| 00:21 | <jgraham> | Hixie: I can't see any obvious errors in the rest of the algorithm, but I didn't rigorously check everything, particularly the places where you changed the constraints so x_current always <= x_width and similarly with y_current |
| 00:25 | <Hixie> | k |
| 00:25 | <Hixie> | as always, let me know if you run into problems ever :-) |
| 00:46 | <gsnedders> | Can someone get me held back in school, just so I can stay in Secondary 5? |
| 07:46 | <hsivonen> | http://lists.w3.org/Archives/Public/www-tag/2008Mar/0033.html |
| 08:07 | <hsivonen> | http://www.dehora.net/journal/2008/03/20/html5-obs/ |
| 08:26 | <hsivonen> | http://lists.w3.org/Archives/Public/www-tag/2008Mar/0058.html |
| 08:27 | <hsivonen> | http://lists.w3.org/Archives/Public/www-tag/2008Mar/0061.html |
| 09:09 | <svl> | To see "Aunt Tilly" and "simply deploy RDF documents" in the same sentence... |
| 09:10 | <hsivonen> | regarding UTF-8 as a token default alongside windows-1252: it seems that people who use UTF-8 have a clue about declaring it |
| 09:58 | <hsivonen> | Philip`: did you examine the behavior of getProbableCharsets() in jchardet? Is the *last* entry in the array always the most probable? |
| 10:04 | <hsivonen> | hmm. I guess I should just let the detection fail if UTF-16 is the most probable |
| 10:04 | <hsivonen> | instead of trying to take the next most probable |
| 11:50 | <hsivonen> | Hixie, annevk: the discussion in this Mozilla bug might be of interest: https://bugzilla.mozilla.org/show_bug.cgi?id=289938 |
| 12:29 | <virtuelv> | I believe annevk is on vacation |
| 12:31 | <hsivonen> | right. I forgot |
| 12:40 | <hsivonen> | http://wilbur.bytowninternet.com/pipermail/list_html4all.org/2008-March/000660.html |
| 13:16 | <BenMillard> | hi all, Simon 'Zcorpan' Pieters told me Hixie had gotten to table issues |
| 13:17 | <BenMillard> | he linked me to the tracker and the recent IRC chat |
| 13:24 | <MikeSmith> | hsivonen - have you tested install of your conformance checker on OSX? |
| 13:27 | <MikeSmith> | virtuelv - I really wish there were some way I could easily install the latest opera snapshot/weekly without needing to the goddamn desktop team blog page to check and manually get the link |
| 13:27 | <virtuelv> | MikeSmith: you want build feeds? |
| 13:33 | <MikeSmith> | Dashiva - I was going to say to virtuelv that I really don't care what the mechanism is as long as I've got some automated way to install the test build when I want to |
| 13:35 | <hsivonen> | MikeSmith: I develop on Mac OS X, but I haven't tested a clean install on OS X in a while |
| 13:37 | <MikeSmith> | hsivonen - OK, just wondering. I've been messing around a bit lately with shell stuff under Darwin on OSX and just wondering if there were any gotchas |
| 13:37 | <MikeSmith> | one thing I noticed so far is that it's not really obvious what JAVA_HOME is |
| 13:38 | <MikeSmith> | but eventually managed to figure out it's /System/Library/Frameworks/JavaVM.framework/Versions/CurrentJDK/Home |
| 13:49 | <BenMillard> | I've told Accessify Forums (as "Cerbera") about the tables section of HTML5 changing. I hope it's a fair summary: http://www.accessifyforum.com/viewtopic.php?p=59125#59125 |
| 14:16 | csarven | finds http://csseleven.com/ lame on so many levels |
| 14:38 | <hsivonen> | ouch. the spec has grow past the Validator.nu size limit again |
| 14:40 | <zcorpan_> | Hixie: while you're working on tables, i think the table processing should look at child <col> elements of <table> |
| 14:40 | <zcorpan_> | (and that it should be allowed and perhaps that the html parser shouldn't imply <colgroup> -- i've sent email about this if i recall correctly) |
| 14:41 | <hsivonen> | zcorpan_: re: attributes defaulted from a DTD: If the user chooses DTD processing, Validator.nu processes the DTD. If that sucks for the user, the users shouldn't ask for DTD processing |
| 14:41 | <zcorpan_> | hsivonen: fair enough :) |
| 14:41 | <hsivonen> | zcorpan_: however, I have considered cloning Gecko's DTD behavior and offering that as an option |
| 14:42 | <hsivonen> | but I don't want to do it unless there's agreement that Gecko's DTD behavior should be grandfathered into the Web platform |
| 14:58 | <Philip`> | hsivonen: I haven't tried using getProbableCharsets at all |
| 14:59 | <hsivonen> | Philip`: ok |
| 15:00 | <hsivonen> | csarven: is there any point other than self-promotion with pictures and minibios? |
| 15:01 | <jgraham_mibbit> | BenMillard: I think tha's a fair summary although I guess Hixie can come up with an entirely novel algorithm |
| 15:03 | <csarven> | hsivonen My thoughts exactly! |
| 15:16 | <BenMillard> | jgraham_mibbit, thanks and that's a possibility zcorpan mentioned. |
| 17:30 | <Hixie> | hsivonen: if you want to let the html4all people know, there will be a gigantic e-mail coming out relatively soon discussing all the table edits in great detail |
| 17:30 | <Hixie> | hsivonen: i just didn't want to delay the edits until i'd done editing the spec |
| 17:31 | <Hixie> | er, until i'd done replying to the e-mail, i mean |
| 18:54 | <Philip`_> | http://wiki.whatwg.org/wiki/Special:Recentchanges - Mpsingh looks like a spammer |
| 19:24 | <andersca> | Hixie: found a tiny typo in the spec |
| 19:24 | <andersca> | Explicit entries can also be marked as foreign, which means that they have an manifest attribute but that it doesn't point at this cache's manifest. |
| 19:24 | <andersca> | an manifest => a manifest |
| 19:30 | <hsivonen> | Hixie: mentioned as part of replying on html4all |
| 19:48 | <andersca> | Hixie: also, ping |
| 19:48 | <andersca> | :) |
| 19:53 | Pavlov_ | summons hixie too, for good measure |
| 19:54 | <andersca> | cool |
| 20:25 | <Hixie> | andersca, Pavlov_: here |
| 20:26 | <Hixie> | andersca: and sorry to be a pain, but if you could, please mail the typo to me :-) (either the list or ian⊙hc if you don't want to spam the list) |
| 20:26 | <andersca> | Hixie: I would rather have all those typos exposed on the whatwg list ;) |
| 20:27 | <Hixie> | sure thing :-) |
| 20:27 | <andersca> | Hixie: also, I was wondering why null characters in the manifest file should be converted to fffd? |
| 20:28 | <Hixie> | to avoid any problems with implementors wanting to use null-terminated strings |
| 20:29 | <Hixie> | it's been a source of many problems in other parts of the web platform |
| 20:30 | <andersca> | ap: did you see that |
| 20:31 | <ap> | Hixie: the problems in other parts were fixed without changing decoding of nulls - I think it would be better to have a consistent approach |
| 20:31 | <Hixie> | the html5 parser handles nulls in much the same way |
| 20:31 | <ap> | Hixie: hmm, do any browsers follow that? |
| 20:32 | <ap> | Hixie: for all I know, WebKit just treats nulls as normal characters |
| 20:32 | <Hixie> | some do. others treat them as nulls, and yet others have security bugs (they just strip out the nulls, or truncate at nulls, both of which can be used for xss or similar behaviours) |
| 20:33 | <Hixie> | to be honest it really doesn't matter much what we do with nulls here, since none of the magic strings contain nulls, and URIs aren't allowed to contain nulls. |
| 20:33 | <Hixie> | (or fffds) |
| 20:33 | <Hixie> | i don't really understand why it's a problem |
| 20:33 | <Hixie> | so long as the behaviour is indistinguishable from treating nulls as FFFDs |
| 20:33 | <ap> | Hixie: I'd just prefer to avoid introducing yet another mode in the decoder |
| 20:34 | <Hixie> | the spec doesn't say anything about how you implement it |
| 20:34 | <ap> | Hixie: hmm, I think andersca was talking about this phrase from the spec: "All U+0000 NULL characters must be replaced by U+FFFD REPLACEMENT CHARACTERs." |
| 20:35 | <Hixie> | right, so long as the result is equivalent to this, it doesn't matter how you do it |
| 20:35 | <Hixie> | in fact i think that line is effectively redundant, so long as you correctly detect whether URIs are valid or not |
| 20:36 | <othermaciej> | why not just wait for Opera to fix their null character handling bugs |
| 20:36 | <Hixie> | since both FFFD and 0000 are invalid in URIs |
| 20:36 | <ap> | Hixie: ok, I guess I didn't have enough context, and thought that this had practical consequences |
| 20:36 | <andersca> | oh, sorry about that :( |
| 20:36 | <Hixie> | othermaciej: this isn't about those bugs |
| 20:36 | <Hixie> | othermaciej: but the null->fffd stuff in the parser is there because nulls have caused far many more problems than just in opera |
| 20:37 | <Hixie> | e.g. the status bar bug several years back |
| 20:37 | <Hixie> | and since there really is no use case for U+0000 on the Web, it seems far safer to just remove it altogether |
| 20:37 | <othermaciej> | nulls are certainly risky when you mix with C APIs |
| 20:37 | <othermaciej> | but converting them in some places but not others is an annoying burden |
| 20:37 | <ap> | Hixie: sounds like a warning could have better effect than a somewhat misguiding requirement in a sample algorithm |
| 20:38 | <othermaciej> | and you really want your code to be safe against nulls in strings anyway |
| 20:38 | <Hixie> | ap: send mail :-) |
| 20:38 | <ap> | Hixie: k |
| 20:38 | <Hixie> | thanks |
| 20:38 | <othermaciej> | but if the conversion doesn't affect results of the algorithm then it does not matter much in this case |
| 20:39 | <Hixie> | othermaciej: in the manifest case, i don't think there's any noticeable effect, and we can probably remove it |
| 20:39 | <Hixie> | othermaciej: in the html parser case, i agree that it is unfortunate that you can still introduce nulls through the DOM APIs |
| 20:39 | <Hixie> | othermaciej: not much i can do about that, though |
| 20:40 | othermaciej | blames the Unicode consortium |
| 20:40 | <Hixie> | othermaciej: still seems far safer and better to make nulls become FFFDs, since you already have to have code to turn a number of other things into FFFDs and have to do newline normalisation anyway |
| 20:40 | <Hixie> | anyway |
| 20:40 | <Hixie> | please discuss this on a mailing list :-) |
| 20:40 | Hixie | is working on tables |
| 20:42 | <gsnedders> | eww! you can't use tables! they're presentational! divs ftw! |
| 20:46 | hsivonen | likes \0 to FFFD substitution in the parser |
| 20:52 | <BenMillard> | hixie, is there anything I can be doing to help your work on tables? |
| 20:58 | Philip` | sees the last paragraph of http://html4all.org/mailman/archives/list_html4all.org/2008-March/000663.html and conducts some research |
| 20:58 | <Philip`> | Out of about 130,000 random pages from dmoz.org, the 'axis' attribute is used on http://www.cdsm.co.uk/ and http://centerville.lionwap.org/ and http://gazianteplionsclubtr.lionwap.org/ |
| 20:58 | Philip` | concludes that it's quite rare, and that he has insufficient data to conclude anything else |
| 20:59 | <hsivonen> | Philip`: I assume you keep a local cache of 130000 pages? |
| 20:59 | <Philip`> | hsivonen: Yes |
| 21:00 | <hsivonen> | how much disk space does the cache take? |
| 21:00 | <Philip`> | I wouldn't want the cost of downloading ~3GB of content every time I wanted to find a little more information :-) |
| 21:00 | <Philip`> | hsivonen: ~3GB |
| 21:00 | <Philip`> | (having clamped each page to 256KB) |
| 21:01 | <Philip`> | (since the (unclamped) mean size of a page is infinite) |
| 21:02 | <Philip`> | If disk space was a concern, I imagine compression could help significantly |
| 21:05 | <Philip`> | (http://triin.net/archive/kool/webstat/figure-8.png suggests almost everything is <=256KB, so that seems a reasonable limit to impose) |
| 21:05 | <Philip`> | (though actually I picked the value pretty arbitrarily) |
| 21:07 | <hsivonen> | If I have understood axis correctly (not a safe assuption), http://gazianteplionsclubtr.lionwap.org/ uses axis in a useless way |
| 21:07 | <hsivonen> | as does http://centerville.lionwap.org/ |
| 21:08 | <hsivonen> | and http://www.cdsm.co.uk/ |
| 21:09 | jgraham_ | is pretty sure axis is useless |
| 21:12 | <Philip`> | At least it's abused far less frequently than many other attributes |
| 21:24 | BenMillard | agrees with jgraham |
| 21:29 | <BenMillard> | I'm off for some supper, some TV and then bed. bye all |
| 21:29 | <hsivonen> | ye |
| 21:29 | <hsivonen> | bye even |
| 21:30 | <BenMillard> | cya |
| 22:11 | <jgraham_> | http://james.html5.org/graph_2.svg -> possibly all the state transitions made in the html5lib treebuilder testcases (assuming I didn't screw up) |
| 22:12 | jgraham_ | will work out how to distinguish "real" transitions from "reprocess as if in" transitions next (that graph should have both) |
| 22:13 | <Hixie> | how do you go from AfterBodyPhase to InFramesetPhase? |
| 22:14 | <Hixie> | oh that's the "process using the rules for" thing |
| 22:14 | <Hixie> | nevermind |
| 22:14 | <Hixie> | there are iirc only four insertion modes that can be subject to that, btw |
| 22:14 | <Hixie> | if that helps optimise code |
| 22:49 | <Philip`> | jgraham_: That's quite nice :-) |
| 22:52 | <Philip`> | <text style="font-family:Times New Roman;font-size:14.00;"> - doesn't it need to quote the string? |
| 22:53 | <Hixie> | no |
| 22:53 | <Philip`> | Oh |
| 22:53 | <Hixie> | it does, however, need units on the font-size |
| 22:54 | <Philip`> | Oh |
| 22:54 | <hsivonen> | Hixie: in SVG? |
| 22:54 | <Hixie> | yup |
| 22:54 | <Hixie> | unless you use the font-size="" attribute |
| 22:55 | <hsivonen> | in retrospec, it might have been good to make unitless default to px in CSS even in the standards mode |
| 22:55 | <Hixie> | except that would clash with unitless numbers in line-height, font, and other properties |
| 22:55 | <hsivonen> | unitless line-height is redundant with % |
| 22:56 | <Hixie> | nope |
| 22:56 | <Hixie> | they inherit differently |
| 22:56 | <hsivonen> | too difficult :-) |
| 22:56 | <Hixie> | (numbers allow lines to expand to fit children with bigger font sizes, %s don't) |
| 22:56 | <Hixie> | i'm not defending the design of the language |
| 22:57 | <Hixie> | just explaining why the svg group got it wrong |
| 22:59 | <Philip`> | http://www.gerv.net/security/link-fingerprints/ - "half a loaf is better than no bread" - I think that's not true in general - the half loaf might be covered in mould, and it'd be better if you didn't have any bread at all |
| 23:00 | <Hixie> | i think we already discussed fingerprints on the list once |
| 23:00 | <hsivonen> | http://mindforks.blogspot.com/2008/03/aria-templateid-explained.html |
| 23:00 | <Hixie> | and came to the conclusion it wasn't a good idea |
| 23:00 | <Hixie> | i was hoping someone would find the reference and reply to him before i had to :-) |
| 23:00 | <hsivonen> | I wonder how aria-templateid scales on the time axis given the update cycle of AT |
| 23:00 | <hsivonen> | compared to the update cycle of Web 2.0 apps |
| 23:00 | Hixie | wonders what happens if he sets aria-templateid="google.com/gmail"; on his blog |
| 23:01 | <othermaciej> | what's aria-templateid? |
| 23:01 | <Hixie> | <hsivonen> http://mindforks.blogspot.com/2008/03/aria-templateid-explained.html |
| 23:02 | Philip` | tends to download files by copying the URL and pasting it into wget, and then getting bash errors and doing it again with quotes around the URL argument, which would avoid the benefits of link hashing |
| 23:02 | <othermaciej> | that sounds like an amazingly bad idea |
| 23:02 | <Hixie> | othermaciej: you hate blind people! |
| 23:03 | <hsivonen> | othermaciej: you are not the first person with that initial reaction |
| 23:04 | <takkaria> | I can't imagine AT vendors actually implementing support for it |
| 23:04 | <Philip`> | Hixie: If you did that, it would make your blog less usable to some people with AT, and there are much easier ways you can cause problems for those people, so I don't see why it matters |
| 23:05 | <Hixie> | Philip`: it seems like reducing the ways that authors can (accidentally) screw over their users would be a good thing |
| 23:06 | <Hixie> | Philip`: and, similarly, increasing the number of ways that authors can screw over their users would be a bad thing |
| 23:06 | <Philip`> | Why would someone accidentally set aria-templateid="google.com/gmail"; on their site? |
| 23:06 | <othermaciej> | the saddest thing about it is that it seems to presume that web sites themselves won't use ARIA to give a good enough accessibility experience, so that AT needs special-case hacks for different sites |
| 23:06 | <Hixie> | Philip`: why would someone accidentally put <!DOCTYPE PHP PUBLIC "-//W3C//DTD PHP 4.01//EN"> as their doctype? |
| 23:06 | <Hixie> | othermaciej: firevox shows that this is already happening |
| 23:07 | <Hixie> | othermaciej: though why a site would care enough to put an aria-templateid attribute but not enough to make their site usable, i dunno |
| 23:07 | <Dashiva> | I find the AT upgrade cycle a much more persuasive argument :) |
| 23:08 | <othermaciej> | we have indeed heard many times that the brutally slow AT upgrade cycle makes changing accessibility features fruitless |
| 23:08 | <Philip`> | Hixie: There's an infinite number of errors like that that people can accidentally introduce; but there's only a (pretty small) finite number of aria-templateid="..." strings that any AT would care about, so there's a vastly smaller chance of that happening by accident |
| 23:08 | <othermaciej> | for instance that a new table header association algorithm was worthless, since it would be forever until AT supports it, and twice forever until users who need it have that version |
| 23:09 | <Hixie> | Philip`: except that the sites that would use it are those most likely to be held up as examples to learn from |
| 23:09 | <Hixie> | othermaciej: way to rain on what i'm doing on my other monitor :-P |
| 23:10 | <Philip`> | aria-templateid allows competition in (e.g.) the Gmail-usability space, because lots of vendors (and extension developers etc) can all try to make it work as well as possible; otherwise only Google could improve Gmail usability, so they're going to do less well than when there are many competing attempts |
| 23:11 | <Hixie> | nonsense |
| 23:11 | <othermaciej> | Hixie: I'm not saying I agree with such sentiments |
| 23:11 | <Hixie> | Philip`: ATs could easily just hardcode the gmail uris |
| 23:11 | <Dashiva> | Philip`: Replace templateid with the URL and you have a user script system that works today |
| 23:12 | <Philip`> | Maybe Google Maps is a better example, since that often gets embedded in other web pages and you can't hardcode the page URLs |
| 23:12 | <Philip`> | (Yay mashups!) |
| 23:13 | <Hixie> | that might hypothetically work, if templateid is on a per-subtree level, but that would need extremely careful definition |
| 23:14 | <hsivonen> | (the visual part of Google maps is very visual though) |
| 23:14 | <Hixie> | horrah, ms feedback on storage |
| 23:15 | <Philip`> | I suppose the ARIA spec is a little vague on aria-templateid, but maybe someone will blog about it in an extremely careful detailed way |
| 23:15 | <Hixie> | man, i should use that model to write html5 |
| 23:15 | <Hixie> | that would be so much easier |
| 23:16 | <Philip`> | hsivonen: (There are many disabled people who aren't blind, so AT improvements on primarily visual content would still help some users) |
| 23:16 | <hsivonen> | Philip`: you are aware of the wiki guide, right? |
| 23:17 | <Philip`> | hsivonen: I'm not, and Google just tells me about a science fiction manga |
| 23:17 | <Philip`> | Oh, is that the Mozilla one? |
| 23:17 | <hsivonen> | Philip`: http://developer.mozilla.org/en/docs/ARIA_UA_Best_Practices |
| 23:18 | <Philip`> | "aria-templateid: XXX not sure what to do with this one" |
| 23:18 | <Philip`> | Hmm |
| 23:22 | Philip` | just sees http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2006-November/thread.html#7825 about the link fingerprint thing |
| 23:23 | <Hixie> | that's the one |
| 23:26 | <Hixie> | thanks Philip` |
| 23:28 | Philip` | thanks the "Search Mail" button |
| 23:33 | Hixie | wonders how strange a validator feature request would have to be before hsivonen would just reject it out of hand |
| 23:33 | <Hixie> | hsivonen: so i'm tempted to ask for support for an equivalent of the #line preprocessor directive... |
| 23:33 | Philip` | wonders what happens if the target of checksummed link is replaced by an HTML file containing a malicious script followed by enough stuff to make a browser want to incrementally render before it's finished downloading and can check the checksum |
| 23:33 | <Hixie> | (i edit a file that is then concatenated to another one before validation, so the line numbers are all 82 lines off for me) |
| 23:34 | <Hixie> | crap, gotta go |
| 23:34 | <Hixie> | if people want to check it out, i just regenned the spec with the new algorithm for tables |
| 23:34 | <Hixie> | bbl |
| 23:36 | <csarven> | Are there any UAs that support inline pseudo-classes? |
| 23:38 | <andersca> | Hixie: cool |
