Options:
- # Session Start: Wed Apr 16 00:00:00 2008
- # Session Ident: #html-wg
- # [00:01] * Quits: adele (adele@17.203.14.240) (Ping timeout)
- # [00:14] * Quits: adele_ (adele@17.203.14.214) (Quit: adele_)
- # [00:38] * Quits: heycam (cam@124.168.30.80) (Quit: bye)
- # [00:41] * Joins: adele (adele@17.203.14.214)
- # [00:44] * Quits: adele (adele@17.203.14.214) (Ping timeout)
- # [00:49] * Joins: adele (adele@17.203.14.214)
- # [00:59] * Quits: adele (adele@17.203.14.214) (Quit: adele)
- # [01:04] * Joins: mjs (mjs@17.203.15.181)
- # [03:13] * Disconnected
- # [10:08] * Attempting to rejoin channel #html-wg
- # [10:08] * Rejoined channel #html-wg
- # [10:08] * Topic is 'HTML WG chat http://www.w3.org/html/wg/tracker (logs: http://krijnhoetmer.nl/irc-logs/ ) '
- # [10:08] * Set by DanC_lap on Mon Mar 10 03:08:44
- # [10:41] * Disconnected
- # [10:41] * Attempting to rejoin channel #html-wg
- # [10:41] * Rejoined channel #html-wg
- # [10:41] * Topic is 'HTML WG chat http://www.w3.org/html/wg/tracker (logs: http://krijnhoetmer.nl/irc-logs/ ) '
- # [10:41] * Set by DanC_lap on Mon Mar 10 03:08:44
- # [11:10] * Attempting to rejoin channel #html-wg
- # [11:10] * Disconnected
- # [11:10] * Attempting to rejoin channel #html-wg
- # [11:10] * Rejoined channel #html-wg
- # [11:10] * Topic is 'HTML WG chat http://www.w3.org/html/wg/tracker (logs: http://krijnhoetmer.nl/irc-logs/ ) '
- # [11:10] * Set by DanC_lap on Mon Mar 10 03:08:44
- # [11:13] <MikeSmith> aaronlev, you there?
- # [11:13] <chaals> MikeSmith: back from reading miles of ranting and ill-informed rubbish about alt...
- # [11:14] <aaronlev> hi MikeSmith
- # [11:14] <MikeSmith> chaals, there's some money in rubbish. ask the Mafia
- # [11:14] <aaronlev> chaals: the problem is when there is an a11y issue that everyone can understand
- # [11:14] <chaals> indeed, part of the idea is that if you have too much alt text then it starts to be painful and counter-productve - at which point you should shift it out of the main flow. longdesc does that - in HTML5 currently there isn't anything equivalent to that.
- # [11:14] <aaronlev> it's an opportunity to argue endlessly
- # [11:14] <MikeSmith> aaronlev, hey, was wondering something about AT apps
- # [11:15] <aaronlev> but ther eare so many more issues
- # [11:15] <chaals> aaronlev: or even "almost understand" :(
- # [11:15] <aaronlev> heh
- # [11:15] <aaronlev> chaals: you're almost better than coffee
- # [11:15] * Quits: Lachy (Lachlan@88.91.106.102) (Quit: This computer has gone to sleep)
- # [11:15] <MikeSmith> aaronlev, do you know, is there some technical reason why an AT app couldn't by default speak just the alt text that is below some particular user-configurable limit?
- # [11:16] <MikeSmith> and then prompt the user to choose whether or not to speak any alt text longer than that?
- # [11:16] <MikeSmith> e.g., provide a generated hyperlink to any alt text longer than whatever limit the user has chosen?
- # [11:16] <chaals> hsivonen: it isn't JAWS in particular - that just happens to be the screen reader most Americans are familiar with (and a few people in countries where the government buys JAWS as the answer to all blind folks' problems, like Finland and Denmark.
- # [11:17] <chaals> it isn't even screen readers in particular, although they are the clearest consumer (since people forget that search engines have to be engineered rather than just being some amazing black magic) so the thing that discussions center around.
- # [11:17] <aaronlev> MikeSmith: i'm missing the context i haven't keept up with the alt discussion
- # [11:17] <aaronlev> as far as I know, alt="" means don't speak anyting for the image, because it is decorative or something
- # [11:17] <chaals> MikeSmith: no technical reason, but it would be pretty un user-friendly
- # [11:17] <aaronlev> no alt means it is not provided and there is an opportjunity to repair
- # [11:17] <aaronlev> and anything else means, this is the text to speak for the image
- # [11:17] <MikeSmith> aaronlev, there is a WCAG draft test that sets a limit of 100 chars on alt content
- # [11:18] <aaronlev> that's a bullshit old requirement
- # [11:18] <aaronlev> for some older version of JAWS
- # [11:18] <MikeSmith> well, some WCAG checkers still seem to flag it
- # [11:18] <aaronlev> let me ask marco
- # [11:18] <chaals> having to listen to a half-sentence, and then figure out when you want to hear the rest, is not exactly conducive to concentrating on the flow of 600wpm from a crappy voice synthesizer
- # [11:18] <MikeSmith> http://www.w3.org/WAI/GL/WCAG20/tests/test3.html
- # [11:18] <aaronlev> chaals: but JAWS has fixed that for a while right?
- # [11:19] * chaals believes so.
- # [11:19] <chaals> I would ask a jaws user...
- # [11:19] <aaronlev> we're at JAWS 9 now
- # [11:19] <chaals> which reminds me...
- # [11:19] <aaronlev> the oldest versions i hear people are still using are like JAWS 6
- # [11:19] <aaronlev> generally
- # [11:19] <MikeSmith> chaals, why would it be any more user-unfrienly than a longdesc link
- # [11:20] <MikeSmith> what I mean to say is that as far as the end-user experience, it would be the same
- # [11:20] <MikeSmith> except that the longdesc link would be generated by the AT app
- # [11:20] <MikeSmith> instead of provided by the author
- # [11:20] <chaals> MikeSmith: because a half-sentence that cuts off throws your concentration, whereas some signal that there is a description available can fit into the flow of what you get.
- # [11:20] <MikeSmith> ah
- # [11:20] <MikeSmith> I see
- # [11:20] <MikeSmith> yeah, I can see that
- # [11:21] <hsivonen> chaals: I think government contracts where the buyer is very different from the user lead to economic incentives that lead to less usable software
- # [11:21] <hsivonen> not only for AT but for different kinds of software
- # [11:21] <chaals> true as a general principle of economics.
- # [11:22] <aaronlev> MikeSmith: longdesc allows rich content
- # [11:22] <aaronlev> you can have a table for example, which is the alternative for a chart
- # [11:22] <aaronlev> but aria lets you do that anyway
- # [11:22] <chaals> (i.e. also applies to roads and methadone programs)
- # [11:22] <MikeSmith> aaronlev, If you know anybody at Freedom Scientific or at other AT vendors, can you suggest how we might be able to get them to participate directly on public-html discussions around accessibility issues?
- # [11:22] <hsivonen> aaronlev: speaking of which, is there a plan to move to exposing emphasis and links in aria-describedby referent?
- # [11:24] <aaronlev> hsivonen: i don't understand the question
- # [11:25] <aaronlev> MikeSmith: they won't spend resources on that but may be willing to engage in specific questions
- # [11:25] <aaronlev> we sometimes get Glen Gordon to participate in an ARIA discussion
- # [11:25] <aaronlev> MikeSmith: why does it have to be freedom sci
- # [11:25] <aaronlev> there are other screen reader developers
- # [11:27] * Joins: Lachy (Lachlan@213.236.208.22)
- # [11:29] <MikeSmith> aaronlev, Freedom Scientific because they seem to have big market share and because some guidelines in WCAG seem to be getting produced in order to work around specific limitations in JAWS
- # [11:29] <aaronlev> even in WCAG 2?
- # [11:29] <aaronlev> i would have thought it's removed in wcag 2
- # [11:30] <MikeSmith> http://www.w3.org/WAI/GL/WCAG20/tests/test3.html is part of "HTML Test Suite for WCAG 2.0"
- # [11:31] <MikeSmith> Last-Modified: Thu, 11 Aug 2005 18:19:51 GMT
- # [11:31] <hsivonen> aaronlev: I mean exposing the stuctured content instead of plain string for describedby
- # [11:31] <aaronlev> hsivonen: we expose it both ways
- # [11:32] <hsivonen> aaronlev: ah ok
- # [11:32] <aaronlev> hsivonen: we provide accDescription which is plaintext, and an accessible relation, which is basically a pointer to another place in the DOM tree
- # [11:32] <aaronlev> however, if the text is display:none we end up just providing the plaintext, unfortunately
- # [11:33] <aaronlev> so putting using ARIA to provide an alternative table for a chart would not work well at this pt
- # [11:33] <aaronlev> you'd have to do something ugly like place it offscreen
- # [11:45] <chaals> MikeSmith: if it is a warning, that you quite possibly have too much stuff in alt and should check it, then it's probably not a bad test. If it is a hard requirement then it's rubbish.
- # [11:46] <chaals> [[[This test case was created by the ATRC at the University Of Toronto. It should not be construed as required for conformance with the proposed WCAG2.]]] is a bit of a hint though...
- # [11:47] <chaals> [[[Procedure
- # [11:47] <chaals> For each img element, calculate the length of text within the alt attribute value.
- # [11:47] <chaals> If the Alt text is greater than 100 characters (English) then it must be shortened or the user must confirm that it is the shortest Alt text possible.]]]
- # [11:47] <chaals> if it is the shortest alt possible, that's reasonable. Sometimes alt needs to be long. But most times it is a sign of bad alt text.
- # [11:48] <MikeSmith> it just seems arbitrary
- # [11:48] <MikeSmith> like I said, ideally it should be up the user to choose
- # [11:48] <MikeSmith> for them to choose what the consider excessively long
- # [11:48] <MikeSmith> they consider
- # [11:50] <chaals> it is arbitrary. Blame ATRC at the University of Toronto for having arbitrary ideas.
- # [11:51] <chaals> ideally the author would know what is too long and what is good alt text, and be qualified to make that choice. The user is going to have a hard time adapting to each author unless you have a huge semweb-backed system for sharing information about who writes excessive amounts of drivel, who writes moderate amounts of drivel, and who writes short crpytic comments that need to be expanded by reference to other stuff they write...
- # [11:52] <MikeSmith> yep
- # [11:53] <anne> if you're not complaining about long paragraphs complaining about long alt= values is silly
- # [11:53] <aaronlev> heck, in german there are some words that are 40 letters long :)
- # [11:55] <aaronlev> give davidb a hard time when he's on #wai-aria later today
- # [11:55] <MikeSmith> I think the requirement should be that the author and checker have to time how long it actually takes to say it aloud.
- # [11:55] <aaronlev> he's at ATRC
- # [11:55] <aaronlev> he likes getting flack
- # [11:55] <aaronlev> :)
- # [11:56] <MikeSmith> I'm good at flack.
- # [11:56] <MikeSmith> I should at that to my CV.
- # [11:57] <aaronlev> i want to add street hustling to mine
- # [11:57] <MikeSmith> I wonder what character limits they specify Japanese and Chinese
- # [11:57] <MikeSmith> pimpin', mackin'
- # [11:57] <MikeSmith> and slackin'
- # [11:59] <MikeSmith> one character in Japanese can often be two or more spoken syllables
- # [11:59] <MikeSmith> 魚
- # [11:59] <MikeSmith> sakana
- # [12:01] <chaals> prolly 100 characters. That's what the text says.
- # [12:02] <chaals> (The ATRC guys used to want to make hard limits and one of the things I pointed out was that different langauges work differently like that.)
- # [12:03] <chaals> It is true that having alt text too long is counter-productive. A *description* of a complex image isn't generally useful in the flow of text - you should be able to skip over it or read it, based on a short functional bit of text.
- # [12:04] <MikeSmith> chaals, see also the "Language Specific" table there
- # [12:04] <MikeSmith> which says 115 for ger and 90 for kor
- # [12:05] * MikeSmith wonders why they hell the author of that page uses "eng" "ger" and "kor" abbreviations in that table
- # [12:05] * chaals is busy and doesn't go looking
- # [12:06] * Joins: Dashiva (noone@80.202.220.46)
- # [12:07] <MikeSmith> 90-character limit is not consistent with 100-character limit for English
- # [12:08] <MikeSmith> wondering what process if any they actually used to arrive at those limits
- # [12:08] <chaals> MikeSmith: stop worrying and do something useful :P
- # [12:10] <anne> hard limits make for easy programming
- # [12:10] <anne> (unless you define it in vague terms like "characters" and try to implement it correctly :) )
- # [12:12] <Dashiva> It's an arbitrary limit, so trying to make sure it's "correct" seems like misplaced effort
- # [12:15] * Quits: Dashiva (noone@80.202.220.46) (Ping timeout)
- # [12:15] <anne> it's called being an asshole: http://diveintomark.org/archives/2004/08/16/specs
- # [12:16] * Joins: Dashiva (noone@80.202.220.46)
- # [12:19] <MikeSmith> heh, "Virtually every useful tutorial in the world was written by a moron-turned-expert."
- # [12:21] <Dashiva> I wonder who wrote the rest, assholes or sociopaths
- # [12:22] <anne> I'm not sure I agree with angels and work of fiction though. hsivonen seems to fit that description. He even feeds his output back into the spec. (Though that also matches asshole somewhat :-) )
- # [12:30] * Quits: Dashiva (noone@80.202.220.46) (Quit: Dashiva)
- # [12:33] * Joins: Dashiva (noone@80.202.220.46)
- # [13:02] * Quits: aaronlev (chatzilla@217.228.202.109) (Quit: ChatZilla 0.9.81 [Firefox 3.0pre/2008041306])
- # [13:14] * Joins: aaronlev (chatzilla@217.228.202.109)
- # [13:20] * Joins: Sander (svl@86.87.68.167)
- # [13:21] * Quits: aaronlev (chatzilla@217.228.202.109) (Quit: ChatZilla 0.9.81 [Firefox 3.0pre/2008041506])
- # [13:21] * Joins: aaronlev (chatzilla@217.228.202.109)
- # [13:50] * Joins: myakura (myakura@122.29.8.215)
- # [14:05] * Joins: zcorpan (zcorpan@88.131.66.80)
- # [14:24] * Quits: myakura (myakura@122.29.8.215) (Ping timeout)
- # [14:31] <Philip> http://dev.w3.org/2008/mobile-test/canvas.js - is canvii a legitimate pluralisation?
- # [14:34] <Philip> It sounds kind of like the -i suffix is only used for certain Latin nouns imported into English, and canvas has too remote an etymology from Latin for that to apply...
- # [14:34] <tlr> also, -i wouldn't be the latin plural for it...
- # [15:19] * Joins: matt (matt@128.30.52.30)
- # [15:19] * Joins: myakura (myakura@122.29.8.215)
- # [15:51] * Quits: matt (matt@128.30.52.30) (Quit: matt)
- # [15:52] * Joins: matt (matt@128.30.52.30)
- # [16:15] * Quits: fearphage (fearphage@66.68.52.63) (Ping timeout)
- # [16:15] * Joins: fearphage (fearphage@66.68.52.63)
- # [16:27] * Joins: billmason (billmason@69.30.57.200)
- # [16:45] * Joins: aroben (aroben@71.58.57.150)
- # [16:51] <gsnedders> Philip: the Latin plural can be used for most words taken directly from Latin. It's just people try and use it on words that aren't Latin too :)
- # [16:54] * Quits: matt (matt@128.30.52.30) (Quit: matt)
- # [16:54] * Joins: matt (matt@128.30.52.30)
- # [16:58] * tlr observes there's more than one latin plural
- # [16:58] <anne> there's 5
- # [16:59] <Philip> Only 5 Latin words are plurals?
- # [16:59] <gsnedders> Philip: No, there are five singular forms and five plural forms of most Latin nouns
- # [16:59] <Philip> Oh
- # [17:00] <gsnedders> Only one case is commonly used in English, though
- # [17:00] <Philip> Sounds like quite an overengineered language
- # [17:00] <tlr> philip, you should look at Polish or Greek...
- # [17:00] <gsnedders> Hungarian has a large number of cases, no?
- # [17:00] <tlr> also, I wasn't talking about cases, but declinations
- # [17:01] <gsnedders> tlr: ah
- # [17:01] * gsnedders is dumb
- # [17:01] <gsnedders> there are seven cases :P
- # [17:01] <gsnedders> and five declensions.
- # [17:02] <tlr> whooops, you caught me with a Germanism there.
- # [17:02] <gsnedders> Did I?
- # [17:02] <gsnedders> Oh.
- # [17:03] <tlr> I said "declination" instead of declension
- # [17:03] * tlr off for a conf call
- # [17:04] * Quits: Lachy (Lachlan@213.236.208.22) (Quit: This computer has gone to sleep)
- # [17:11] * Joins: mjs_ (mjs@64.81.48.145)
- # [17:11] * Quits: mjs (mjs@64.81.48.145) (Connection reset by peer)
- # [17:21] <DanC> The s/n ratio in the alt discussion is pretty low. Chaals seems to be keeping the issue tracker mostly up to date. http://www.w3.org/html/wg/tracker/issues/31
- # [17:21] <DanC> If html5 conformance were kept to machine-checkable constraints, I might be able to help the discussion some. But as it is, I don't see much I can do.
- # [17:24] <DanC> hmm... I'm not alone... "For example, HTML 5 seeks to make layout tables non-conforming, which
- # [17:24] <DanC> I think it is an exercise in futility."
- # [17:24] <DanC> -- <http://www.w3.org/mid/212F8C99-9310-4C63-85C9-903514DD9BD1@iki.fi>
- # [17:25] <anne> machine checkable conformance is a subset of conformance that HTML5 does acknowledge somehow
- # [17:25] <DanC> oh?
- # [17:25] <zcorpan> DanC: that doesn't mean that hsivonen doesn't want non-machine-checkable conformance criteria in general
- # [17:25] <anne> (that too)
- # [17:25] <DanC> sure
- # [17:26] <anne> "Conformance checkers are exempt from detecting errors that require interpretation of the author's intent"
- # [17:26] <DanC> hsivonen, do you want any non-machine-checkable conformance criteria?
- # [17:27] <Philip> anne: That sounds slightly vague, and makes it hard to produce a specific list of the conformance criteria that a conforming conformance checker must check
- # [17:27] <DanC> it also makes the label "conformance checker" pretty misleading
- # [17:28] <zcorpan> can a human be a conformance checker?
- # [17:28] <anne> maybe it should differentiate between automated conformance checkers and human conformance checkers :)
- # [17:28] <anne> but maybe it's not a big deal
- # [17:29] <Philip> anne: What about people with cyborg brains?
- # [17:30] <aaronlev> MikeSmith: what did you think about the answer from the JAWS folks, about long alt text?
- # [17:30] <anne> Philip, why do you assume they're mutually exclusive?
- # [17:30] <DanC> chaals, are you available to attend tomorrow's telcon? (1700Z, I think). I'd like your help with the alt issue
- # [17:31] * Joins: Lachy (Lachlan@88.91.106.102)
- # [17:31] <DanC> no, 1600Z http://www.timeanddate.com/worldclock/fixedtime.html?month=04&day=17&year=2008&hour=16&min=00&sec=0&p1=0
- # [17:32] <Philip> anne: I don't assume that; there's just a continuous spectrum between "automated" and "human", so if you want to differentiate between them you'd have to define a specific cut-off point
- # [17:33] <anne> Philip, that'd be part of their definitions
- # [17:34] <Philip> anne: It might not be possible to define it precisely
- # [17:35] <DanC> anne, do you know what part of the planet chaals is on today/tomorrow?
- # [17:35] * aaronlev is now known as aarondinner
- # [17:35] <anne> I know he's going to China at some point (for the AC)
- # [17:36] <anne> so no :)
- # [17:38] <gsnedders> Do you know whether chaals will be on the planet?
- # [17:47] * Quits: chaals (chaals@84.77.0.120) (Connection reset by peer)
- # [17:48] * Quits: beowulf (beowulf@208.113.221.22) (Quit: leaving)
- # [17:49] * Joins: beowulf (beowulf@208.113.221.22)
- # [18:10] * aroben is now known as aroben|lunch
- # [18:10] * Joins: dbaron (dbaron@67.160.251.228)
- # [18:17] * Quits: myakura (myakura@122.29.8.215) (Quit: Leaving...)
- # [18:34] * Quits: zcorpan (zcorpan@88.131.66.80) (Ping timeout)
- # [18:40] * Joins: adele (adele@17.203.14.240)
- # [19:03] * Quits: mjs_ (mjs@64.81.48.145) (Quit: mjs_)
- # [19:07] <anne> Is William F Hammon, signing with Bill, really suggesting text/html might have two code paths? One for HTML and one for XML?
- # [19:07] <anne> I thought we were getting somewhere in the discussion...
- # [19:09] * Philip was about to reply with why doctype sniffing for XHTML is not the bestest idea
- # [19:09] <Philip> (About two thirds of XHTML-doctype pages are not even well-formed)
- # [19:11] <gsnedders> That sounds far too low
- # [19:12] <gsnedders> Then there's the issue that the requirements for UAs for XHTML 1.0 apply to more than just strictly conforming documents, i.e., those without doctypes too
- # [19:15] * Quits: matt (matt@128.30.52.30) (Quit: matt)
- # [19:24] <anne> Philip, that might be cool if he questions my "The Web" reply :)
- # [19:24] <anne> (the more interesting part of my reply was the link I found somewhere on my own site to the HTML WG decision of September 2000)
- # [19:36] * Joins: hober (ted@68.101.220.172)
- # [19:37] <Philip> gsnedders: You can complain to the libxml developers if you think it should be finding well-formedness errors on more pages than it is :-p
- # [19:37] <gsnedders> Philip: or your sample size isn't big enough :P
- # [19:38] <gsnedders> (how big is it, FWIW?)
- # [19:38] <gsnedders> I remember Hixie quoting a far large figure, I think, from one of his huge looks at billions of documents
- # [19:39] <Philip> gsnedders: A hundred, with "XHTML 1.0 Strict" in them somewhere (probably the doctype)
- # [19:39] <Philip> I don't think Hixie has ever said how many XHTML-doctyped pages are well-formed
- # [19:39] <gsnedders> Philip: I'd think that's probably too small of a sample
- # [19:39] <Philip> gsnedders: I wouldn't think it is
- # [19:40] <Philip> or at least it's not too small to make imprecise claims like "two thirds"
- # [19:41] <Philip> Standard deviation is something like sqrt(n*p*(1-p)) = 5 so it's like (64+/-10)% of the dmoz.org XHTML 1.0 Strict pages being ill-formed
- # [19:43] <gsnedders> 10% is fairly significant, though
- # [19:43] <Philip> Depends how much you care about accuracy
- # [19:43] <Philip> and in this case I don't care much
- # [19:43] <gsnedders> I'm an asshole :P
- # [19:43] <gsnedders> What do you expect me to be like?
- # [19:44] <Philip> so I'm happy to just know that it's between a half and three quarters, and I'm happy to call that two thirds
- # [19:46] <Philip> Anyway, this is the internet, I can just make up facts here - nobody's going to bother confirming my claims
- # [19:49] <gsnedders> :P
- # [20:03] * Joins: matt (matt@128.30.52.30)
- # [20:07] * Quits: heycam (cam@124.168.30.80) (Ping timeout)
- # [20:13] * Joins: heycam (cam@124.168.100.30)
- # [20:23] * Joins: mjs (mjs@17.255.104.166)
- # [20:23] * Joins: mjs_ (mjs@17.203.15.181)
- # [20:26] * Quits: mjs (mjs@17.255.104.166) (Ping timeout)
- # [20:35] * aroben|lunch is now known as aroben
- # [20:39] <hsivonen> DanC: I think I want non-machine-checkable criteria on the level of "authors must used elements and attributes only according to their specified semantics", but I think stipulating very specific non-machine-checkable criteria risks becoming dead letters of the spec.
- # [20:41] <hsivonen> DanC: as for the conformance checker label, I think it should be a given that a software labeled "conformance checker" can only check machine-checkable criteria
- # [20:42] <hsivonen> So that part can't be expanded. But non-machine-checkable criteria could be narrowed if having non-machine-checkable criteria is considered problematic
- # [20:42] <hsivonen> the above goes for HTML5
- # [20:43] <DanC> hmm
- # [20:43] <mjs_> perhaps conforming to machine-checkable criteria and conforming to all criteria should be given different labels
- # [20:43] * mjs_ is now known as mjs
- # [20:43] <hsivonen> for WCAG 2.0, having mostly non-machine-checkable criteria comes with the territory and I think trying to forcibly cast it as machine-checkable is bad
- # [20:44] <mjs> (for example, "valid" or "syntactically conforming" on the one hand and "fully conforming" on the other)
- # [20:44] <DanC> re "authors must use elements only according to their semantics", I tend to look at it another way: if an author writes <p>, and the spec say <p> means paragraph, then readers get to conclude that the author meant paragraph.
- # [20:44] <hsivonen> I'd like "valid" to mean "conforms to machine-checkable criteria"
- # [20:45] <mjs> having a subjective conformance class seems problematic
- # [20:45] <Hixie> DanC: there are two issues, what the reader can assume, and what the author meant.
- # [20:45] <hsivonen> DanC: in practice though, the reader can assume less than what Hixie stipulates in the spec
- # [20:46] <Hixie> DanC: the reader can certainly assume that <cite> meant "title of work", but if the author actually used it to mean "quote", then the document isn't conforming and the reader isn't going to be getting useful results from his assumptions.
- # [20:46] <Hixie> DanC: but there's little that can be done to machine-check that case
- # [20:46] <DanC> I don't see much relevance to what the author meant, I guess. the spec is the community agreement about what the language means. If I write "$200" in the amount field of a check but say "oh; I meant $20", too bad for me. $200 is coming out of my account.
- # [20:46] <hsivonen> Did anyone file a bug on OO.o for that one yet?
- # [20:47] <Hixie> DanC: sure, but there's still an error, as the person losing $180 would presumably emphatically agree
- # [20:47] <Hixie> DanC: and it'd be the error of the author
- # [20:47] <DanC> right, but it's the author's error, not a problem with the format of the check
- # [20:47] <Hixie> DanC: right
- # [20:47] <Hixie> DanC: in html5 terms, there's a non-machine checkable conformance error
- # [20:48] <DanC> I don't see why authoring errors should be considered conformance errors
- # [20:48] <Hixie> i don't understand how they could not :-)
- # [20:48] <hsivonen> DanC: as a practical matter, I complain less when Hixie makes non-machine-checkable stipulations unrealistic in my opinion :-)
- # [20:48] <Hixie> what else can a conformance error be?
- # [20:48] <DanC> it seems more valuable to the community that conformance should be a property of a document alone, independent of the author's intent.
- # [20:49] <Hixie> if the author screws up, there's an error, whether or not you can tell by inspecting the document out of context
- # [20:49] <Hixie> that seems obvious to me
- # [20:50] <DanC> yes, but it's not clear that it's useful for the spec to bother with that class of error
- # [20:50] <DanC> do you really want the same sequence of characters to be sometimes conforming and sometimes not, depending on what the author meant?
- # [20:50] <hsivonen> I think machine-checkable errors should be decidable from Content-Type plus finite byte stream
- # [20:51] <hsivonen> (i.e. no dependencies on external image resources, etc.)
- # [20:51] <hsivonen> or the base URI
- # [20:51] <Hixie> danc: if i write <p>And they Jerry said <cite>Oh noes</cite>.</p>, that's obviously an error, and as an author i care that that is an error. it's valuable to me to know that it's an error (e.g. if someone points it out to me) so that i can fix it and get better behaviour out of reader's clients.
- # [20:51] <Hixie> readers'
- # [20:52] <Hixie> DanC: conformance is something primarily useful to authors, not the community
- # [20:52] <Hixie> it's a way of determining if you will get readers to interoperate with you
- # [20:52] <Hixie> and your intent
- # [20:52] <hsivonen> DanC: as for the alt issue at hand, I think it doesn't belong in the machine-checkable syntax definition and therefore doesn't belong in the validation function
- # [20:53] <hsivonen> but since there clearly is demand for alt inspection in a validation service, I'm addressing the issue in the Validator.nu UI outside the validation function
- # [20:53] <DanC> wild. I strongly disagree, Hixie. conformance is absolutely a market-maker, i.e. a community asset.
- # [20:54] <mjs> I think as long as machine-checkable conformance has a decent name that people can refer to, it still provides that value
- # [20:54] <hsivonen> it seems to me that in the case of Web specs, error handling is a market maker
- # [20:54] <Hixie> DanC: how so?
- # [20:55] <mjs> however, if semantics are to be inferred, I think it has to be defined what semantic uses of elements are correct, and therefore any use that doesn't match that is some kind of error
- # [20:55] * Quits: dbaron (dbaron@67.160.251.228) (Quit: 8403864 bytes have been tenured, next gc will be global.)
- # [20:55] <hsivonen> I agree that abusing defined semantics is some kind of error even though non-machine-checkable by necessity
- # [20:57] <DanC> I'm stumped. I can't seem to unroll the notion that conformance is a community asset. it's so obvious to me that I can't figure out why I think it's true.
- # [20:59] <hsivonen> DanC: I think the internal Microsoft emails about cloning Lotus and Novell software (from court) documents were spot on
- # [20:59] <hsivonen> DanC: error handling behavior causes lock-in
- # [20:59] <hsivonen> or complex internal behavior in general
- # [21:00] <DanC> complexity brings lock-in, yes, that's pretty clear
- # [21:00] <Philip> Putting lots of work into your product makes it hard to clone
- # [21:00] <hsivonen> I, therefore, conclude that commoditizing error handling is a more important market-maker than handling the non-erroneous case
- # [21:00] <Philip> because the cloner has to put in lots of work too
- # [21:01] <Philip> which kind of sounds entirely obvious and unavoidable
- # [21:01] <mjs> not all kinds of work are equal in this regard
- # [21:02] <mjs> the same amount of effort will not necessarily add the same amounts of behavioral complexity
- # [21:02] <Hixie> indeed
- # [21:02] <mjs> (Apple for instance is known for spending a lot of effort to add as little behavioral complexity as possible)
- # [21:02] <DanC> especially if the 1st implementor achieved complexity thru *lack* of effort
- # [21:02] <hsivonen> DanC: conformance as a market maker seems to work when it can push the lock-in aspect of secret-sauce error handling to insignificance
- # [21:02] <mjs> and yes, much observed behavioral complexity is accidental
- # [21:03] <hsivonen> when a significant proportion of the document population is non-conforming, commoditized error handling becomes the market maker
- # [21:04] <hsivonen> I don't believe HTML5 validators can shift the proportions enough to bring us to the case where almost everything is conforming and conformance is the market maker
- # [21:04] <hsivonen> so I try to position V.nu as a tool that reduces author bewilderment and saves their time
- # [21:04] <DanC> I think HTML is a special case in that respect
- # [21:05] <hsivonen> as opposed to a tool that makes the Web conforming
- # [21:06] <anne> DanC, you don't think the same goes for HTTP and CSS?
- # [21:06] <anne> DanC, or EMCAScript (might be harder to check)
- # [21:06] <DanC> no. with http, if you send the wrong bytes, you get no joy
- # [21:06] <hsivonen> HTTP is less crazy but still crazy
- # [21:06] <DanC> also, there are no web designers hand-coding http headers
- # [21:07] <Hixie> HTTP has a ton of error handling stuff going on
- # [21:07] <hsivonen> DanC: actually, with PHP, there are
- # [21:07] <Hixie> it's just not visible when you view source
- # [21:07] <Hixie> and there are _plenty_ of bogus headers out there
- # [21:07] <Philip> DanC: I do 'print "Content-Type: text/html\n\n"' a lot
- # [21:07] <hsivonen> Philip: that should be \r\n!!!
- # [21:08] * hsivonen seriously dislikes CRLF
- # [21:08] <anne> yeah, HTTP has a lot of error handling too DanC
- # [21:08] <Philip> hsivonen: I don't know how Apache munges my headers, so I just do whatever I feel like as long as it works
- # [21:08] <DanC> the standards process around CRLF is broken indeed, hsivonen
- # [21:08] <anne> hsivonen, right, CRLF is violated all over I think... I noticed that on my homepage I mix all sorts of line endings :)
- # [21:09] <DanC> "I just do what I feel like as long as it works" is what we can expect from most parties.
- # [21:09] <Philip> http://www.rusit.de/ typos Content-Type as ContentType but still seems to work
- # [21:09] <DanC> ?!
- # [21:11] <DanC> can I go back to my little world where stuff actually works like it's supposed to now?
- # [21:11] <anne> Philip, interesting, maybe more than just Content-Type is supported :)
- # [21:12] <Hixie> Philip: probably due to content sniffing
- # [21:12] <Hixie> Philip: rather than ContentType being supported
- # [21:12] <Hixie> DanC: probably not if you want to chair this group :-)
- # [21:12] <Philip> Oops, my mistake
- # [21:12] <Philip> It sends both Content-Type and ContentType
- # [21:13] <mjs> HTTP is pretty crazy
- # [21:14] <mjs> in practice you have to do a lot of error handling, and in fact have to handle responses with no headers at all
- # [21:14] <mjs> so it is indeed the case that almost any random byte sequence will do something
- # [21:14] <mjs> though perhaps not what you intended
- # [21:14] <mjs> (and I'm not sure there are any popular HTTP Validator services for servers)
- # [21:15] * Joins: dbaron (dbaron@63.245.220.241)
- # [21:16] <Philip> http://philip.html5.org/data/http-headers.txt
- # [21:17] <Philip> I can't actually see many common ones that look like typos
- # [21:17] <Philip> Content_Length, I suppose
- # [21:18] <Philip> "http" is indicative of a certain bug
- # [21:18] <Philip> and "Character set '#30' is not a compiled character set and is not specified in the 'c" indicates a somewhat different bug
- # [21:19] <Philip> (That header has value "\mysql\share\charsets\Index")
- # [21:20] <hsivonen> Philip: are all X-Bender headers from diveintomark.org or are there others using it?
- # [21:21] <Philip> hsivonen: All the ones I saw are Slashdot and csdaily.com
- # [21:21] <hsivonen> oh. interesting
- # [21:54] * Quits: tlr (tlr@128.30.52.30) (Quit: tlr)
- # [22:00] * Joins: tlr (tlr@128.30.52.30)
- # [22:03] * anne has a hard time following http://www.w3.org/2008/04/16-xhtml-minutes.html#item03 but suspects it's somehow relevant to "us"
- # [22:03] * Joins: Zeros (Zeros-Elip@129.2.175.74)
- # [22:14] <Philip> "TBL surprised to hear no one used XHTML; more than 50% of top 20 web sites using XHTML"
- # [22:15] <Philip> When I last looked, of the Alexa top 100, 35 used XHTML and 28 of those were not well-formed XML
- # [22:16] <Philip> Aha - http://lists.w3.org/Archives/Public/public-html/2007Aug/1248.html
- # [22:17] <Philip> (Summary: 67 of top 200 claimed XHTML; 51 ill-formed; ampersands are hard)
- # [22:17] <Philip> and only one site in the top 500 serves application/xhtml+xml to browsers like Firefox
- # [22:18] * Quits: ROBOd (robod@89.122.216.38) (Quit: http://www.robodesign.ro )
- # [22:20] * Quits: ChrisWilson (cwilso@131.107.0.73) (Ping timeout)
- # [22:26] * Joins: ChrisWilson (cwilso@131.107.0.102)
- # [22:35] <Zeros> Philip, "using xhtml" is very vague
- # [22:36] <Zeros> many many websites "use XHTML" just because it's "newer" and the false notion that the /> gains you something
- # [22:36] <Zeros> I can even count the number of developers I've talked to that had no notion of mime types of XHTML at all
- # [22:37] <Zeros> books don't talk about it, tutorials don't talk about, schools don't talk about it
- # [22:37] <aarondinner> Zeros: w3.org pages generally got that wrong, i don't know if they've corrected it finally
- # [22:37] <aarondinner> they used text/html for XHTML pages
- # [22:39] <Zeros> s/can/can't/
- # [22:39] <aarondinner> nope, it's still an issue even on the xhtml2 main page
- # [22:39] <Zeros> do they really send a xml mime now?
- # [22:39] <Zeros> didn't think so
- # [22:39] <Zeros> browser sniffing has issues too. Webkit sends */* for it's Accept header
- # [22:39] <Zeros> and Webkit has issues with entities in older versions
- # [22:40] <aarondinner> don't need to browser sniff, you can check the accepts header
- # [22:40] <Zeros> that's my point
- # [22:40] <Zeros> if you check the accept header then Webkit always gets text/html
- # [22:40] <aarondinner> right
- # [22:40] <gsnedders> Zeros: It doesn't have "issues", it just doesn't read the DOCTYPE
- # [22:40] <gsnedders> or rather, external entities
- # [22:40] <aarondinner> i see
- # [22:40] <gsnedders> which is completely fine
- # [22:40] <Philip> gsnedders: That doesn't make it not an issue
- # [22:40] <Zeros> gsnedders, that is an issue, because it puts a bunch of errors at the top of the page about the entities
- # [22:40] <aarondinner> firefox 3 is changing to prefer html over xhtml because it works better for progressive rendering i believe
- # [22:40] * aarondinner is now known as aaronlev
- # [22:41] <Philip> gsnedders: since incompatibility with the real world is always an issue
- # [22:41] <Zeros> gsnedders, errors that Opera and Firefox won't generate
- # [22:41] <gsnedders> Zeros: peh. the pages are broken. they rely on optional features.
- # [22:41] <Zeros> heh, no Webkit is (was, I think it's fixed) broken
- # [22:41] <gsnedders> Zeros: it's been fixed for a while
- # [22:41] <Philip> Just agree that *everyone* is broken
- # [22:42] <gsnedders> Remember as was agreed earlier: everything I say is bullshit :P
- # [22:49] * Joins: adele_ (adele@17.203.14.214)
- # [22:51] * Quits: adele (adele@17.203.14.240) (Ping timeout)
- # [22:54] * Quits: aaronlev (chatzilla@217.228.202.109) (Ping timeout)
- # [22:55] <Philip> <meta name=pics-label> is used about 1% as much as <meta http-equiv=pics-label>
- # [22:55] <Philip> I guess authors get confused occasionally
- # [22:57] * Quits: Zeros (Zeros-Elip@129.2.175.74) (Quit: This computer has gone to sleep)
- # [23:15] * Quits: gsnedders (gsnedders@217.44.37.113) (Quit: Partying in teh intarwebs)
- # [23:48] * Quits: heycam (cam@124.168.100.30) (Quit: bye)
- # [23:58] * Joins: mjs_ (mjs@17.255.104.166)
- # Session Close: Thu Apr 17 00:00:00 2008
The end :)