| 00:54 | <roc> | creative use of <canvas> and <audio>: http://www.bluishcoder.co.nz/2008/09/javascript-space-invaders-emulator.html |
| 02:47 | <Hixie> | you can now click on <dfn> elements in html5 |
| 03:25 | <MikeSmith> | Hixie: click-on-<dfn> thing is great |
| 03:25 | <MikeSmith> | Ted Nelson would be proud |
| 04:14 | <kangax> | what would be the best way to draw ellipse on canvas? |
| 04:28 | <roc> | kangax: transform a circle? |
| 04:28 | <kangax> | that's what i'm doing now |
| 04:29 | <GregHouston> | kangax: if the ellipse is the only thing you are drawing then you could just draw a circle with arc and scale it in one direction. Otherwise you might want to go with a couple bezierCurveTo()s or plot it point by point with a parametric equation. |
| 04:29 | <kangax> | would it be possible to make one with 2 bezier? |
| 04:29 | <kangax> | GregHouston: right. now i just want to find out which one would be faster |
| 04:33 | <GregHouston> | kagnax: faster for you or the browser. I couldn't say for the browser. The easiest way is going to be the scaled circle. At the bottom of this page is the parametric equation. You can use it like the spirograph example at developer.mozilla.org. http://en.wikipedia.org/wiki/Ellipse |
| 04:38 | <kangax> | GregHouston: for the browser |
| 04:38 | <kangax> | I'll go with transformed circle, thanks |
| 04:39 | <kangax> | (building svg to canvas parser) |
| 04:39 | <GregHouston> | Nice. What language is it written in? |
| 04:41 | <kangax> | GregHouston: JS |
| 04:43 | <kangax> | It's mostly a general purpose library that makes it easy to draw basic shapes on canvas (with support for drag/drop and any affine transformations) |
| 04:44 | <GregHouston> | Do you have a blog? I would be interested in keeping tabs on it. |
| 04:44 | <kangax> | I do, but it's not related to canvas : ) |
| 04:45 | <kangax> | GregHouston: http://thinkweb2.com/projects/prototype/ |
| 04:50 | <GregHouston> | kangax: Cool. I'll add you to my feeds and see if you don't post anything about it. I wrote a function for drawing a rectangle with rounded corners if you haven't got to that one yet. |
| 04:50 | <kangax> | nice |
| 04:50 | <kangax> | haven't yet |
| 04:54 | <GregHouston> | kangax: I pasted it here: http://paste.mootools.net/f3be6b0e8 |
| 04:55 | <kangax> | didn't know mootools have pastie |
| 04:55 | <kangax> | : ) |
| 04:55 | <GregHouston> | :) |
| 08:06 | <Hixie> | you know, when we started html5, xforms and xhtml2 provided a kind of motivation to do a good job and do it quick |
| 08:06 | <Philip`> | http://crypto.stanford.edu/websec/chromium/ - ooh, an HTML5 page |
| 08:06 | <Hixie> | as competition is wont to do |
| 08:07 | <Hixie> | but our success has removed the credibility from those efforts |
| 08:09 | <MikeSmith> | I guess Flash and Silverlight provide a little more motivation |
| 08:09 | <MikeSmith> | or even Gears |
| 08:10 | <MikeSmith> | especially now that we have a competing browser with built-in Gears support |
| 08:13 | <takkaria> | Philip`: HTML5 with XHTML talismans, too :) |
| 08:13 | <othermaciej> | I'm not worried about XHTML2 but I am worried about the open web platform vs. proprietary alternatives |
| 08:13 | <othermaciej> | and I am not even so afraid of that (not as much as roc anyway) |
| 08:14 | <takkaria> | ugh, another +1 post on public-html |
| 08:17 | <Hixie> | the problem with flash and silverlight is that they're competitors to the overall platform, not to html5 specifically |
| 08:17 | <Philip`> | othermaciej: Are you not so afraid because you don't think it's so likely, or because you don't think it's so bad if a better (but proprietary) alternative wins? |
| 08:17 | Hixie | isn't worried as much as he was because of the former |
| 08:18 | <othermaciej> | Philip`: I don't think it's so likely |
| 08:18 | <hsivonen> | It's sad that in this day and age, *anyone* can launch a new Flash-like thing |
| 08:18 | <hsivonen> | i.e. Silverlight |
| 08:18 | <othermaciej> | I feel like Flash is close to peaking, and I doubt Silverlight will see much some success in the grand scheme of things |
| 08:18 | <hsivonen> | though I am hoping that Silverlight scares Adobe enough for them to open-source all parts of Flash that are licensable by Adobe |
| 08:19 | <othermaciej> | and I think Google is sensitive enough to criticism that native web platform features will mostly obsolete Gears (or reduce it to an IE uncripplement add-on) |
| 08:20 | <hsivonen> | also, I'm wondering about the strategic smartness of the whole Novell Mono thing |
| 08:20 | <hsivonen> | though things could be different if OpenJDK had arrived earlier |
| 08:21 | <hsivonen> | now it's like Mono tries to rescue legacy Java code to run on it |
| 08:21 | <hsivonen> | instead of OpenJDK rescuing legacy C# code |
| 08:21 | <othermaciej> | wait, what? |
| 08:22 | <hsivonen> | othermaciej: you can use ikvm.net to run Eclipse on Mono |
| 08:22 | <othermaciej> | it translates JVM bytecode to MSIL? |
| 08:22 | <hsivonen> | but as far as I'm aware, there isn't a viable project that allowed one to take a C# project on bring it over to the JVM |
| 08:22 | <hsivonen> | othermaciej: either bytecode or Java source |
| 08:23 | <hsivonen> | but to come to back to the Web relevance of Mono |
| 08:23 | <hsivonen> | it seems to me that Moonlight is helping Silverlight when it's done so early |
| 08:23 | <othermaciej> | it definitely is |
| 08:24 | <Philip`> | That's kind of its intention |
| 08:24 | <othermaciej> | Miguel has a weird love of Microsoft technologies |
| 08:25 | <Hixie> | to be fair, C# and .Net are pretty solid technologies, from a purely technical perspective |
| 08:25 | <Hixie> | i mean they're not perfect by any means |
| 08:25 | <Hixie> | but they compare pretty favourably compared to the rest of the industry's best efforts |
| 08:26 | <Hixie> | i don't think they're overly suitable for the web |
| 08:26 | <Hixie> | and i think their microsoft pedigree and lack of truly open development process are fatal flaws, though |
| 08:26 | <hsivonen> | Hixie: it seems to me that giving control of the GNU platform direction to MS isn't a good idea |
| 08:27 | <Hixie> | oh i don't think mono is strategically sound |
| 08:27 | <hsivonen> | Hixie: I'd prefer to get the good stuff from C# bolted onto Java even if it meant some more warts |
| 08:28 | <Hixie> | i'm not a big java fan myself, i'd rather have many different languages available |
| 08:28 | <Hixie> | but that's besides the point |
| 08:29 | <hsivonen> | the problem is that there are only two big multi-language server-side VMs |
| 08:29 | <hsivonen> | I'm not expecting Parrot to rock the market |
| 08:29 | <hsivonen> | but on the client side, the approach GWT is taking is *very* interesting |
| 08:29 | <hsivonen> | I'd like to see that for more languages |
| 08:29 | <hsivonen> | particularly Python |
| 08:29 | <Hixie> | the mono effort could have been strategically sound -- and still could be, actually -- if they used a slightly different path |
| 08:30 | <Hixie> | if they can get enough critical mass of people moving in both directions (mono <-> .net) then they could start introducing "innovations" |
| 08:30 | <hsivonen> | It seems to me that the direction the client VM is going is better for concurrency than the direction the JVM went |
| 08:30 | <othermaciej> | JS implementations (other than IE) are getting fast enough that compiling to JS might not be totally insane |
| 08:30 | <Hixie> | i.e. embrace-and-extend |
| 08:31 | <hsivonen> | that is, running normal-looking (not erlang) language with workers instead of shared-memory threads |
| 08:31 | <Hixie> | and that would end up forcing microsoft's hand a bit with silverlight |
| 08:31 | <Hixie> | which would make it harder for them to abuse the platform |
| 08:31 | <Hixie> | but as i understand it right now mono is just playing follow-the-leader with mono in the role of follower |
| 08:31 | <Hixie> | anyway |
| 08:32 | <Hixie> | othermaciej: isn't that basically what v8 does? |
| 08:32 | <Hixie> | othermaciej: oh, you mean compiling _to_ js |
| 08:32 | <Hixie> | s/mean/said/ |
| 08:32 | <Hixie> | i read just "compiling js" |
| 08:35 | <Hixie> | i wonder who the +1 posts are intended for |
| 08:43 | <othermaciej> | every JS implementation "compiles" in some sense |
| 08:55 | <hsivonen> | has Google announced anything about <video> in Chrome or Android WebKit? |
| 08:55 | <othermaciej> | I have not heard announcements but I believe <video> does not work in Chrome currently |
| 08:56 | <othermaciej> | it has the Safari 3.1 featureset in the engine minus (at least) @font-face, <video>, text-shadow and Database |
| 08:56 | <hsivonen> | yeah, but they must have a plan |
| 11:22 | <Philip`> | Surprisingly many people use the script from http://cow.neondragon.net/stuff/reflection/ (or derivatives), which uses canvas (on non-IE), so that's quite a lot of indirect canvas users |
| 11:23 | <Philip`> | (Sadly it uses "if (document.all && !window.opera)" to trigger the use of 'filter: flipv progid:DXImageTransform.Microsoft.Alpha(...)' instead of canvas) |
| 11:26 | <hsivonen> | regarding the discussion the other day: it seems that Windows XP SP3 still doesn't install IE7 as an automatic update, but it does install it as an Express Update if you go to Windows Update and accept the defaults |
| 11:29 | <Philip`> | Is Silverlight still non-selected by default? |
| 11:33 | <hsivonen> | Philip`: I don't see Silverlight being offered through Windows Update at all |
| 11:35 | <hsivonen> | Windows Update tells me to upgrade to Microsoft Update |
| 11:35 | <hsivonen> | but it's not at all clear why I, as a user, should have to make a decision between two update mechanisms |
| 11:37 | <Philip`> | Maybe you have to make that decision because it would be unfair for Microsoft to leverage its privileged OS-developer position to automatically provide additional services for its own non-OS products |
| 11:37 | <hsivonen> | I see |
| 11:38 | <Philip`> | (Or maybe that's totally wrong) |
| 11:48 | <zcorpan_> | hsivonen: i'm on windows using cygwin |
| 11:48 | <hsivonen> | zcorpan_: are you using easy_install? |
| 11:48 | <zcorpan_> | hsivonen: no |
| 11:49 | <hsivonen> | zcorpan_: in that case, I suggest trying it |
| 11:51 | <zcorpan_> | python ez_install.py |
| 11:52 | <zcorpan_> | Fishished processing dependencies for setuptools==0.6c7 |
| 11:52 | <zcorpan_> | but running validator-tester.py still says ImportError: No module named simplejson |
| 11:53 | <hsivonen> | ez_setup.py should install a script names easy_install |
| 11:53 | <hsivonen> | then easy_install simplejson installs simplejson |
| 11:54 | <zcorpan_> | there's no easy_install :( |
| 11:54 | <hsivonen> | that's weird |
| 11:55 | <hsivonen> | you can get a new one from here http://peak.telecommunity.com/DevCenter/EasyInstall |
| 11:55 | <hsivonen> | (the site loads very slowly) |
| 11:56 | <zcorpan_> | thanks |
| 12:01 | <hsivonen> | validator.nu goes offline for a moment. sorry |
| 12:04 | <hsivonen> | back up |
| 12:08 | <zcorpan_> | easy_install.exe is in my Python24\Scripts directory but trying to use it doesn't work |
| 12:09 | <hsivonen> | :-( |
| 12:10 | <Lachy> | just found out there will be a WebJam on while I'm in Aus. I should try to organise a 3 minute presentation for it. |
| 12:10 | <hsivonen> | I'm completely inexperienced when it comes to troubleshooting Windows Python problems |
| 12:11 | <Lachy> | zcorpan_, it probably means that the directory isn't in the PATH environment variable |
| 12:11 | <zcorpan_> | Lachy: i added it to PATH... do i need to restart cmd for it to start working? |
| 12:11 | <Lachy> | yes |
| 12:11 | <zcorpan_> | ah |
| 12:12 | <Lachy> | you could just type echo %PATH% and see if it was really in it |
| 12:13 | <Lachy> | aargh, I hate how Windows update sometimes installs critical updates and then does a restart without even asking. Especially, when I'm not around to stop it! |
| 12:13 | <zcorpan_> | it worked after i did echo %PATH% but not before |
| 12:15 | <zcorpan_> | ok simplejson is installed |
| 12:16 | <Lachy> | hmm, maybe I should get http://html5.lachy.id.au/beta/ finished and present that at WebJam, since it will have a lot of nice features for web developers |
| 12:16 | <Philip`> | Lachy: Set it to ask before installing updates, and then it won't do it silently |
| 12:16 | <Philip`> | (It'll still force you to reboot after the installation, though) |
| 12:16 | <Lachy> | Philip`, I'm sure I have |
| 12:17 | <Philip`> | Oh |
| 12:17 | <Lachy> | ah, crap. That machine was still set to automatic :-( |
| 12:17 | <Lachy> | fixed now |
| 12:17 | <zcorpan_> | or i think it was installed... but validator-tester.py still says it's missing |
| 12:18 | <zcorpan_> | Lachy: you can stop the automatic updates service to make it not reboot |
| 12:19 | <Philip`> | zcorpan_: You could just copy simplejson's .py files into the right place, and not bother with the whole installation setup thing |
| 12:19 | <Philip`> | (It has a _speedups.so (or .dll or whatever) too but that appears to be optional) |
| 12:19 | <didymos> | Lachy, I experienced that same thing yesterday, and it was a bit annoying, because I was in the middle of a conversation, and couldn't take the time to disable it |
| 12:19 | <zcorpan_> | Philip`: which .py files and where's the right place? |
| 12:19 | <didymos> | so I just had to click 'later' everytime it tried to restart |
| 12:20 | <BenMillard> | didymos, I have the same experience. To me, the "later" button should mean "when I decide to do it myself". |
| 12:21 | <didymos> | BenMillard, certainly |
| 12:21 | <Philip`> | zcorpan_: The really lazy way is to copy the 'simplejson' directory from http://pypi.python.org/packages/source/s/simplejson/simplejson-1.9.2.tar.gz into the directory you're running Python from, and I think then it should pick it up properly |
| 12:21 | <BenMillard> | rather than "bug me again about this in 5 minutes, then do it of your own accord if I'm not here when you ask, losing all my unsaved work for me". |
| 12:22 | <didymos> | unfortunately, it seems they have to go for the lowest common denominator, if they want it to be installed on non-techie users' machines |
| 12:22 | <BenMillard> | didymos, I doubt non-techies leave their machines running for many days in a row |
| 12:22 | <BenMillard> | didymos, so interpreting "later" as I describe would be fine on non-technie machines, since they don't run for that long at a time anywy. |
| 12:22 | <didymos> | good point, although I have a case to prove you wrong, but in theory I believe you'd be right ;) |
| 12:23 | <Philip`> | BenMillard: Now that Vista's hidden the 'power off' button, I'd guess many people just suspend or hibernate it and never actually reboot |
| 12:23 | <didymos> | BenMillard, my mother has a laptop in the kitchen, which she leaves on so she can get to it fast -- it can easily run for weeks in a row |
| 12:23 | <zcorpan_> | wonder if Python24 is the right directory because i have python 2.5 |
| 12:23 | <Philip`> | particularly on laptops where you just shut the lid and it'll happily suspend itself for days |
| 12:24 | <didymos> | exactly |
| 12:24 | <zcorpan_> | how do i figure out which directory my python is running from |
| 12:24 | <Philip`> | zcorpan_: 'which python' |
| 12:25 | <Philip`> | (assuming Cygwin with suitable packages installed) |
| 12:25 | <zcorpan_> | Philip`: /usr/bin/python |
| 12:26 | <Lachy> | ideally, computers should never need to reboot to apply a patch. They should just apply hot patches to everything |
| 12:26 | <Philip`> | zcorpan_: Is that a symlink to a python2.x file? |
| 12:26 | <Philip`> | (ls -l /usr/bin/python) |
| 12:26 | <didymos> | Lachy, Steve Yegge, I believe? :) |
| 12:27 | <didymos> | at least he said it at some point |
| 12:27 | <Philip`> | Lachy: The problem with that is that it can get extremely complicated, particularly if you care about efficiency in the normal not-applying-patches state |
| 12:27 | <zcorpan_> | wait is it my cygwin/bin directory? |
| 12:28 | <Philip`> | zcorpan_: It's probably cygwin/usr/bin |
| 12:28 | <Philip`> | (in the Windows view of the filesystem) |
| 12:28 | <zcorpan_> | Philip`: ok |
| 12:29 | <zcorpan_> | Philip`: it's empty but i'll try pasting the simplejson directory there |
| 12:29 | <Philip`> | zcorpan_: Oh, no |
| 12:29 | <Philip`> | zcorpan_: I meant the directory which you are in when you run Python, not the directory which the Python you run is in |
| 12:29 | <zcorpan_> | Philip`: ah |
| 12:30 | <Philip`> | zcorpan_: Ah, if cygwin/usr/bin is empty then I guess /usr/bin (which sounds non-empty) must be somewhere else, but I don't know where |
| 12:32 | <virtuelv> | is Cameron McCormack usually present on IRC? |
| 12:32 | <zcorpan_> | ok now i got a different error running validator-tester.py |
| 12:32 | <Philip`> | zcorpan_: That sounds promising |
| 12:32 | <Lachy> | virtuelv, that's heycam |
| 12:34 | <zcorpan_> | in dumpReference simplejson.dump({uri:self.database[uri]}, handle, cls=ValidationErrorMessageEncoder, sort_keys=True) |
| 12:35 | <virtuelv> | Lachy: thanks |
| 12:38 | <Philip`> | zcorpan_: What is the error message there? |
| 12:38 | <hsivonen> | zcorpan_: do you have an up-to-date svn checkout? |
| 12:38 | <zcorpan_> | Philip`: KeyError: 'http://simon.html5.org/test/validator/character-encoding/non-ascii.html'; |
| 12:39 | <hsivonen> | that line should now read simplejson.dump({uri:self.database[uri]}, handle, cls=ValidationErrorMessageEncoder, sort_keys=True, indent=2) |
| 12:39 | <hsivonen> | zcorpan_: I think it throws a key error if you don't have the URI you gave it already in the database |
| 12:40 | <hsivonen> | the database being a file called db.json in the working directory by default |
| 12:41 | <zcorpan_> | hsivonen: aha |
| 12:42 | <zcorpan_> | i should be using dumpuri |
| 12:49 | <zcorpan_> | hsivonen: but if i write a test, don't you need to add it to your database? |
| 12:53 | <zcorpan_> | hsivonen: so i can just point to a testcase in a bug report and when you have fixed the bug you do adduri and i don't have to get this working |
| 12:54 | <hsivonen> | zcorpan_: yes. so I'd need you to send me your db.json file (preferable containing tests only form you) so I can merge it in |
| 12:54 | <hsivonen> | zcorpan_: that would work, too |
| 12:54 | <zcorpan_> | hsivonen: ok. then i won't spend more time trying to get it working :) |
| 12:55 | <hsivonen> | OK. sorry about not thinking about this before |
| 12:55 | <zcorpan_> | that's ok |
| 12:57 | <zcorpan_> | hsivonen: your sniffer is still looking at 512 bytes right? |
| 12:57 | <hsivonen> | zcorpan_: yes |
| 12:57 | <zcorpan_> | hsivonen: you think it will continue to do so? |
| 12:58 | <hsivonen> | I think it will continue to look for n bytes where n is likely to be 512 or 1024 depending on what Hixie puts in the spec |
| 12:58 | <hsivonen> | 512 is a good guess |
| 12:58 | <hsivonen> | though IIRC, WebKit switched to 1024 |
| 12:58 | Philip` | suggests 768 |
| 12:58 | <hsivonen> | Philip`: jokingly or seriously? |
| 12:58 | <Philip`> | hsivonen: I'm just trying to be awkward |
| 13:00 | <Philip`> | I presume "one packet" could be the most sensible thing for a browser to use, rather than a fixed number of bytes, in terms of being able to render the page as quickly as possible - maybe it'd be interesting to work out how many bytes of content do fit in the first packet... |
| 13:01 | <hsivonen> | Philip`: my sniffer works with blocking Java IO |
| 13:01 | <hsivonen> | Philip`: so I don't see packets |
| 13:02 | <hsivonen> | Implementations where the network layer pushes data to parser could be different |
| 13:02 | <Philip`> | hsivonen: Your abstractions need more leaks |
| 13:03 | <Philip`> | although I suppose it's Java's abstractions in that case |
| 13:03 | <zcorpan_> | now ftp won't connect |
| 13:03 | <hsivonen> | I couldn't follow the Gecko source on this topic with full confidence, but I think it scans the first packet |
| 13:04 | <hsivonen> | Philip`: If I wrote a NIO driver and if NIO can be told to push data on a per packet basis, then it would be doable |
| 13:04 | <hsivonen> | Philip`: but SAX assumes IO--not NIO |
| 13:05 | <hsivonen> | Philip`: the Validator.nu HTML Parser core assumes neither and, by design, fits a NIO/Gecko-like IO setup as well |
| 13:13 | <hsivonen> | Hixie: I implemented the <rp> spec literally without allowing space characters around the paren |
| 13:20 | <zcorpan_> | Lachy: could you use a different favicon? my brain thinks that the html5 tools tab is the spec |
| 13:38 | <Philip`> | Looks like the mean HTTP header size (from a very small sample from dmoz.org) is around 290 bytes (and the median is 260, and very few are above ~500) |
| 13:38 | <Philip`> | so you should get at least a kilobyte of content in the first packet |
| 13:38 | <Lachy> | zcorpan_, ok |
| 13:38 | <Lachy> | zcorpan_, any suggestions for what I should use? |
| 13:38 | <Philip`> | (Interestingly enough, Google seems to send 1418 byte TCP packets, whereas everyone else sends 1448 bytes) |
| 13:39 | <zcorpan_> | Lachy: anything but the whatwg icon. no icon wfm :) |
| 13:39 | <zcorpan_> | Lachy: don't you have one for you blog? |
| 13:39 | <Lachy> | no |
| 13:39 | <Lachy> | I need one though |
| 13:39 | <Lachy> | I'll remove it for now, and replace it later |
| 13:39 | <zcorpan_> | ok |
| 13:39 | <Philip`> | Use the Space Invaders favicon |
| 13:40 | <Philip`> | Oops, it was Defender |
| 13:41 | <Lachy> | done |
| 13:41 | <Lachy> | I'll find an icon of some tools or something |
| 13:44 | <Lachy> | I need to figure out how to set up my localhost server to use custom .localhost domains (or whatever) so they can mirror my web server better |
| 13:45 | <Lachy> | some of the things are broken in my local copy because I have paths like http://localhost/html5.lachy.id.au/ and that page has non-relative paths like "/script/...", which break |
| 13:46 | <Philip`> | You could run each on a different port |
| 13:46 | <Lachy> | yeah, maybe. I'll have to figure out how to get MAMP to do that |
| 13:46 | <Philip`> | (and multiple servers, rather than one server with virtual hosts) |
| 13:47 | <Lachy> | I'd have have a domain like http://html5.lachy.id.au.local/ or something |
| 13:47 | <BenMillard> | Lachy, I use Apache with <virtualhost> in the httpd.conf to produce http://calthorpe.local/ and http://ben.local/ and so forth |
| 13:47 | <BenMillard> | Apache as part of the WAMP package, I mean. |
| 13:47 | <Lachy> | ok, I can probably set that up |
| 13:47 | <Lachy> | yeah, apache is in MAMP too |
| 13:48 | <BenMillard> | <VirtualHost 127.0.0.1> |
| 13:48 | <BenMillard> | DocumentRoot "C:\Program Files\WAMP\www\ben" |
| 13:48 | <BenMillard> | ServerName ben.local |
| 13:48 | <BenMillard> | </VirtualHost> |
| 13:48 | <BenMillard> | is one of them |
| 13:49 | <BenMillard> | oh, you have to edit the "hosts" file if your on Windows so the "ben.local" part works |
| 13:49 | <BenMillard> | e.g.: 127.0.0.1 ben.local |
| 13:49 | <Lachy> | yeah, I knew about the hosts file |
| 13:50 | <BenMillard> | Lachy, ok I'll leave you to it. :) |
| 13:50 | <Lachy> | apparently MAMP Pro offers easy setup of virtual hosts like that |
| 13:50 | <Philip`> | Raw Apache httpd.conf offers easy setup like that :-) |
| 13:50 | <Lachy> | but it seems rather expensive. I'll see if I can set it up manually myself instead |
| 13:50 | <Philip`> | and is free |
| 13:51 | <Lachy> | yeah, I'm not paying about $75 for something I can do myself for free |
| 13:53 | <Philip`> | With that kind of attitude, why are you a Mac user? :-p |
| 14:00 | <Lachy> | Philip`, because there are no decent alternatives |
| 14:04 | zcorpan_ | briefly scans through http://www.w3.org/mid/2d509b1b0809101646y4bbda2f9k56474ee3cd72003d⊙mgc |
| 14:05 | hsivonen | notes the TOC doesn't use HTML lists |
| 14:06 | <hsivonen> | I guess that counts as a failure of HTML/CSS to produce the wanted result reliably |
| 14:07 | <takkaria> | "To facilitate independent evolution of producers and consumers, languages in distributed systems should be Extensible." |
| 14:07 | <takkaria> | (from Dave Orchard's post) |
| 14:07 | <takkaria> | surely producers that produce something that consumers don't understand doesn't help anyone, so independent evolution isn't helpful in the slightest |
| 14:08 | <BenMillard> | hsivonen, numbered lists with decimal places are faked in a variety of ways. |
| 14:08 | <Philip`> | Producers that produce something not all consumers understand still helps the consumers that do understand it |
| 14:10 | <Philip`> | though I suppose that's not "independent evolution", since at least some consumers would have to evolve to understand exactly what the producers produce |
| 14:10 | <BenMillard> | hsivonen, maybe <li value> should just allow any string, like <li value="3.2"> or <li value="2a">? Or maybe <ol type> should have a couple more values to do this automatically... |
| 14:11 | <zcorpan_> | BenMillard: you could use <ul><li>3.2 |
| 14:12 | <BenMillard> | zcorpan_, that's one of the ways I see it faked. |
| 14:12 | <BenMillard> | layout tables and with <br> are others |
| 14:12 | <zcorpan_> | <ul> seems better |
| 14:13 | <BenMillard> | yes, but <ol> with the desired markers (by some method) seems best. :) |
| 14:13 | <zcorpan_> | why? |
| 14:15 | <BenMillard> | firstly, <ol> is a closer match to the content than <ul> (unless we think <ol> is a pointless element) |
| 14:15 | <zcorpan_> | are markers searchable? |
| 14:16 | <BenMillard> | zcorpan_, wait...are you asking whether <ol> is at all useful? |
| 14:16 | <zcorpan_> | BenMillard: no i'm asking whether it's useful for TOCs |
| 14:18 | <BenMillard> | zcorpan_, since TOCs are just a numbered list of links, <ol> seems as useful for them as it is any other numbered list (whether it has links or not). |
| 14:19 | <Lachy> | cool, I got the virtual hosts to work. That was easy |
| 14:19 | <zcorpan_> | but for a TOC you want 1, 2, 2.1, 2.2, and not 1, 2, 1, 2 |
| 14:19 | <zcorpan_> | and it's nice to be able to search for "2.1" |
| 14:19 | <BenMillard> | zcorpan_, I've seen TOCs with either |
| 14:20 | <BenMillard> | zcorpan_, I see no reason for browsers to make list markers non-searchable |
| 14:20 | <zcorpan_> | BenMillard: file a bug :) |
| 14:21 | <BenMillard> | zcorpan_, I don't search for section numbers in TOCs, so I don't mind either way. |
| 14:21 | <BenMillard> | but if you want it to work, I guess you should indeed file a bug. |
| 14:22 | <zcorpan_> | searching works for me since specs i care about use <ul> :) |
| 14:24 | <BenMillard> | zcorpan_, searching also works in the ones that fake it with or layout tables. |
| 14:24 | <zcorpan_> | BenMillard: yes but i imagine they are less usable with ATs |
| 14:25 | <zcorpan_> | but i could be wrong |
| 14:26 | <BenMillard> | zcorpan_, using <ul> for a numbered list produces bullets and numbers when author CSS is disabled or overridden by user CSS. |
| 14:27 | <Philip`> | http://timepedia.org/chronoscope/ is an interesting use of GWT + <canvas> |
| 14:27 | <BenMillard> | and using the semantics of <ul> when a TOC is better matches by <ol> means any device which makes use of those semantics might be getting short-changed (such as what announcements you get in screenreaders or in a Braille output, maybe) |
| 14:28 | <BenMillard> | zcorpan_, something like <ol type="decimal-compound"> (or something along those lines) could avoid authors having to type in all the section names |
| 14:29 | <BenMillard> | and it doesn't rely on CSS and does have the <ol> semantics which best fit a TOC |
| 14:30 | <Philip`> | They'd still have to type the section numbers into the section <h2>s, though |
| 14:30 | <Philip`> | and if those are entered manually while the TOC is numbered automatically, they're more likely to get out of sync |
| 14:30 | <Philip`> | Section numbering is an outdated concept in a world of hyperlinks, anyway :-p |
| 14:32 | <BenMillard> | Philip`, for that to happen the TOC would have to differ from the sections in the document, so the TOC would be wrong anyway |
| 14:33 | <Philip`> | BenMillard: Indeed, but the effects of an error could be localised, rather than cascading through the rest of the TOC and destroying all of the numbers |
| 14:34 | <BenMillard> | Philip`, I agree that authors would still have to type into the headings and when making references within the document. |
| 14:35 | <BenMillard> | the error would only cascade to siblings that came after it and the descendents, AFAICT |
| 14:35 | <BenMillard> | and if the TOC isn't matching the document, making that error as obvious as possible actually seems like an asset to me :P |
| 14:38 | <BenMillard> | s/better matches/better matched/ |
| 14:38 | <Philip`> | This is HTML, not XHTML - errors are meant to be papered over, not made as obvious as possible :-p |
| 14:38 | <BenMillard> | s/all the section names/all the section numbers/ |
| 14:42 | <BenMillard> | time for lunch... |
| 15:04 | <hsivonen> | http://html5.validator.nu/ now supports ruby |
| 15:11 | <Philip`> | http://html5.validator.nu/?doc=http%3A%2F%2Fwww.city.kawasaki.jp%2F25%2F25zinken%2Fhome%2Fkidspage%2F seems to work, at least in the sense that it's only complaining about <rb> |
| 15:13 | <Philip`> | It would be user-friendlier if the validator recognised <rb> and gave an informative message like "You're not allowed to use <rb> and have to go to loads of effort to remove it from all your pages, because it has no effect and we don't want to allow it like we allow xmlns and xml:lang and trailing slashes" |
| 15:14 | <Philip`> | otherwise people will have no idea why it's suddenly become an error |
| 15:14 | <hsivonen> | Philip`: or we could allow it as a talisman |
| 15:14 | <Philip`> | hsivonen: That too |
| 15:15 | <Philip`> | http://html5.validator.nu/?doc=http://www.geocities.co.jp/Hollywood-Spotlight/2135/ seems to fail at encoding sniffing |
| 15:15 | <zcorpan_> | ruby isn't used much though, is it? and <rb> even less |
| 15:15 | <zcorpan_> | (compared to xmlns and xml:lang) |
| 15:16 | <Philip`> | hsivonen: and that page there gets error reports on line 0, which doesn't exist |
| 15:17 | <Philip`> | hsivonen: Also, the "show source" view has something totally bogus on line 45 that isn't in the real source |
| 15:17 | <Philip`> | (It says: "http://www.geocities.co.jp/Hollywood-Spotlight/2135/2pigpigsan.jpg";, "tiiojpol "http://www.geocities.co.jp/Hollywood-Spotlight/2135/2pigpigsan2.jpg",↩) |
| 15:18 | <hsivonen> | Philip`: the validator doesn't run heuristic sniffing (parsetree does) |
| 15:18 | <hsivonen> | Philip`: should the validator run a heuristic sniffer, too? |
| 15:19 | <hsivonen> | line 45 is certainly interesting |
| 15:20 | zcorpan_ | ponders about "Error: The font element is obsolete." vs "Error: Element rb not allowed as child of element ruby in this context. (Suppressing further errors from this subtree.)" in <RUBY><RB><FONT face="ＭＳ Ｐゴシック">区役所</F |
| 15:20 | <hsivonen> | zcorpan_: the implementations are different |
| 15:20 | <hsivonen> | should I give <rb> the <font> treatment? |
| 15:20 | <Philip`> | hsivonen: It would make the output more useful if it didn't let one error (lack of explicit encoding) cause the rest of the error report to become unreadable, since maybe people have a good reason for not fixing the first error (e.g. they're not the site author and can't fix it straight away) but they'll still want to know about subsequent errors |
| 15:21 | <zcorpan_> | hsivonen: seems reasonable if <rb> is used enough to care |
| 15:21 | <hsivonen> | now I'm most immediately curious about line 45 |
| 15:22 | <Philip`> | zcorpan_: http://lists.w3.org/Archives/Public/public-html/2008May/0620.html - 0.3% of the .jp pages use ruby, and I think most (or at least many) of those use <rb> |
| 15:22 | <hsivonen> | the unmappably byte sequence errors repeat themselves |
| 15:23 | <zcorpan_> | i get var args = 6 * div;↩ on line 45 |
| 15:25 | <zcorpan_> | Philip`: ok |
| 15:30 | <Philip`> | zcorpan_: That's the right line but the wrong page |
| 15:31 | <Philip`> | zcorpan_: It was http://html5.validator.nu/?doc=http%3A%2F%2Fwww.geocities.co.jp%2FHollywood-Spotlight%2F2135%2F&showsource=yes#l45 but now that looks correct to me |
| 15:32 | <hsivonen> | there has to be something very wrong with the show source code |
| 15:32 | <hsivonen> | something that depends on how buffer boundaries fall |
| 15:32 | <hsivonen> | sigh |
| 15:33 | <hsivonen> | why can't show source ever just be easy |
| 15:34 | <Philip`> | It is easy: slurp the whole HTTP response into a string, then parse and validate and print it out again :-) |
| 15:35 | <hsivonen> | Philip`: can't slurp without the parser figuring out the encoding |
| 15:36 | <Philip`> | s/string/byte buffer/ |
| 15:44 | <hsivonen> | well, the code I used for calculating the UTF-16-level line and col when a byte-level error occurs was utterly bogus now that the tokenizer has undergone notable changes |
| 15:45 | <hsivonen> | and that caused assertions to be violated in the show source package |
| 15:46 | <takkaria> | I do worry a bit about adding parse error reporting to Hubbub and how much that'll complicate things |
| 15:47 | <hsivonen> | takkaria: error reporting as such is not bad |
| 15:47 | <hsivonen> | takkaria: approximate error locations aren't bad |
| 15:48 | Philip` | would typically prefer a fast parser that doesn't report errors, since he's usually parsing other people's junk and doesn't care as long as it works as quickly as possible |
| 15:48 | <hsivonen> | takkaria: error locations that are good enough for the kind of lighlighting that Validator.nu does are a major PITA |
| 15:52 | <hsivonen> | Philip`: I've been toying with the idea of having a preprecessor for removing error reporting code and source location tracking code |
| 15:55 | <Philip`> | hsivonen: Would that be better than having the error reporting code etc being in methods in a class which you can switch for a minimal stub implementation class, and then let the JVM inline all the stub methods and eliminate the dead code? |
| 15:55 | <hsivonen> | Philip`: the location tracking code goes pretty intimately in the hottest method of the tokenizer |
| 15:56 | <hsivonen> | Philip`: also, I don't want too many Javaisms so that the code can be mechanically translated to C++ without having to use virtual methods |
| 15:57 | <Philip`> | Could you write two implementations of that hottest method, one without the location tracking, and choose which one to use at runtime? |
| 15:57 | <Philip`> | ...using virtual method calls |
| 15:57 | <hsivonen> | Philip`: It has crossed my mind but only recently |
| 15:57 | <Philip`> | In C++ you can just use templates - they solve all problems |
| 15:58 | <hsivonen> | "now you have two problems" :-) |
| 15:59 | <Philip`> | But now they're more fun problems |
| 16:01 | <hsivonen> | Philip`: but you are right, I could make the Tokenizer parametrized by a buffer management class using generics in Java and make the parametrization templatized in C++ |
| 16:01 | <Philip`> | (Except for the problem that compile times become huge, which is never fun) |
| 16:01 | <hsivonen> | Philip`: but then, mechanical translation to C++ already needs to mung stuff, so... |
| 16:01 | <Philip`> | Oh, I hadn't thought of generics |
| 16:02 | <hsivonen> | Philip`: no point in using generics for this expect to make it use the <foo> syntax in both languages |
| 16:02 | <Philip`> | (probably because they're boring and don't let you do anything you couldn't do before - C++ templates are much more exciting) |
| 16:05 | <hsivonen> | aargh. I have to deal with CRLF again |
| 16:12 | <hsivonen> | CRLF is such a pain |
| 16:13 | <jmb> | yes |
| 16:16 | <zcorpan_> | hsivonen: hmm. what are the intended html4-only errors? xmlns, xml:lang, />, </x in rcdata... |
| 16:17 | <zcorpan_> | unquoted attributes with special characters |
| 16:17 | <zcorpan_> | non-boolean empty attributes |
| 16:19 | <hsivonen> | zcorpan_: that seems about right, except there's an open bug (from you) about xmlns and /> |
| 16:20 | <zcorpan_> | hsivonen: yeah i know, i'm writing tests for it :) |
| 16:20 | <hsivonen> | zcorpan_: I started fixing that bug, but then this more immediate show source bug came up |
| 16:20 | <zcorpan_> | hsivonen: ok |
| 16:20 | <hsivonen> | (crazy show source stuff must be fixed while the Web remains constant, because even taking a local copy changes buffering) |
| 16:21 | <Philip`> | (Do you have a way of writing test cases that trigger buffer-handling bugs?) |
| 16:22 | <hsivonen> | Philip`: I don't. |
| 16:24 | <hsivonen> | ok. now I have the bytes-to-UTF-16 line and col counting fixed |
| 16:24 | <hsivonen> | or so I think |
| 16:25 | <hsivonen> | but an assertion still fires in show source... |
| 16:29 | <zcorpan_> | hsivonen: hmm wouldn't it be good if the tests had a pass condition? "there should be no errors" |
| 16:32 | <hsivonen> | zcorpan_: I don't understand the question |
| 16:33 | <zcorpan_> | hsivonen: wouldn't it be good if the test file said whether an error is expected or no errors are expected? e.g. <!doctype html><title></title><p>There should be no errors. |
| 16:34 | <hsivonen> | zcorpan_: yeah, that would be good |
| 16:35 | <zcorpan_> | added pass conditions |
| 19:37 | <Lachy> | I don't get why a separate wiki page had to be created on the W3C wiki http://esw.w3.org/topic/HTML/MultimediaAccessibilty |
| 19:38 | <Lachy> | it just puts the info in 2 places instead of one, and the ESW wiki sucks |
| 19:42 | <Lachy> | looks like some people still don't understand what use cases are :-( |
| 19:42 | <Dashiva> | The use case is getting your pet feature into the spec :P |
| 20:49 | <Lachy> | Using the ESW wiki, how do you markup code samples? |
| 20:54 | <Lachy> | damn, I hate the ESW wiki, when there are edit conflicts :-( |
| 20:54 | <Lachy> | in fact, I just hate the ESW wiki alogether |
| 20:55 | <Lachy> | *altogether |
| 21:09 | <smedero> | Lachy: ESW is a MoinMoin wiki I believe. |
| 21:09 | <smedero> | so I think you can do {{{inline code}}} |
| 21:09 | <smedero> | or `inline code` |
| 21:09 | <Lachy> | smedero, yeah, I found that out by looking at another page |
| 21:09 | <virtuelv> | I hate wiki syntax |
| 21:09 | <smedero> | um, blocks are |
| 21:09 | <smedero> | oh ok |
| 21:09 | <smedero> | sorry |
| 21:10 | <Lachy> | virtuelv, me too. But atleast I'm relatively familiar with MediaWiki syntax |
| 21:10 | <Lachy> | and MediaWiki allows people to edit specific sections instead of the whole page, which reduces the chance of conflicts |
| 21:17 | <Lachy> | aargh! Laura Carlson just went and removed everything |
| 21:17 | <Lachy> | I wrote on the wiki |
| 21:18 | <Lachy> | ok, well, that's the *last time* I use the ESW WIki. I'm off to add the real info the the WHATWG wiki |
| 21:30 | <Lachy> | I suppose I should have realised this would happen. The same thing happened when I once tried to fix the so-called "Use Cases" on the LongdescRetention page a long time ago |
| 21:32 | <smedero> | "Reinstaing Lachlan's lost the use case" |
| 21:32 | <smedero> | is the last change... |
| 21:32 | <smedero> | but that's not quite what you intended... |
| 21:34 | <Lachy> | well, that's good. But it's still listing descriptions of disabilities as use cases |
| 21:35 | <Lachy> | and it doesn't include the authoring and end user requirements section. It just goes straight to the proposed solutions, which starts with a requirement, not a solution |
| 21:57 | <Hixie> | http://www.w3.org/mid/48C979AD.1040904⊙ac |
| 21:57 | <Lachy> | I get internal server error from that |
| 21:57 | <smedero> | ditto |
| 22:03 | <Lachy> | Hixie, I'm guessing that email is from Shane McCarron based on the message ID, but I can't see it on any of the obvious mailing lists that he posts to |
| 22:03 | <Hixie> | www-tag |
| 22:04 | <othermaciej> | tritto |
| 22:04 | <Lachy> | http://lists.w3.org/Archives/Public/www-tag/2008Sep/0037.html |
| 22:05 | <hober> | yeah, I facepalmed when reading that. |
| 22:07 | <Lachy> | I don't understand what he's trying to say |
| 22:07 | <roc> | what am I afraid of? |
| 22:11 | <Dashiva> | commitment? |
| 22:11 | <Lachy> | roc, spiders? |
| 22:11 | <Lachy> | snakes? |
| 22:11 | <gsnedders> | me? |
| 22:11 | <Lachy> | I don't know. Why are we playing this game? |
| 22:13 | <gsnedders> | Lachy: from my reading, a CURIE is wholly equivalent to a IRI |
| 22:13 | <Dashiva> | gsnedders: Where are you reading that? |
| 22:13 | <gsnedders> | "The value space for CURIE is IRI." |
| 22:14 | <Dashiva> | That's syntactically, not semantically |
| 22:15 | <Dashiva> | (and also the problem) |
| 22:15 | <Lachy> | he may have meant semantically, since a CURIE is just a shorter, less useful, IRI abstraction mechanism |
| 22:17 | <Dashiva> | Well, the problem remains. You can't tell them apart, which is why safe CURIEs were created. |
| 22:17 | <Lachy> | I don't know what a safe CURIE is |
| 22:17 | <gavin> | roc: <othermaciej> I'm not worried about XHTML2 but I am worried about the open web platform vs. proprietary alternatives |
| 22:17 | <gavin> | <othermaciej> and I am not even so afraid of that (not as much as roc anyway) |
| 22:17 | <roc> | yeah I found the IRC logs thanks |
| 22:17 | <Dashiva> | safe_curie := '[' curie ']' |
| 22:36 | <Philip`> | "value space" is semantic, not syntactic - the lexical space is the syntax |
![http://www.geocities.co.jp/Hollywood-Spotlight/2135/2pigpigsan.jpg"](http://www.geocities.co.jp/Hollywood-Spotlight/2135/2pigpigsan.jpg"){kind=link}
![http://www.geocities.co.jp/Hollywood-Spotlight/2135/2pigpigsan2.jpg",↩](http://www.geocities.co.jp/Hollywood-Spotlight/2135/2pigpigsan2.jpg",↩){kind=link}