| 00:18 | <Philip`> | Preliminary canvas test results: Opera 9.5 scores 75.4%, Firefox 3.0 scores 67.4%, Safari 3.1 scores 76.5% |
| 00:19 | <Philip`> | s/Safari 3.1/WebKit nightly (on OS X where it doesn't crash like on Windows)/ |
| 00:21 | <othermaciej> | Philip`: where's the test suite again? |
| 00:21 | <othermaciej> | there's gotta be some low hanging fruit in those compliance bugs |
| 00:21 | <othermaciej> | especially since I think we've got most of the missing features covered |
| 00:22 | <Philip`> | othermaciej: It's currently on 192.168.2.10 |
| 00:23 | <othermaciej> | Philip`: well that I probably can't run :-) |
| 00:23 | <othermaciej> | I'm surprised the WebKit nightly isn't higher, because I vaguely recall we were close to the other browsers even when we were missing get/putImageData and getDataURL and such entirely |
| 00:25 | <Philip`> | othermaciej: There's at least one trivial fix that could make about two dozen test cases pass |
| 00:25 | <Philip`> | (on setting fillStyle = 'something invalid') |
| 00:26 | <othermaciej> | Philip`: what do we do that's wrong in that case? |
| 00:27 | <othermaciej> | I'd actually love to have these tests folded into our regression test suite |
| 00:27 | <othermaciej> | maybe I can get someone enthusiastic about doing that |
| 00:28 | <Philip`> | othermaciej: It seems to set to rgba(0,0,0,0) instead of ignoring the set |
| 00:28 | Hixie | 's mind explodes as he tries to deal with test cases that involve five distinct character encodings |
| 00:28 | <Hixie> | http://hixie.ch/tests/adhoc/uri/encoding/016.html |
| 00:28 | <gavin_> | phillip`: how big the delta is between your current test suite and the tests that were landed for mozilla earlier? |
| 00:29 | <Philip`> | othermaciej: I've already got lots of these in Mozilla's regression test suite, just by modifying the test-file-generation script to generate stuff in the expected format, which is reasonably straightforward (except my code is nasty and duplicative) |
| 00:30 | <Philip`> | (I did that for Mozilla since it seemed to be in more dire need of regression tests than WebKit ;-) ) |
| 00:31 | <othermaciej> | Philip`: for our regression tests, just about any format should do as long as you can tell failure from either a plaintext dump or from the structure of the render tree |
| 00:31 | <Philip`> | gavin_: There was already a big delta between my test suite at the time of generating the Mozilla tests, and the Mozilla tests |
| 00:32 | <othermaciej> | I think the way your tests work, it should be possible to do the check based on text along, at least if getImageData is there so there's no need to check purely visually |
| 00:32 | <Hixie> | hm, i wonder what xhr says about dealing with non-ascii characters in urls tin its arguments |
| 00:32 | <othermaciej> | *on text alone |
| 00:32 | <Philip`> | since I skipped all the ones (about a quarter?) that didn't pass in Firefox on Linux and couldn't easily be marked as 'todo' |
| 00:33 | <Philip`> | Between the test suite at that time, and the test suite now, I've forgotten how many changes I've made but it's probably quite a few :-) |
| 00:33 | Hixie | now needs annevk and hsivonen |
| 00:33 | <othermaciej> | Philip`: if we can get the test suite in the tree, even the failing ones, I bet many of the remaining tests would get fixed pretty quick |
| 00:34 | <Philip`> | othermaciej: There's about five or ten that require visual checking, but the rest can all work just with getImageData |
| 00:35 | <Philip`> | I expect a number of the tests are buggy, and many are unreasonably pedantic, so I would be surprised if anyone ever did pass them all :-) |
| 00:35 | <othermaciej> | I would definitely want either the tests or our implementation fixed in all such cases |
| 00:36 | <Philip`> | othermaciej: I probably won't have time tonight to try anything like that, but I'll see if I don't feel too lazy/bored tomorrow |
| 00:36 | Philip` | is currently just waiting for the 10.5.3 update to install, which is taking approximately forever |
| 00:37 | <roc> | 10.5.3 is evil, I'm holding out for 10.5.4 |
| 00:38 | <Philip`> | roc: Hmm, the "Stop" button is greyed out so I guess it's too late for me to be saved :-( |
| 00:38 | <roc> | does ebutler know about your test suite update? |
| 00:38 | <roc> | are you announcing it on a blog or something? |
| 00:39 | <Philip`> | I don't have a blog |
| 00:39 | <Philip`> | I haven't even finished it or uploaded it yet :-) |
| 00:39 | <Hixie> | feel free to use blog.whatwg.org |
| 00:39 | <Philip`> | s/it/the test suite/ |
| 00:40 | <roc> | he should be able to update the Mozilla suite and maybe even fix some of the bugs ... although it'll be interesting to see how many are actually cairo bugs |
| 00:41 | <Philip`> | I'm planning to run it on Windows and Linux and OS X to see how many platform-dependent bugs there are |
| 00:42 | <Philip`> | but my results table will have too many columns, so I need some way to fix that... |
| 00:42 | <roc> | zoom |
| 00:43 | <Philip`> | Nobody seems to have implemented unidirectional zoom yet, so it'll make the rows too short |
| 00:44 | <othermaciej> | roc: what's wrong with 10.5.3? |
| 00:44 | <Philip`> | While I'm waiting for the OS X updater to finish being incredibly slow, I suppose I could make a fancy dynamic column-enabler thing |
| 00:46 | <roc> | othermaciej: my officemate has some problems with it, not sure of the details. And there is the insane bug regarding VerifiedDownloadPlugin which forced us do an RC3 |
| 00:46 | <gavin_> | yeah I've heard of people having trouble with it too |
| 00:47 | <roc> | where if you just dlopen VerifiedDownloadPlugin, it "works" but then the OS hangs on shutdown |
| 00:47 | <gavin_> | I've avoided it so far |
| 00:47 | <roc> | and if you dlopen it again without a reboot in between, the process hangs in an unkillable way |
| 00:48 | <roc> | and no-one can figure out what that plugin is actually for, since it's not a real plugin |
| 00:55 | Philip` | wonders if the updater is actually doing anything |
| 00:58 | <othermaciej> | roc: I asked around, it will be fixed in the next update |
| 00:58 | <othermaciej> | roc: the plugin is for Apple's signed Dashboard widget downloads |
| 00:58 | <othermaciej> | roc: I think it affects more than just Firefox and will be fixed RSN |
| 00:58 | <roc> | ok |
| 00:58 | <roc> | thanks |
| 01:00 | <roc> | we worked around it for FF3, but it affects FF2. Not too much since most of the time we won't reregister plugins on startup, but it will happen once every FF2 update and maybe when certain extensions are installed |
| 01:15 | <Philip`> | Alas, IE+excanvas doesn't have getImageData |
| 01:15 | Philip` | clicks five hundred pass/fail buttons over a slightly laggy rdesktop connection |
| 01:24 | <Philip`> | Hixie: The commit-watchers subject lines seem to be unhelpfully truncated to lengths that assume I'm using a buggy 20-year-old mail client on a 72-character display, when actually I've got about half a thousand horizontal pixels waiting to be filled :-p |
| 01:25 | <Hixie> | heh |
| 01:25 | <Hixie> | it's 40 characters now i think |
| 01:25 | <Hixie> | what would you want? 80? |
| 01:28 | <Philip`> | Oh, maybe Gmail already truncates subjects when they're more than about 80ish characters |
| 01:28 | <Philip`> | in which case there's not enough space for great things in there :-( |
| 01:29 | <Philip`> | but if you put something like 80 in the emails then the subjects wouldn't be truncated any earlier than necessary |
| 01:29 | <Philip`> | so that would hopefully be a bit better |
| 01:29 | <Hixie> | i wonder how often i can ask microsoft a question before it becomes rude |
| 01:29 | <Philip`> | (Hooray, OS X finished updating itself) |
| 01:30 | <Hixie> | sure, i'll up it to 80 |
| 01:37 | <Philip`> | Hmm, WebKit apparently loses 1.9% when moving from 10.5.2 to 10.5.3 |
| 01:38 | <Hixie> | correctness or perf? |
| 01:38 | <Philip`> | Also, the 10.5.3 upgrade re-enables the incredibly annoying dim-the-screen-when-I-move-my-hands-near-the-keyboard misfeature |
| 01:38 | <Philip`> | Hixie: Correctness |
| 01:38 | <Philip`> | or incorrectness depending on whether my tests are correct or not |
| 01:38 | <Hixie> | fun |
| 01:39 | <roc> | could be a quartz issue |
| 01:39 | <roc> | quartz has some interesting bugs in 10.4 |
| 01:40 | <Philip`> | Hmm, I think it's partly due to nondetermistic output from the test runner |
| 01:40 | <Philip`> | which is WebKit's fault so I'll still blame them |
| 01:43 | <Philip`> | Firefox 3 gains 1.5% by moving from Linux to OS X |
| 01:46 | <Philip`> | Actually, 2.1% if I avoid being too strict on some fuzzy gradient matches |
| 01:46 | Hixie | wonders how they will "conclude" |
| 01:46 | <Hixie> | http://lists.w3.org/Archives/Public/public-html/2008Jun/0293.html |
| 01:47 | <Hixie> | (i thought i'd been pretty "concludey" already) |
| 01:51 | <Philip`> | IE+excanvas scores 31.7% |
| 01:53 | <Philip`> | (+/- some margin of error in my recording of its results) |
| 01:56 | <takkaria> | Hixie: do you plan to reply to that? |
| 01:58 | <Hixie> | not sure how to |
| 01:58 | <Hixie> | so probably not |
| 01:58 | <Hixie> | i'm happy to continue responding to the technical comments on that thread |
| 01:58 | <Hixie> | such as they are |
| 01:59 | <roc> | someone should reply and point out that it's default-deny; vendor extensions should be prefixed until they're accepted. |
| 01:59 | Hixie | begins going through the entire spec removing "URI" and "IRI" and replacing them with the pointers to the URL section |
| 01:59 | <Hixie> | roc: be my guest |
| 01:59 | <roc> | you're forcing me to join public-html aren't you |
| 02:00 | <Hixie> | yup |
| 02:00 | <Philip`> | You can post to public-html without joining it |
| 02:00 | <roc> | true |
| 02:00 | <roc> | alright |
| 02:01 | <Hixie> | yay, go roc |
| 02:01 | <roc> | now if only GMail supported mailto: URLs |
| 02:01 | <takkaria> | dunno if people have seen http://blogs.msdn.com/ie/archive/2008/06/10/introducing-ie-emulateie7.aspx |
| 02:01 | <Hixie> | roc: i thought it did with ff3 |
| 02:01 | <roc> | nah |
| 02:01 | <Hixie> | oh? |
| 02:01 | <roc> | not realy |
| 02:01 | <Hixie> | oh is the support half hearted? |
| 02:01 | <roc> | there's still no GMail URL in the right format |
| 02:02 | <Hixie> | i thought that had been fixed |
| 02:02 | <roc> | I wish |
| 02:02 | Hixie | investigates |
| 02:02 | <roc> | there's https://mail.google.com/mail/?view=cm&tf=0&to=%s |
| 02:02 | <roc> | but that just takes an address |
| 02:04 | <Hixie> | was that added for firefox? |
| 02:04 | <roc> | no |
| 02:05 | <Hixie> | hm |
| 02:05 | <takkaria> | I can only assume that concluding will happen with the spec is in LC...# |
| 02:07 | <Hixie> | Philip`: is it better now? |
| 02:07 | <Philip`> | http://philip.html5.org/tests/canvas/suite/tests/results.html |
| 02:08 | <Philip`> | Hmm, still need to test Konqueror |
| 02:11 | <Philip`> | The main conclusion of my testing: Konqueror has the coolest looking <button>s |
| 02:11 | Hixie | wonders how to define "parse a URL" |
| 02:14 | <Philip`> | http://philip.html5.org/tests/canvas/suite/tests/results.html - now with Konqueror, which is approximately tied with Opera and WebKit |
| 02:18 | <Philip`> | Hixie: That does look better |
| 02:18 | <Philip`> | (Thanks :-) ) |
| 02:20 | <Hixie> | http://www.abyssandapex.com/200710-wikihistory.html is funny as hell |
| 02:27 | Philip` | sees http://blogs.msdn.com/ie/archive/2008/06/23/securing-cross-site-xmlhttprequest.aspx |
| 02:59 | <Hixie> | oh jeez |
| 02:59 | <Hixie> | parsing a URI reference requires knowing the scheme of the base URI |
| 02:59 | <Hixie> | someone please write URI5 so i don't have to deal with this nonsense |
| 03:00 | <Hixie> | bbl |
| 06:13 | <Hixie> | does <base href=""> affect scripts? |
| 06:14 | <Hixie> | answer, yes. |
| 06:15 | <Hixie> | ...in safari. |
| 06:15 | <Hixie> | ...but not in IE |
| 06:15 | <Hixie> | nothing. is ever simple. |
| 06:51 | <mcarter> | i didn't know that URIs were so complicated |
| 07:03 | <Hixie> | man i wish the uri spec defined how to handle ill-formed uris |
| 07:04 | Hixie | tries to work out how to mangle the bnf to support random characters in a ua-compatible way without redefining the whole parsing himself |
| 07:05 | <takkaria> | epic fail, from the sounds of it :/ |
| 07:20 | <Dashiva> | [acgiowrt] (2) Define 'URL' and 'valid U [...] |
| 07:24 | <Dashiva> | Hixie: What's the [\w*] part first in the subject mean? |
| 07:25 | <Hixie> | who is affected |
| 07:25 | <hsivonen> | Hixie: you mentioned needing hsivonen and annevk |
| 07:25 | <Hixie> | hsivonen! |
| 07:25 | <Hixie> | hsivonen: http://www.whatwg.org/specs/web-apps/current-work/#urls |
| 07:25 | <Dashiva> | So each letter is one person? |
| 07:25 | <Hixie> | hsivonen: "valid URL" |
| 07:26 | <Hixie> | Dashiva: g=gecko, v=validators, t=tools, r=gears, etc |
| 07:26 | <Dashiva> | ah |
| 07:26 | <Dashiva> | e=everyone? |
| 07:26 | <Hixie> | Dashiva: anne's script uses it for icons |
| 07:26 | <Hixie> | e=editorial only |
| 07:27 | <Dashiva> | I wonder about that [] edit then :) |
| 07:27 | <Hixie> | hsivonen: do you think that's ok, or do you think we should also make things that aren't valid URIs and IRIs valid? |
| 07:27 | <Hixie> | Dashiva: it's not purely editorial, but it doesn't yet affect anyone (e.g. nobody implements that yet) |
| 07:27 | <hsivonen> | Hixie: at least now your definition of validity isn't as bad as I expected it to be |
| 07:28 | <Hixie> | hsivonen: or should we go further, and make even URIs that have %-encoded bits that aren't UTF-8 invalid? |
| 07:28 | <Dashiva> | Hixie: Now that you're doing changes to the commit-watchers, could it be possible to get source above index as well? |
| 07:28 | <hsivonen> | Hixie: I think we shouldn't, since that may be the only way to link to a resource |
| 07:29 | <Hixie> | Dashiva: no idea how to do that |
| 07:29 | <Hixie> | Dashiva: i'm just using the script that came with svn |
| 07:29 | <hsivonen> | the last bullet point that depends on the encoding of the document sucks big time considering my current impl |
| 07:29 | <Hixie> | yeah |
| 07:29 | <Hixie> | required pretty massive changes to the spec, too |
| 07:31 | <Hixie> | right now the definition is the exact definition that takes what browsers do, and only allows valid URIs and IRIs to come out of it |
| 07:31 | <hsivonen> | Hixie: do browsers always convert the path component to URI according to UTF-8? |
| 07:31 | <Hixie> | i.e. (it's the least delta from the status quo that i could do) |
| 07:31 | <Hixie> | hsivonen: as far as i can tell, yes |
| 07:31 | <Hixie> | hsivonen: and they use the document encoding for the query component |
| 07:32 | <Hixie> | which is wacked beyond words |
| 07:32 | <hsivonen> | I take it that the behavior is interoperable and changing it would break the Web? |
| 07:32 | <Hixie> | so it seems |
| 07:32 | <hsivonen> | :-( |
| 07:32 | <Hixie> | yeah no kidding |
| 07:33 | <Hixie> | http://hixie.ch/tests/adhoc/uri/encoding/012.html |
| 07:33 | <Hixie> | if you want a headache |
| 07:33 | <Hixie> | there are up to six encodings involved in that testcase depending on how you count them |
| 07:34 | <hsivonen> | I wonder if I should extend Jing to pass around the character encoding of the document or if I should do something else |
| 07:34 | <hsivonen> | leaky abstractions |
| 07:34 | <Hixie> | documents have two character encodings |
| 07:35 | <Hixie> | the original one and the one after you set .charset |
| 07:35 | <Hixie> | browsers differ on which one to use when encoding the query component |
| 07:35 | <hsivonen> | crazy |
| 07:35 | <hsivonen> | well, Validator.nu doesn't support setting .charset |
| 07:36 | <Hixie> | right now i'm ignoring the original and only using the one you set using .charset |
| 07:36 | <Hixie> | but i don't know if that'll owrk |
| 07:36 | <Hixie> | work |
| 07:36 | <Hixie> | so anyway |
| 07:36 | <hsivonen> | and after all this, someone wanted to use non-UTF-8 with XHR |
| 07:37 | <Hixie> | yeah that's what i wanted to speak to anne about -- as far as i can tell, xhr doesn't mention any of this |
| 07:37 | <hsivonen> | this an insane amount of complexity to support people who can't be bothered to use UTF-8 |
| 07:37 | <Hixie> | yes. |
| 07:44 | <Hixie> | hsivonen: anyway, does what the spec say now seem like the best compromise, or should we allow or disallow something that is currently disallowed or allowed? |
| 07:45 | <hsivonen> | Hixie: I may have a different opinion after I've thought about it more, but at the moment, it appears that the current text makes sense for authors |
| 07:45 | <Hixie> | k |
| 07:45 | <hsivonen> | although I don't like the idea of having to implement the last bullet point |
| 07:45 | <Hixie> | yeah |
| 07:45 | <Hixie> | i hear you |
| 07:46 | <Hixie> | like i said, speccing it was bad enough, i don't even want to think about what it means for you |
| 08:23 | <Hixie> | some pretty funny things happen if you visit data://example.com/ |
| 08:24 | <hsivonen> | Firefox leaks the URI to Google |
| 08:24 | <Hixie> | firefox just assumes it's a random non-uri string and does an "i'm feeling lucky" search |
| 08:24 | <Hixie> | opera is the funniest |
| 08:25 | <Hixie> | it tries to download a file "default" |
| 08:36 | <virtuelv> | http://www.bbc.co.uk/blogs/radiolabs/2008/06/removing_microformats_from_bbc.shtml |
| 08:38 | <Hixie> | yeah someone really should tell them about <time> |
| 08:42 | <virtuelv> | mwah, proprietary registration system |
| 08:45 | <Lachy> | I tried posting a comment, but it rejected it |
| 08:45 | <Lachy> | ah, it didn't like me using "<time>" because it complained about invalid XML. |
| 08:46 | <virtuelv> | Lachy: perhaps post the link plain-text? |
| 08:47 | <Lachy> | done |
| 09:21 | Hixie | begins to define how to resolve urls |
| 09:26 | <Hixie> | data:text/html;charset=iso-8859-1,<!DOCTYPE html><meta charset%3D"iso-8859-1"><script>document.write(document.charset)<%2Fscript><p><a href%3D"http%3A%2F%2Fdamowmow.com%3F%26%23x263a%3Bé">test<%2Fa> |
| 09:34 | <annevk> | Hixie, I was thinking of referencing HTML5 URL from XMLHttpRequest... |
| 09:34 | <annevk> | Hixie, hope that answers your question |
| 10:51 | <Hixie> | annevk: cool |
| 10:51 | <Hixie> | annevk: (i bet the people asking for the removal of references to html5 are going to love that! ;-) ) |
| 11:02 | <Philip`> | Hixie: "(Such scripts would not be conforming, however, as xml:base attributes as not allowed in HTML documents.)" - second "as" should be "are" |
| 11:03 | <hsivonen> | Hixie: after head is now eating all end tags, and the spec text seems to have an error in the case label ("other" without any other cases). Is it intentional to eat *all* end tags there? |
| 11:04 | <hsivonen> | (I'm changing test cases to assume all end tags get eaten) |
| 11:13 | <hsivonen> | what's the difference between having a mailing list and having a WG in IETF terms? |
| 11:13 | hsivonen | thought IETF mailing lists were effectively WGs |
| 11:15 | <gsnedders> | hsivonen: mailing lists can continue to exist after a charter has ended |
| 11:16 | <hsivonen> | (aside: I think chartering a WG to clarify but not to fix bugs is weird) |
| 11:19 | <gDashiva> | Only if you believe there are bugs :) |
| 11:20 | <hsivonen> | btw, eating end tags in after head as opposed to in head produces some unintuitive error messages |
| 11:20 | <hsivonen> | (I'm not sure which is better for comment and white space placement) |
| 11:22 | <gsnedders> | Anyone got any suggestions for learning Haskell/functional programming in general? |
| 11:23 | <hsivonen> | gsnedders: Scheme has got a nice book: SICP |
| 11:23 | <hsivonen> | Structure and Interpretation of Computer Programs |
| 11:23 | <takkaria> | ANSI Common Lisp is a fairly well-recommended book, too |
| 11:23 | <gsnedders> | I also must say I do rather suck at doing things from books: some interesting project and decent documentation is better for me |
| 11:24 | <hsivonen> | gsnedders: regarding SICP, see http://www.joelonsoftware.com/articles/ThePerilsofJavaSchools.html |
| 11:27 | <gsnedders> | I still agree with ChrisWilson's comments a while back: "[23:33] <ChrisWilson> ...AAAAAAAnd this is why I think the first round of computer classes should be taught in C. zero-based indexing, sane pointers, and plenty of rope to hang yourself with." |
| 11:27 | <gsnedders> | (plenty is probably an understatement for someone new to programming) |
| 11:28 | <takkaria> | someone made a fair point that people who learn something like Java first are going to be better at thinking in OO, I don't remember where |
| 11:28 | <Philip`> | gsnedders: I've found the best suggestion is to have some program you want to write, for which $new_language is a suitable tool - otherwise I read a few pages of tutorials and then get fed up and do something else instead |
| 11:28 | <takkaria> | I started as a BASIC programmer, then learnt C, and I still don't get the fuss over OO :) |
| 11:29 | <gsnedders> | Philip`: I'm not sure what program I want to write in it yet though :) |
| 11:29 | <Philip`> | takkaria: Sounds like a good reason to not learn something like Java first ;-) |
| 11:29 | <gsnedders> | I started in BASIC, then moved to OO-free PHP, then to OO-full PHP, then to Python |
| 11:29 | <takkaria> | Philip`: hehe |
| 11:30 | <gsnedders> | (with odd brief flirtations with various branches of C) |
| 11:30 | <MikeSmith> | people should learn XSLT first, if they really want to fug up their minds |
| 11:30 | <takkaria> | I also use Lua, which is a sufficiently different language from C that I program in it idiomatically |
| 11:30 | <Philip`> | gsnedders: Port html5lib to Haskell! |
| 11:30 | <gsnedders> | Philip`: I think that's probably a bit too big a first project |
| 11:30 | <Philip`> | (Warning: probably very bad idea) |
| 11:31 | <gsnedders> | (most of the unis in the UK still have functional programming in the first or second semester, FWIW) |
| 11:32 | <hsivonen> | takkaria: I like to think of Java as C++ minus a lot of brokenness |
| 11:32 | <Philip`> | I like to think of Java as C++ minus all the fun bits |
| 11:32 | <jgraham> | gsnedders: I started working through YAHT FWIW |
| 11:32 | <hsivonen> | Java and garbage collection is nice. C and malloc is OK. I don't like thinking of what smart pointers are partially hiding. |
| 11:33 | <takkaria> | at one point I knew the syntax of C++, but that's slipped away over the years... all I see people who start in Java/C++ doing is creating lots of unnecessary classes for things which shouldn't be classes |
| 11:33 | <Philip`> | They're not hiding anything, they're just reference counting and RAII :-) |
| 11:33 | <jgraham> | http://darcs.haskell.org/yaht/yaht.pdf |
| 11:33 | <takkaria> | well, that's a lie, I clearly see them writing a lot of widely-used code too |
| 11:33 | <gsnedders> | jgraham: as has already been said, I can't work from tutorials really |
| 11:33 | <gsnedders> | jgraham: I get a bit into it, then I just can't be bothered. Give me a project to work on! |
| 11:34 | <Philip`> | If it's Haskell, I guess it's best to have projects that don't involve IO |
| 11:34 | <jgraham> | gsnedders: I find the same, but I need to understand a little about how the language is supposed to work before I try doing stuff |
| 11:34 | gsnedders | has never read a Python tutorial in his life |
| 11:34 | <hsivonen> | Philip`: I find Objective-C's manual refcounting more tracktable than smart pointers |
| 11:35 | jgraham | only read the one on python.org |
| 11:35 | <jgraham> | Which is very short and incomplete |
| 11:35 | <hsivonen> | s/ck/c/ |
| 11:35 | <gsnedders> | Obj-C is a nice language in a lot of ways |
| 11:35 | <jgraham> | but enough to get started |
| 11:36 | Philip` | has been writing manual-refcounting code in plain C and doesn't really like it at all :-( |
| 11:37 | <hsivonen> | Philip`: did you have an autoreleasepool? |
| 11:37 | <jgraham> | gsnedders: I guess you could find a book, ignore all the text and just do the excercises |
| 11:37 | <jgraham> | ;) |
| 11:37 | <gsnedders> | jgraham: The exercises tend to be rather pointless and boring :P |
| 11:37 | <gsnedders> | (I did that with Obj-C) |
| 11:37 | <takkaria> | disagree :P |
| 11:38 | <takkaria> | I find it gets your head into the right mode of thinking |
| 11:38 | <takkaria> | but whatever floats your boat, I guess |
| 11:38 | <gsnedders> | takkaria: I've already drowned :) |
| 11:38 | <takkaria> | it's like doing logic derivations or exercise after excercise of maths, and I always found those useful |
| 11:39 | <MikeSmith> | hey, I seem to remember news about Forth being used by some Web-related project recently. Anybody else remember that? Maybe I dreamed it |
| 11:39 | <MikeSmith> | though that would be a really lame dream to have |
| 11:39 | <Philip`> | MikeSmith: Probably the Tamarin tracing thing |
| 11:39 | <doublec> | most likely |
| 11:40 | <MikeSmith> | ah yes |
| 11:40 | <MikeSmith> | doublec: yeah, was on your blog |
| 11:40 | <Philip`> | http://www.bluishcoder.co.nz/2008/05/extending-tamarin-tracing-with-forth.html etc |
| 11:40 | <MikeSmith> | yup yup |
| 11:40 | <doublec> | that'll be it :) |
| 11:42 | <Philip`> | hsivonen: No, though I don't see how that'd help |
| 11:42 | <hsivonen> | Philip`: the autoreleasepool makes refcounting not suck |
| 11:42 | <Philip`> | (possibly because I just don't understand, but possibly because the returning-refcounted-objects-from-functions thing wasn't a problem since I just return stuff with a refcount of 1) |
| 11:42 | <Philip`> | *don't understand it |
| 11:45 | <hsivonen> | Philip`: the autoreleasepool keeps retain/alloc and (auto)releasing together |
| 11:46 | <hsivonen> | I'm now again at a point where the V.nu parser passes all html5lib tokenizer and tree builder tests |
| 11:46 | <hsivonen> | time to implement the new li/dd/dt/p stuff and break tests |
| 11:47 | takkaria | is currently working on hubbub's (C html5 parser) treebuilder, implementing a chunk of it for the first time |
| 12:10 | <roc> | Is hubbub open source or an Opera thing? |
| 12:10 | <takkaria> | it's unrelated to opera entirely :) |
| 12:10 | <takkaria> | it's open-source, under the MIT licence |
| 12:10 | <takkaria> | it's being written as part of a web browser called NetSurf but is an independent library |
| 12:12 | <hsivonen> | svn: OPTIONS of '/svn/trunk/testdata/tree-construction': 502 Bad Gateway (https://html5lib.googlecode.com) |
| 12:12 | <hsivonen> | hmm. |
| 12:13 | <hsivonen> | works now |
| 12:13 | <hsivonen> | weird |
| 12:16 | <jgraham> | hsivonen: I see errors like that from googlecode occasionally |
| 12:17 | <hsivonen> | jgraham: ok. |
| 12:17 | <jgraham> | takkaria: Is the hubbub code avaliable anywhere? |
| 12:17 | <takkaria> | jgraham: http://www.netsurf-browser.org/projects/hubbub/ |
| 12:18 | <takkaria> | I hope to have the treebuilder mostly-complete by the end of next week, at which point it might actually become useful for someone |
| 12:19 | <jgraham> | takkaria: Cool |
| 12:19 | <takkaria> | the tokeniser has something like 99.3% code coverage, so I'm fairly sure it's OK :) |
| 12:19 | <takkaria> | s/code/test/ |
| 12:23 | <hsivonen> | takkaria: I see you are using function per state. Can you suspend anywhere? |
| 12:24 | <takkaria> | AFAICS, there's no need to |
| 12:25 | <takkaria> | since each state quite happily takes one character at a time |
| 12:26 | <hsivonen> | ah. I didn't look far enough to see if you had split the multicharacter spec states |
| 12:28 | <takkaria> | those which do loop in the state will return if they get an out-of-data character, so it works quite well |
| 12:33 | hsivonen | wishes the new P steps didn't have "act as if" |
| 12:34 | <hsivonen> | with the foreign content stuff, I guess I should just really run the "as if" code since inlining the behavior becomes complex |
| 13:19 | <hsivonen> | whee! the new p/li/dd/dt stuff broke test cases |
| 13:26 | <jcranmer> | new list stuff? |
| 13:27 | <hsivonen> | jcranmer: the handling of list item tags changed a while ago |
| 13:31 | <hsivonen> | now I'm confused. the old spec matched IE8 with <ul><li></li><div><li></div><li><li><div><li><address><li><b><em></b><li></ul> |
| 13:31 | <hsivonen> | is the new spec deliberately not matching or did I misimplement the new spec? |
| 13:49 | <hsivonen> | hmm. looks like I misimplemented it |
| 13:54 | <hsivonen> | now only one test doesn't pass |
| 13:54 | <hsivonen> | let's see if that's intentional... |
| 13:57 | <Philip`> | weinig: http://philip.html5.org/tests/canvas/suite/tests/results.html#2d.composite.canvas.destination-over (WebKit in 7th/8th columns) is tests for the drawImage(canvas, ...) problems I was having |
| 15:08 | <weinig> | thanks Philip` |
| 15:15 | <annevk> | http://lists.w3.org/Archives/Public/uri/2008Jun/ :/ |
| 15:18 | <annevk> | I'm not sure why he talks about HTML form parameters, but maybe I'm missing something |
| 15:30 | <annevk> | Ah, common sense prevails: http://lists.w3.org/Archives/Public/uri/2008Jun/0011.html |
| 15:51 | <gDashiva> | Deep: "W.r.t. XHTML 2.0 - it is not a big step from 1.1 - except perhaps philosophically." |
| 15:55 | <zcorpan> | gDashiva: pointer? |
| 15:56 | <gDashiva> | http://lists.w3.org/Archives/Public/www-html/2008Jun/0030.html |
| 16:21 | <gDashiva> | It's kinda refreshing to read Luca's replies, free of political restraint :) |
| 16:36 | <zcorpan> | it seems that Luca's problem is that the BP reference XHTML 1.1 Basic instead of HTML5...? |
| 16:43 | <annevk> | good laugh: http://www.w3.org/2001/tag/2008/06/19-tagmem-minutes#item04 |
| 16:48 | <annevk> | notably "I think this fork in the web is a really bad thing..." and 'document.write is too far deployed to say "don't do that"?' |
| 16:49 | <annevk> | oh, and "document.write is the actual dividing issue" and "xhtml + css is a hands-down winner with authors and to add html tag soup is pushed by the browsers who have the implementations" |
| 16:52 | <hober> | Why do people think HTML UAs are "legacy?" |
| 16:53 | <Philip`> | hober: Any code or content is legacy as soon as it is deployed :-) |
| 16:53 | <hober> | heh. |
| 16:56 | <zcorpan> | that's why xhtml2 UAs aren't legacy |
| 16:56 | <Dashiva> | (and never will be, zing) |
| 16:56 | <hsivonen> | hah |
| 17:18 | <jgraham> | sometimes I wonder if most of the TAG have ever gone to a random web page and used View Source |
| 17:21 | <Philip`> | You have to be careful how you pick a "random" page |
| 17:22 | <jgraham> | Well in lieu of a good way of selecting a rnadom web page, a selection of popular commercial sites would be fine. |
| 18:21 | <virtuelv> | sigh |
| 18:21 | <virtuelv> | people in 2008 still seem to think that XML/XHTML and namespaces is some magic bullet that frees them from every problem in the universe |
| 18:22 | <virtuelv> | http://www.reddit.com/r/programming/info/6oo5r/comments/ |
| 18:22 | <virtuelv> | I don't think I'm even going to bother replying |
| 18:23 | <Philip`> | It would be kind of bad for accessibility tools if everyone used namespaces so there were three billion possible elements and attributes that they would need to know contain human-readable text that should be presented to the user |
| 18:25 | <jgraham> | virtuelv: It seems worth replying because, on the face of it, their argument sounds sensible. The kind of argument that Philip` just made and that hsivonen makes against namespaces in HTML in principle (rather than XML namespace syntax) is rather subtle |
| 18:26 | <virtuelv> | jgraham: while it might be worth replying, I'm having one of those days where I doubt I would bother staying civilized |
| 18:26 | <Philip`> | Presumably the good solution is to use the single <abbr title> thing for everything that's acceptably close to human-readable abbreviation titles, and then use something else for the cases like hCalendar where that's just wrong |
| 18:26 | <Philip`> | and valid HTML4 doesn't provide any hooks for you to do that "something else" |
| 18:26 | <jgraham> | virtuelv: Oh well in that case maybe worth someone else replying :) |
| 18:27 | <Philip`> | so people are forced to make a bad decision |
| 18:27 | <hober> | Cooking up a good solution in #microformats |
| 18:27 | <tantek> | indeed |
| 18:28 | <Philip`> | but namespaces do allow that extensibility so they don't force you to shove inappropriate things into <abbr title> and so they allow good designs and are therefore good |
| 18:28 | <tantek> | the promise of solution via extensibility is a false promise |
| 18:28 | <tantek> | extensibility = babel |
| 18:29 | <Philip`> | Extensibility isn't a solution itself, but inextensibility sometimes prevents you from finding a solution |
| 18:30 | <Philip`> | (unless the inextensible system is sufficiently complete that it provides a solution already) |
| 18:31 | <Dashiva> | The extensibility thing reminds me of something Spolsky said abouts spec writing. |
| 18:31 | <Dashiva> | "And now certain geeks go off to a very dark place where they start thinking about automatically compiling specs into programs, and they start to think that they’ve just invented a way to program computers without programming." |
| 18:32 | <jgraham> | Dashiva: I thought that was what RDF was for... |
| 18:33 | <Dashiva> | Yeah, RDF is awesome, you just have to download the internet to resolve all the references |
| 18:34 | <Philip`> | Dashiva: That's not a problem - just delegate the job to Google |
| 18:44 | <tantek> | I've come to realize that XML should really expand to "eXperimentation Markup Language" - as in, XML may be good for experiments, for explorations, but not for actual solutions. |
| 18:48 | <virtuelv> | well, if someone decides to reply to the xhtml apologists in that reddit thread, why not just make it a blog post, which can be discussed independently? |
| 18:49 | <Philip`> | virtuelv: That loses the immediacy of arguing directly in the comment thread, and most people will be too lazy to read the blog post |
| 20:30 | <takkaria> | hmm, the TAG don't really think document.write() will ever go away, do they? |
| 20:31 | <gsnedders> | takkaria: Oh, probably they do think it will. |
| 20:31 | <takkaria> | I was hoping DanC wouldn't think that, really... |
| 20:34 | <Philip`> | The need to use document.write() when writing new content may go away |
| 20:34 | <Philip`> | which seems a much more realistic goal |
| 20:35 | <Philip`> | and arguably a useful one since document.write() is crazy and confusing and so it'd be nice if alternatives were provided so that all use cases could be solved without document.write() |
| 20:36 | <hsivonen> | Philip`: there's no way browser vendors would pull document.write() support, though, so it needs to be specified and implmented |
| 20:36 | <Philip`> | hsivonen: Indeed |
| 20:36 | <jgraham> | Philip`: Authors seem to like document.write which suggestss that they just avoid the crazy bits |
| 20:37 | <hsivonen> | yeah. It's not too crazy when there's only one level of it |
| 20:37 | <jgraham> | (much as I dislike document.write, I guess it is one of the most successful parts of the DOM in terms of usage) |
| 20:38 | <Philip`> | There's the "I don't want to switch my site to XHTML because it'll break all my (ad provider's) document.write code" which could be solved if more people didn't think document.write was a useful/necessary feature |
| 20:38 | <Philip`> | *the "..." argument |
| 20:39 | <Philip`> | which is relevant in terms of the everyone-should-gradually-migrate-to-XHTML view of the future |
| 20:39 | <jgraham> | You seem to have misse out the word dystopian somewhere |
| 20:40 | jgraham | decides it is time to go home |
| 20:41 | <Philip`> | Hmm, now I remember why it's a bad idea to provide test results |
| 20:41 | <Philip`> | I want to make some minor updates to my tests, but I don't want the published results to be out of sync with the tests, but I don't want to run the tests in ten different browsers again :-( |
| 20:42 | <Philip`> | I guess I'll leave the updates offline for now, until I get non-lazy enough to do another iteration of the whole process |
| 20:48 | <Hixie> | hsivonen: yeah, it's intentional to eat all end tags, and "all other" is just for consistency with the rest of the spec, though i guess i should make that clear. i've noted your e-mail on the subject. |
| 20:52 | <hsivonen> | Hixie: ok. thanks. Did you notice that deferring tag eating until after head (as opposed to in head) makes error situations a bit weird? |
| 20:52 | <Hixie> | i noticed you noticing it, i don't recall it being intentional |
| 20:53 | <hsivonen> | Hixie: ok. might be worth fixing if it doesn't cause badness with comments and whitespace. (I haven't investigated if that is the case.) |
| 20:53 | <Hixie> | you sent mail right? |
| 20:53 | <hsivonen> | I didn't. |
| 20:53 | <Hixie> | ah ok |
| 20:53 | <Hixie> | send mail :-) |
| 20:54 | <Hixie> | right now i'm so far from caring about hte parser section that it's not even funny :-) |
| 20:54 | <Hixie> | too deeply stuck in the mud that is uris |
| 20:57 | <hsivonen> | Hixie: I filed a bug. |
| 20:58 | <Hixie> | that works too |
| 20:58 | <hsivonen> | Hixie: should I send mail to whatwg about SVG <defs><font> or does the remark in the middle of a public-html email count? |
| 21:01 | <Philip`> | It's already in http://canvex.lazyilluminati.com/misc/cgi/issues.cgi/message/%3C14B846F4-08C1-4A5D-8E19-B4ECCF25122F%40iki.fi%3E |
| 21:01 | <hsivonen> | thanks |
| 21:05 | <Hixie> | hsivonen: the remark counts |
| 21:05 | <Hixie> | hsivonen: however i don't know what to do about it |
| 21:06 | <Hixie> | hsivonen: the svg <font>s overwhelm the html <font>s in my data, so visual inspection doesn't work |
| 21:06 | <Hixie> | hsivonen: and i'm not sure how else to distinguish them |
| 21:06 | <Hixie> | ooh, maybe attributes |
| 21:07 | <hsivonen> | I was thinking that a <defs> parent would reduce the probability of <font> being a cargo cult HTML <font> |
| 21:08 | <Hixie> | right but i still have to test it |
| 21:11 | <Philip`> | What the children of <font>? If a child is an HTML element or plain text, then presumably it's an HTML <font>; otherwise the children are probably SVG elements |
| 21:11 | <Philip`> | *What about the |
| 21:11 | <hsivonen> | hmm. looks like valid SVG can have <font> outside <defs> :-( |
| 21:12 | <hsivonen> | Philip`: making a decision based on children would seriously suck in the tree builder |
| 21:12 | <hsivonen> | Philip`: because you haven't seen the children when you need to deal with <font> |
| 21:12 | <hsivonen> | inspecting attributes seems like a much better idea |
| 21:13 | <Philip`> | hsivonen: The <font> could always be interpreted as an SVG <font>, and the tree builder could break out of SVG mode once it's realised a child is wrong |
| 21:13 | <Philip`> | which would slightly break the formatting of the HTML content, since the <font> would no longer be an HTML <font> |
| 21:14 | <Philip`> | but that's only a minor formatting problem, not a make-the-whole-rest-of-the-page-invisible problem |
| 21:17 | tantek | wonders if Google Adsense provides an XHTML-friendly way to include Adsense ads in pages (as opposed to the document.write methods). And if not, is Hixie talking with those folks? |
| 21:19 | Philip` | imagines they might have concerns about what to do when people use the XHTML-AdSense inside text/html pages |
| 21:20 | <Philip`> | and I guess they'd prefer not to serve code for both HTML and XHTML insertion and dynamically choose between them, since that makes the ads bigger and slower |
| 21:21 | <hsivonen> | IIRC, the conspiracy theory was that Google couldn't figure out how to make AdSense work with XHTML, so they hired Hixie to write HTML5 and kill XHTML |
| 21:21 | <Philip`> | and the risk of confusing the tens of percent of people who think they're using XHTML, just to benefit the zero percent who are really using XHTML, doesn't seem great :-( |
| 21:21 | <hsivonen> | (I don't recall the source of the conspiracy theory) |
| 21:23 | <hsivonen> | Philip`: presumably, a single piece of code could assemble ads in both XHTML and HTML and use if (document.createElementNS) to decide whether to use createElementNS or createElement |
| 21:24 | <Philip`> | hsivonen: That code would be more bytes, so the ads would be bigger and slower |
| 21:25 | <Philip`> | Only a tiny bit, but a tiny bit multiplied by billions of views per day makes the viewers unhappy |
| 22:24 | roc | wonders why CSS 2.1 says that 'overflow' doesn't apply to table row groups, when the Web says it does |
| 22:33 | jgraham__ | wonders why roc is surprised at a disconnect between specs and reality |
| 22:34 | <jgraham__> | Did someone already point out http://mochaui.com/demo/ It claims to be a GUI toolkit that uses lots of <canvas> for rendering |
| 22:35 | <hsivonen> | is it accessible? |
| 22:36 | <roc> | that's pretty cool |
| 22:36 | <roc> | It's great to see canvas and SVG usage increasing |
| 22:38 | <jgraham__> | hsivonen: Well it seems to use real text at least. But I doubt the interactivity is exposed to ATs |
| 22:39 | <roc> | I discovered that the Dirac codec documentation uses SVG and MathML very nicely |
| 22:39 | <roc> | http://dirac.sourceforge.net/documentation/algorithm/algorithm/quantisation.xht |
| 22:39 | <hsivonen> | hmm. Why does mochaui work in Firefox 3 and Opera 9.5 but not in a WebKit nightly? |
| 22:42 | <jgraham__> | roc: Oh, that's pretty. I'm so used to seeing crappy Latex2HTML gifs for maths on the web that it comes as a bit of a surprise that Maths-on-the-web doesn't have to be eye-bleedingly ugly |
| 22:42 | <hsivonen> | yeah, it's cool to see MathML usage on the Web |
| 22:43 | <Philip`> | That MathML is kind of horribly ugly and broken in Opera, though |
| 22:44 | <Philip`> | (partly since it doesn't understand , and partly since things like the floor symbol don't get stretched vertically at all) |
| 22:44 | <roc> | you really have to author specifically for the CSS-MathML subset in Opera, I think |
| 22:45 | <Philip`> | (Also the v-tilde has a whole line-height of space in its middle) |
| 22:46 | <Philip`> | Is this something Opera could fix via CSS, or would they have to really implement MathML? |
| 22:46 | <roc> | I'm too lazy to figure that out rightnow |
| 22:47 | <Philip`> | Fair enough :-) |
| 22:47 | <roc> | huh, Webkit doesn't support scrollable table row gorups |
| 22:47 | roc | wonders why so many people file bugs about them against Gecko, then |
| 22:48 | <roc> | neither does Opera |
| 22:51 | <jgraham__> | roc: The HTML 4 spec suggests that UAs might implement scrollable table row groups |
| 22:54 | <hsivonen> | wishful thinking from many years ago: http://hsivonen.iki.fi/misc/table-scroll/ |
| 22:54 | <roc> | wishful thinking? that would work in Gecko |
| 22:55 | <hsivonen> | roc: I wasn't aware. With what CSS? |
| 22:55 | <hsivonen> | no action in https://bugzilla.mozilla.org/show_bug.cgi?id=53702 |
| 22:56 | <roc> | see this example https://bugzilla.mozilla.org/attachment.cgi?id=310440 |
| 22:56 | <roc> | and ignore the background-painting bug |
| 22:57 | <hsivonen> | roc: cool! do the theads/tfoots of nested tables stack as in my illustration when a nested table is showing? |
| 22:57 | <roc> | no, nothing special there |
| 22:59 | <Hixie> | http://lists.w3.org/Archives/Public/www-html/2008Jun/0036.html is special |
| 23:00 | <Hixie> | "we work with browser developers" "which ones?" "well i hope we work with browser developers" |
| 23:08 | <gsnedders> | browsers? peh! |
| 23:08 | <gsnedders> | Google is the future! |