | Philip`: (the excanvas guys know, i told them yesterday, and they immediately were like "wow, we could totally fix the <canvas>-finding code...") |
| 16:41 | <Hixie> | Lachy: you mean like how they should accept that the author has taken responsibility for themselves and not screw them over when they pick standards mode using a doctype? |
| 16:41 | <Hixie> | jwalden: iirc |
| 16:41 | <Philip`> | annevk: Added a comment to the file to hopefully make it clearer |
| 16:41 | <SadEagle> | Philip`: if you have a canvas A, that had images from domains d1 and d2 painted, and you painted from it on canvas B, d1 and d2 should be included on its list of domains, right? |
| 16:42 | <Hixie> | Lachy: this isn't about them accepting anything, it's about them thinking they know better about what users want as far as compat goes |
| 16:42 | <jwalden> | cool; the exact phrasing was "the document that the script that invoked the methods is associated with", for what it's worth |
| 16:42 | <Hixie> | yeah, i should make that clearer |
| 16:42 | <Philip`> | SadEagle: In at least Firefox, there's just a boolean "unsafe" flag per canvas, which gets set whenever you paint from a different-origin image or from an unsafe canvas |
| 16:43 | <Philip`> | SadEagle: which seems sensible, and I can't think of cases where it's unsafe or overly restrictive, though I know very little about security |
| 16:43 | <gsnedders> | I think part of the issue is the umber of sites using <!--[if IE]> without any version |
| 16:43 | <gsnedders> | *number |
| 16:43 | <SadEagle> | Philip`: yeah, I just realized that 2 sources can never be equal to 1 :-) |
| 16:44 | <annevk> | SadEagle, no need for domain lists, just a "safe" flag |
| 16:44 | <Philip`> | http://philip.html5.org/data/ccs.html |
| 16:44 | <Philip`> | <!--[if IE]> is quite common :-( |
| 16:44 | SadEagle | won't worry too much about websites not working in IE |
| 16:44 | <Philip`> | (Well, 1.5%) |
| 16:44 | <annevk> | you need a "safe" flag on CanvasPattern, HTMLCanvasElement, and maybe HTMLImageElement depending on how you implement things |
| 16:45 | <jwalden> | Hixie: second: what should be the domain/uri when postMessage is called from within a window in which about:blank was initially loaded but which has been document.open|write|closed with different contents? null/"about:blank" is my guess per <http://www.whatwg.org/specs/web-apps/current-work/multipage/section-documents.html#domain>; |
| 16:45 | <Philip`> | (or 1.5% of not-very-complex web pages - these aren't fancy application sites) |
| 16:46 | <Hixie> | jwalden: doesn't document.open() propagate a domain/origin into the about:blank apage? |
| 16:46 | <SadEagle> | annevk: thanks for reminding me of patterns. |
| 16:46 | jwalden | checks |
| 16:46 | <jwalden> | that's the behavior of Moz right now, but I wasn't sure if it was right or not |
| 16:46 | Philip` | wonders if he had a test for drawing data: onto canvas being safe |
| 16:47 | <annevk> | i believe so |
| 16:47 | <annevk> | and apart from Opera most browsers fail iirc :( |
| 16:47 | <Philip`> | Oh, I do |
| 16:47 | <Philip`> | Opera 9.2 fails, Opera 9.5 passes, Firefox fails, Safari doesn't have toDataURL |
| 16:48 | <Hixie> | jwalden: sorry, i'm confused. what is? about:blank, or propagating the domain? |
| 16:48 | <annevk> | Hixie also needs to update the spec to mention that <img src=data:...> is always "safe" |
| 16:48 | <annevk> | for <canvas> purposes |
| 16:49 | <Hixie> | send mail, everyone :-) |
| 16:49 | <annevk> | and that <img src=external.foo -> data:...> is not |
| 16:49 | <jwalden> | Hixie: sorry, Moz currently has the URI/domain of the writer (or possibly opener, I've not tested) |
| 16:49 | <Hixie> | jwalden: that seems reasonable to me. i expect html5 to eventually require that once i have clarified the 'origin' of documents. |
| 16:50 | <jwalden> | Hixie: okay; should I send an email to the list to make sure that gets put in the open() description, as it's not there now? |
| 16:50 | <jwalden> | (for whichever of document or window is causing this behavior, that is) |
| 16:50 | <Philip`> | annevk: I think the spec already requires that - data: URIs have the same origin as the document they're used in, or something like that |
| 16:50 | <Philip`> | (Er, things created from data: URIs have the same ...) |
| 16:51 | <SadEagle> | annevk: the origin of an image element is always the origin of its document, right? |
| 16:51 | <Philip`> | SadEagle: No, it depends on the src |
| 16:51 | <annevk> | SadEagle, <img src="http://external.example.org/img.gif">; in an foobar.invalid document would be an "unsafe" image |
| 16:52 | <annevk> | SadEagle, as it could contain information from an intranet or so |
| 16:52 | <annevk> | (the attack vector is that until <canvas> there was no way to read image data and therefore cross-site images were not an issue) |
| 16:53 | <Hixie> | jwalden: yes please |
| 16:54 | <SadEagle> | annevk: hmm, if you paint it on an external.example.org--owned canvas, it's safe though? |
| 16:54 | <Philip`> | How do you get a reference to a differently-owned canvas? |
| 16:55 | <SadEagle> | one can restrict domains further via JS.. |
| 16:56 | <SadEagle> | hmm, or not. I wish I was the guy who wrote the XSS-checks infrastructure :( |
| 16:58 | <SadEagle> | patterns aren't tied toa particular context, are they? |
| 16:58 | <Philip`> | They shouldn't be |
| 16:59 | <Philip`> | http://philip.html5.org/tests/canvas/suite/tests/2d.pattern.crosscanvas.html |
| 16:59 | jwalden | is rather glad he wasn't the person who wrote the XSS-checks infrastructure |
| 16:59 | <jwalden> | I can armchair-critique what resulted with the best of them, however! |
| 17:00 | <Philip`> | If you change document.domain with JS, I've got entirely no idea what should happen :-) |
| 17:00 | SadEagle | tries to think whether an unsafe bit is really enough for them, then. |
| 17:01 | <SadEagle> | (I think this calls for pen-and-paper... and an auditor) |
| 17:01 | <Philip`> | A list of domains might not be enough, since there could be some other non-image-drawing functions that make the canvas unsafe |
| 17:02 | <Philip`> | like a drawWindow etc |
| 17:02 | <SadEagle> | well, a pattern created from an unsafe canvas is always unsafe.. but what about images? I guess only if the images origin isn't the same as the domain of the document, or? |
| 17:03 | SadEagle | notes the use of the word 'guess', and gets scared |
| 17:03 | <Philip`> | Comparing image origin vs document domain is what people seem to be doing |
| 17:05 | <annevk> | and that makes sense |
| 17:05 | <annevk> | document.domain doesn't really matter here afaict |
| 17:05 | <SadEagle> | since normal checks should prevent that? |
| 17:07 | <annevk> | well, if you have access to some other place because of document.domain you can extract the image data anyhow |
| 17:07 | <annevk> | so the only problem might be that if you compare the origin of the <canvas> on which you draw against the <img> origin which you draw upon it you might be too strict |
| 17:22 | <annevk> | http://www.b-list.org/weblog/2008/jan/23/legacy/ |
| 17:27 | <SadEagle> | well, if they set the default the other way around, they'd be doing everyone else a huge favor, and what's their motivation for doing that? |
| 17:27 | <annevk> | i'm not sure they're doing themselves a favor right now |
| 17:30 | <Camaban> | piss off IE only developers, or piss off everyone else... |
| 17:30 | <SadEagle> | annevk: so to double-check, the only origin of unsafety would be an image element with a mismatch of origin? |
| 17:32 | <annevk> | well, and <canvas> and CanvasPattern objects which have such images painted on them |
| 17:32 | <SadEagle> | thanks. I view that as transitivity and not origin :-) |
| 17:32 | <annevk> | <canvas>.drawImage(<canvas>.drawImage(unsafeimage)) -> both are unsafe |
| 17:32 | <annevk> | ok |
| 17:33 | <annevk> | hmm, so UTF-7 may not be needed after all |
| 17:33 | <annevk> | Philip`, got stats on usage? :) |
| 17:34 | <SadEagle> | And I am curious whether utf-7 or ISO 8859-5 are used less |
| 17:36 | <Philip`> | annevk: I don't think I do, and I wouldn't have a large enough sample to get anything meaningful for things like utf-7 |
| 17:37 | <annevk> | k |
| 17:37 | <Hixie> | annevk: re our earlier conversation |
| 17:37 | <Hixie> | according to my data, as of july last year, the most common doctypes were: |
| 17:37 | <Hixie> | 1. no doctype |
| 17:38 | <Hixie> | 2. html,-//W3C//DTD XHTML 1.0 Transitional//EN,http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd |
| 17:38 | <Hixie> | 3. HTML,-//W3C//DTD HTML 4.01 Transitional//EN,(no uri) |
| 17:38 | <Hixie> | ... |
| 17:38 | <Hixie> | 6. html,-//W3C//DTD XHTML 1.0 Strict//EN,http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd |
| 17:39 | <Hixie> | 7. html5 doctype (no fpi or uri) |
| 17:39 | <Hixie> | 8. HTML,-//W3C//DTD HTML 4.01//EN,http://www.w3.org/TR/html4/strict.dtd |
| 17:39 | <Hixie> | 13. html,-//W3C//DTD XHTML 1.1//EN,http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd |
| 17:39 | <Hixie> | (skipping transationals) |
| 17:39 | <annevk> | 7 html5?! |
| 17:40 | <Hixie> | 23. HTML,-//W3C//DTD HTML 4.0//EN,(no uri) |
| 17:40 | <Hixie> | annevk: they might also have been parse errors, it's unclear |
| 17:40 | <annevk> | k |
| 17:41 | <Hixie> | HTML,-//W3C//DTD HTML 4.01//EN,http://www.w3.org/TR/html4/strict.dtd seems to be the most common non-transitional, non-xhtml doctype |
| 17:41 | <Hixie> | i guess i should use that for acid3 |
| 17:41 | <annevk> | i'd go for 8 then |
| 17:41 | <annevk> | yeah |
| 17:42 | <Hixie> | i'm surprised about #7 |
| 17:42 | <Hixie> | how weird |
| 17:42 | <Hixie> | i should investigate further |
| 17:42 | <Hixie> | but now i have to go to a meeting. bbl. |
| 17:43 | <blooberry> | hixie: html5 doctype was pretty far down on my list |
| 17:45 | <Philip`> | Hixie: Was that using the HTML5 tokenisation algorithm to extract the doctype? |
| 17:45 | <Hixie> | yeah |
| 17:45 | <Hixie> | maybe those were all bogus doctypes |
| 17:45 | <Hixie> | afk |
| 17:46 | <Philip`> | http://philip.html5.org/data/doctypes.txt has zero <!doctype html>s |
| 17:46 | <annevk> | hsivonen, http://www.b-list.org/weblog/2008/jan/23/legacy/#c53749 shows that your tables are slightly confusing |
| 17:47 | <Philip`> | Hixie: Also, were you ignoring case when counting commonness? |
| 17:50 | <Philip`> | Argh, I forgot case-sensitivity in my quirks/standards detector... |
| 17:51 | <blooberry> | philip` my list has it about 150 times (with varying case results) |
| 17:52 | <blooberry> | philip`: Do you keep an active linkage in your research to the results, eg, if you wanted to find a particular URL in that result set, can you? |
| 17:52 | <Philip`> | blooberry: 150 out of how many? |
| 17:52 | <blooberry> | philip`: a lot. ;-} 1681088 domains |
| 17:54 | <blooberry> | (that number didn't count any results with no doctypes in the set. Total domains in set is ~3.2 million) |
| 17:54 | <Philip`> | blooberry: For e.g. doctypes, I generate a list of (uri, matched string)s, and then have a separate program to summarise the data, so I can go back and find the URIs from there |
| 17:54 | <blooberry> | very cool |
| 17:54 | <Philip`> | but also I cache all the pages I downloaded, so I can re-run the whole data-extraction program without any network activity |
| 17:54 | <Philip`> | (so it only takes about thirty seconds for 16K pages) |
| 17:55 | <blooberry> | also useful...what is the disk hit on that? (how much space does it cost) |
| 17:55 | <blooberry> | whoa. Very nice. 8-} |
| 17:55 | <Philip`> | 400MB for the 16K HTML pages (plus HTTP headers, plus serialised Java exceptions) |
| 17:55 | <Philip`> | It's only 14MB for the extracted doctype data |
| 17:56 | <blooberry> | any external object references analyzed or saved? |
| 17:58 | <blooberry> | philip`: oh...I've seen you reference statistics now you have generated from I think at least 3 domains. Do you have any central place where you summarize or link to all the far-flung pieces? |
| 17:58 | <Philip`> | I've checked external references in the past, for e.g. http://canvex.lazyilluminati.com/misc/stats/scripttypes2.html |
| 17:59 | <blooberry> | philip`: hmm. I think that makes it 4 domains now. ;-} |
| 17:59 | <Philip`> | I don't have any central place at the moment, and I keep changing my mind about where I want to bother saving stuff :-) |
| 17:59 | <blooberry> | 8-D |
| 17:59 | <Philip`> | I've been meaning to make a list of all the stuff I've done, so I could do that soon if you're interested in finding the various bits |
| 18:00 | <blooberry> | do you have some sort of list of all the things you have covered? Or is it all in your head? ah, ok. |
| 18:00 | <Philip`> | It's not in my head - I just have to search through all the servers to find anything |
| 18:01 | <Philip`> | Some of it is only exists in mailing list posts, too... |
| 18:01 | <Philip`> | s/ is// |
| 18:05 | <annevk> | http://www.computable.nl/nieuws.jsp?id=2319711 talks about HTML5 making an initial step to the Semantic Web :) |
| 18:05 | <Philip`> | http://philip.html5.org/data/doctypes.txt - now correcter in its quirks/standards determination |
| 18:06 | <Philip`> | and also http://philip.html5.org/data/doctypes-lc.txt - now ignoring trivial differences |
| 18:07 | <Philip`> | XHTML 1.0 Strict beats HTML 4.01 Strict by a lot |
| 18:09 | <hsivonen> | annevk: can you suggest how to make the table clearer? it seems inappropriate to suggest that IE7's standards mode were more standards that Gecko/WebKit/Opera Almost Standards |
| 18:10 | <Philip`> | hsivonen: Label it "standards" but colour it in the "almost standards" darker green? |
| 18:10 | <Philip`> | I'm sure that's bad for accessibility, though |
| 18:12 | <annevk> | http://www.metah.ch/blog/2008/01/23/w3c-working-draft-use-html5-not-flash-or-silverlight/ |
| 18:12 | <annevk> | hsivonen, "IE standards mode" |
| 18:12 | <annevk> | hsivonen, with an additional paragraph that explains the difference |
| 18:13 | <hsivonen> | annevk: hmm... |
| 18:15 | <hsivonen> | annevk: I don't quite like that characterization. but perhaps I have to do something like that when IE8 comes out at the latest |
| 18:16 | <hsivonen> | it isn't quite fair to suggest Netscape 6 or Mac IE 5 standards modes were on par with contemporary standads modes, either |
| 18:18 | <annevk> | at some point you should probably put the historical info in a separate table |
| 18:20 | <hsivonen> | annevk: or do what I did with the gamma article: put the conclusion in the article and move the research results (for those who don't take my conclusions on face value) on another page |
| 18:22 | <hsivonen> | annevk: after all, the point of the table isn't to suggest that Web authors use the data. the point is to convince them that I known what I'm talking about (and to give browser QA something to test with) |
| 18:27 | <annevk> | i've seen at least one post now suggesting that MS might want to push HTML5 forward by accepting that DOCTYPE |
| 18:27 | annevk | wondered when that would happen |
| 18:29 | <hsivonen> | I don't have the bandwidth to correct all the doctype sniffing misconceptions that have been flying around in various blog comment sections. |
| 18:30 | <hsivonen> | I guess I need to update my doctype page sooner or later nonetheless |
| 18:56 | <SadEagle> | Philip`: 83.3% :-) |
| 18:58 | <Hixie> | Philip`: I don't think so |
| 18:58 | <Philip`> | SadEagle: I hope you won't make me feel guilty if I ever suggest that the spec should be changed to be more compatible with Firefox/Safari/Opera, if that'll harm your test results since you've actually been following the spec :-) |
| 18:59 | <SadEagle> | Philip`: that happened already with the path transformation... |
| 19:00 | <SadEagle> | Philip`: it's not that bad, though, since there at least is a spec. |
| 19:00 | <SadEagle> | Though fredrik was still reverse-engineering Safari's arcTo.. |
| 19:02 | <Philip`> | SadEagle: Does http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2007-July/012117.html miss some parts of arcTo? |
| 19:03 | <SadEagle> | Philip`: I won't know, and fredrik isn't around atm |
| 19:47 | <jwalden> | Hixie: HTML5 says data: URIs have no domain; spec says that document.domain must then be null, but postMessage doesn't quite say how "has no domain" maps to the |domain| attribute on the dispatched event -- is it likewise null, or is it the empty string? |
| 19:48 | <hsivonen> | Real pre-ISO-8859-1, pre-HTTP 1.0 Web content: http://montaasi.tky.fi/vanha/johtis/vanhat/1957/Kokous-27.09.1957.txt |
| 19:54 | <hsivonen> | hmm. I need a non-disruptive way to trigger reflow in Gecko from JS |
| 19:55 | <annevk> | jwalden, the URI itself? |
| 19:56 | <jwalden> | annevk: a document contained in a data: URL uses postMessage -- what's the value of the domain attribute on the event the receiver will get, null or ""? |
| 20:03 | <annevk> | oh wait, data: doesn't have a domain so it can't be itself |
| 20:03 | <annevk> | dunno |
| 20:03 | <annevk> | haven't tested that |
| 20:07 | <jwalden> | it's super-edge-case, but I'm implementing it now and can do either way -- figure I should do it right the first time |
| 20:13 | <annevk> | do you have a test using document.postMessage() ? |
| 20:13 | <annevk> | i could test in Opera |
| 20:32 | <h3h> | I'm sufficiently confused now |
| 20:32 | <h3h> | http://blogs.msdn.com/cwilso/archive/2008/01/22/i-feel-happy-too.aspx#7202711 |
| 20:32 | <h3h> | so...the IE version lock-in only applies for HTML4 / XHTML1 doctypes? |
| 20:32 | <h3h> | which means HTML5 can proceed using the latest (most standards-compliant) mode of IE? |
| 20:33 | <jwalden> | annevk: sec, I'll write something up |
| 20:35 | <Philip`> | h3h: Apparently so, though it's kind of hard to tell yet since Microsoft hasn't been giving out many details and people have to try to piece together random blog comments to work out how it works... |
| 20:35 | <jwalden> | annevk: data:text/html,<script>window.onload%20=%20function()%20{%20document.addEventListener("message",%20function(e)%20{%20alert("domain:%20"%20+%20e.domain%20+%20",%20type:%20"%20+%20(typeof%20e.domain));%20},%20false);%20window.postMessage("foo");%20};</script> |
| 20:35 | <h3h> | yeah, kinda ridiculous |
| 20:35 | <h3h> | they need to publish a technical document |
| 20:36 | <h3h> | that seems like kind of a big detail to leave out |
| 20:36 | <h3h> | I think it's the difference between me caring a lot and not caring at all |
| 20:36 | <Philip`> | If <!doctype not-html-4-or-xhtml-1> gets maximum-standards-support, that would prevent much of the criticism that they've been getting, so it does seem odd to have not mentioned it |
| 20:39 | annevk | -> mtg |
| 21:01 | <gsnedders> | I love half the comments on the IE blog post. "Opt-in to standards? Never!" while assuming the current doctype switch is a standard. |
| 21:12 | <gsnedders> | anyone know when unicode support was added into Python? |
| 21:16 | <gsnedders> | 2.0, seemingly |
| 21:17 | <Philip`> | What counts as "support"? |
| 21:22 | <gsnedders> | I mean in having a native Unicode string type |
| 21:25 | Philip` | adds ieparsehack.js to his web page |
| 21:39 | <zcorpan_> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3Cbody%3E%0D%0A%3Cscript%3Edocument.createElement('bar')%3C%2Fscript%3E%0D%0A%3Cbar%2F%3Ex%3Cbar%3Ey%3C%2Fbar%3Ez%3C%2Fbar%3E |
| 21:39 | <mpt> | Philip`, what does it do? Change the render mode between IE7 and IE8 every 0.5 seconds? |
| 21:40 | <Dashiva> | Philip`: Is the first bar supposed to be self-closing? |
| 21:42 | <zcorpan_> | Dashiva: was that meant to me? |
| 21:42 | <Dashiva> | right |
| 21:42 | <zcorpan_> | well, it is in ie |
| 21:42 | <annevk> | jwalden, empty string in Opera |
| 21:43 | <Dashiva> | zcorpan_: Well, it's <bar/>, then <bar></bar> and finally an unmatched </bar> right now |
| 21:43 | <annevk> | jwalden, sorry for the delay, had a meeting and window.postMessage() does not work in Opera; we use document.postMessage() |
| 21:43 | <jwalden> | ah |
| 21:43 | <jwalden> | are you planning on supporting both eventually, then? |
| 21:43 | <zcorpan_> | Dashiva: right |
| 21:44 | <annevk> | jwalden, I suppose so... |
| 21:44 | <jwalden> | we currently return an empty string, for what that's worth |
| 21:44 | <zcorpan_> | Dashiva: my point is that createElement() causes ie to use "xml parsing" for that tag name, same as tags that have colons in them |
| 21:44 | <annevk> | ok, that's the same then |
| 21:44 | <zcorpan_> | i.e. /> is self-closing, unlike for "html parsing" |
| 21:45 | <jwalden> | hm |
| 21:45 | <jwalden> | too bad we can't have the same behavior for both |
| 21:45 | <Dashiva> | zcorpan_: Aren't all unknown elements self-closing by default in IE? |
| 21:45 | <jwalden> | dunno how much a compatibility concern it'd be to make both null, since changing document.domain is risky-ish |
| 21:45 | <zcorpan_> | Dashiva: not <foo:bar> |
| 21:46 | <annevk> | jwalden, does the domain value do what document.domain does? |
| 21:46 | <annevk> | jwalden, I don't think there's content out there that relies on it being "" so that should be changeable |
| 21:46 | <Dashiva> | zcorpan_: Yeah, but the issue is making the new elements work in IE, isn't it? And those don't have colons |
| 21:46 | <jwalden> | annevk: no, data:text/html,<script>alert(document.domain)</script> alerts null for me while the previous was empty string |
| 21:47 | <annevk> | that's the empty string in Opera |
| 21:47 | <zcorpan_> | Dashiva: yep. *but*, integration of svg and mathml has been discussed, with /> syntax |
| 21:47 | <jwalden> | haha |
| 21:47 | <annevk> | thank god |
| 21:47 | <annevk> | we're internally consistent :p |
| 21:47 | <jwalden> | what a mess |
| 21:48 | annevk | tries Safari |
| 21:48 | <annevk> | Safari does empty string too it seems |
| 21:49 | <annevk> | data:text/html,<script>alert(document.domain === "")</script> yields true anyway |
| 21:49 | annevk | hopes Firefox can change |
| 21:49 | <jwalden> | guess Hixie may want to reconsider null vs. empty string |
| 21:49 | <jwalden> | I can certainly change postMessage |
| 21:50 | <jwalden> | document.domain, probably as well |
| 21:50 | <jwalden> | shouldn't be hard, question is whether it breaks people |
| 21:50 | jwalden | would hope not |
| 21:52 | <annevk> | they'd rely on non-standards! |
| 21:52 | <annevk> | can you imagine? |
| 21:53 | <gsnedders> | But they aren't MS, so no need to do much. |
| 21:53 | <Philip`> | mpt: It just calls document.createElement('header') etc |
| 21:54 | <mpt> | Seriously, I wonder whether changing the value of the <meta> element after-the-fact works |
| 21:54 | <gsnedders> | mpt: it won't, the action is done when the parser comes across it |
| 21:56 | <mpt> | that's no fun |
| 21:57 | <gsnedders> | Or at least that's the impression I got from Chris Wilson when talking to him about it over a month ago in #html-wg |
| 21:59 | <Philip`> | I wonder how nicely it will interact with the Live DOM Viewer's method of dynamically constructing pages... |
| 22:25 | <roc_> | of course changing the meta value could not work |
| 22:41 | <jruderman> | Hixie: why doesn't <a ping> send referrer? (or cookies or http auth) |
| 22:41 | <jruderman> | Hixie: not sending referrer seems like it invites CSRF |
| 22:42 | jruderman | finds the changelog entry pointing at http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2008-January/013637.html and starts reading it |
| 22:43 | <jruderman> | ugh |
| 22:44 | <jruderman> | sites that blacklist attributes in an attempt to prevent XSS/CSRF are broken |
| 22:44 | <Hixie> | not sending referrer seems to invite CSRF, but that's blocked by not sending cookies, no? |
| 22:44 | <Hixie> | there's no question that the sites in question are broken |
| 22:44 | <jruderman> | not sending referrer+cookies makes it harder to prevent CSRF against the link tracker |
| 22:44 | <Hixie> | why? |
| 22:45 | <jruderman> | because you have to include a magic token in the ping url |
| 22:45 | <jruderman> | instead of being able to check the referrer |
| 22:47 | <annevk> | Referer is not that reliable... |
| 22:48 | <jruderman> | it's a whole lot more sane than magic tokens |
| 22:52 | <annevk> | dare I suggest using a new HTTP method "PING" ... |
| 22:52 | <Philip`> | Only if I can suggest another new HTTP method "PONG" |
| 22:53 | <Philip`> | which gets sent in response to PINGs, or gets sent if the web server thinks the client smells |
| 22:53 | <gsnedders> | is that case-insensitive like currently defined methods or not? |
| 22:53 | <annevk> | or if the server wants to play PING/PONG |
| 22:55 | <annevk> | maybe PING/PONG is more server-to-server communication |
| 22:56 | <othermaciej> | clearly we should use SYN/ACK |
| 22:56 | <othermaciej> | or GET/200 OK |
| 22:57 | <Hixie> | jruderman: sorry, had to run off briefly |
| 22:58 | <Hixie> | tbl had the interesting idea of using a raw UDP packet |
| 22:58 | <Hixie> | but i don't know that that's where we want to go |
| 22:58 | <Hixie> | jruderman: i'm very open to feedback on what to do about the referer thing |
| 22:59 | <Philip`> | RX packets:100854 errors:1057 dropped:230798 overruns:0 frame:0 |
| 22:59 | <Philip`> | I'm not sure UDP is a great idea :-) |
| 23:00 | <annevk> | how easy is it to script against UDP on the server? |
| 23:01 | <Hixie> | (udp is a bad idea for a host of reasons, i wouldn't worry about seriously considering it) |
| 23:01 | <Philip`> | You need to run a new server process, which is possibly difficult on many web hosts |
| 23:02 | <annevk> | i see, i'd still be interested to know though |
| 23:02 | <Philip`> | although if the protocol is simple enough, your server process could be "netcat -ul 1234 >> ping_log" |
| 23:03 | <annevk> | ah, i see |
| 23:03 | <annevk> | that shouldn't be too tricky i think |
| 23:03 | <Philip`> | (You'd need to be a bit carefuller to avoid data corruption if people violate the protocol when sending you packets) |
| 23:04 | <Philip`> | (Also, it'd be pretty nasty with virtual hosting) |
| 23:05 | <Hixie> | you'd also have troubles with proxies and firewalls |
| 23:08 | <SadEagle> | Philip`: if you care, getImageData/putImageData/toDataURL are in in konq svn (both branches..) |
| 23:09 | <SadEagle> | Philip`: still need to fix the onload bug that may cause some pixmap-based testcases to fail intermittengly.. |
| 23:10 | <Philip`> | SadEagle: Sounds good - I ought to be busy with other things tonight, so I'll try to avoid the distraction of trying it out right now, but I'll have a look soon and see if I can break it :-) |
| 23:10 | <SadEagle> | thanks :-). It's a lot easier to try it out now :-) |
| 23:11 | <fredrikh> | Philip`: SadEagle mentioned you had a question about arcTo in khtml? |
| 23:13 | <Philip`> | fredrikh: I think he mentioned you'd been looking at Safari's arcTo, so I was just wondering if you'd seen http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2007-July/012117.html |
| 23:14 | <fredrikh> | Philip`: i had |
| 23:14 | <Philip`> | fredrikh: Okay |
| 23:14 | <Philip`> | fredrikh: I don't think I have anything new or useful to add now :-) |
| 23:15 | <fredrikh> | Philip`: but there's one thing that you didn't mention there: http://ktown.kde.org/~fredrik/canvas/arcto3.html |
| 23:15 | <fredrikh> | what happens when the lines are too short for the circle to tangent them :) |
| 23:17 | <Philip`> | "the line defined by the points (x0, y0) and (x1, y1)" etc meant the infinite line which passes through those points (assuming the points are not coincident) |
| 23:17 | <fredrikh> | i was also interested in knowing where the final point ends up when the angle between the tangent lines is 0 degrees |
| 23:18 | <fredrikh> | since you can't really add an infinitely long line to the path |
| 23:18 | <fredrikh> | Philip`: ah, i must have missed that |
| 23:18 | <Philip`> | You can do a good approximation of an infinitely long line, as long as you're not silly like Firefox and have a very limited coordinate range |
| 23:20 | <fredrikh> | yeah, the current implementation just adds a very long line segment |
| 23:25 | <SadEagle> | fredrikh: I think I might know what's up with the slightly-off colors.. |
| 23:27 | <SadEagle> | fredrikh: it's probably just AA blending w/the red BG with very high, but not full, pixel coverage |
| 23:55 | <fredrikh> | SadEagle: well, the odd thing about it is that it's pixels in the middle, not near the edges |
| 23:56 | <SadEagle> | well, the one I tested had it on the edge, and not the middle ones... |
| 23:57 | <fredrikh> | in the skewed transform test, and one of the arctTo tests it's in the middle |
| 23:57 | <fredrikh> | or close to it |
![http://external.example.org/img.gif">](http://external.example.org/img.gif">){kind=link}