| 00:02 | <othermaciej> | well, I commented on Molly's post |
| 00:02 | <othermaciej> | although mostly responding to her, not her commenters |
| 00:02 | <othermaciej> | who seem to be off on another tangent altogether |
| 01:17 | <annevk> | the comments on molly.com latest post are just sad |
| 01:25 | <annevk> | xml parser error: http://tommorris.org/blog/ |
| 01:25 | annevk | hits "Reparse document as HTML" |
| 01:37 | <Philip`> | XHTML isn't great at marketing itself, since the only time you ever notice that a site uses XHTML is when it gives you an XML parse error |
| 06:29 | <Hixie> | i don't really understand molly |
| 06:29 | <Hixie> | didn't she used to be pro progress? |
| 06:34 | <othermaciej> | I'm not sure if she's as against progress as she sounds |
| 06:35 | <othermaciej> | I thought it was funny that she picked on AIR but not Silverlight |
| 06:41 | <Lachy> | it looks like JF and others are still complaining about pave the cowpaths (see his comment on molly's blog). I think he just misunderstands it, misrepresents it and then objects to it |
| 06:41 | <Lachy> | but I think we need to reword it to make it a little less vague |
| 06:42 | <Lachy> | this is an alternative wording I've come up with: |
| 06:42 | <Hixie> | i still don't really understand why this was never a problem before the w3c htmlwg |
| 06:42 | <Lachy> | Where possible, evaluate the success and failure of existing practices or solutions and consider adopting and/or improving upon them in preference to forbidding them and/or inventing new features. |
| 06:43 | <tantek> | o hai hixie, othermaciej & lachy |
| 06:43 | <Hixie> | lachy: what's the current wording? |
| 06:43 | <Hixie> | hey tantek |
| 06:43 | <Lachy> | hi tantek |
| 06:43 | tantek | hopes he didn't speak too out of line on molly's post. |
| 06:43 | <Lachy> | Hixie: "When a practice is already widespread among authors, consider adopting it rather than forbidding it or inventing something new." |
| 06:44 | <Hixie> | Lachy: i prefer your new wording, yeah |
| 06:44 | <othermaciej> | hello tantek |
| 06:44 | <othermaciej> | Lachy: I agree it should be reworded, I like your wording |
| 06:44 | <tantek> | Hixie, this *was* a problem before (the new) w3c htmlwg |
| 06:44 | <othermaciej> | Lachy: I do think a lot of people are misunderstanding it |
| 06:45 | <Lachy> | IIRC, a few people have compared pave the cowpaths with adopting tables for layout, since it's so common, but such comparisons miss the point about evaluation the solutions |
| 06:45 | <othermaciej> | Lachy: but some misunderstandings seem to be wildly out of line with even the current wording |
| 06:45 | <Hixie> | lachy: we might want to look at the usemap="" thing as another example for that btw, since it's not accessiblity-related and doesn't involve xml syntax... |
| 06:45 | <othermaciej> | for example, claiming that "pave the cowpaths" would calll for removing particular html features |
| 06:45 | <tantek> | pave the cowpaths is not about adopting markup usage literally - it's about analyzing people's *data* *publication* behaviors |
| 06:45 | <Hixie> | Lachy: it highlights the focus on research |
| 06:46 | <tantek> | at least in the context of microformats |
| 06:46 | <tantek> | people are starting to use "pave the cowpaths" out of context |
| 06:46 | <Hixie> | though as othermaciej says, that's probably best given as a separate principle |
| 06:46 | <Hixie> | to avoid confusing people |
| 06:46 | <othermaciej> | I think the principle might need a new name as well as new text along lines proposed by Lachy |
| 06:47 | <Lachy> | one issue with that new wording is that it sounds a lot like Don't Reinvent the Wheel |
| 06:47 | <tantek> | i'm ok with confused people as long as they keep their dialog civil |
| 06:47 | <Hixie> | tantek: what i meant about the problem is that before the w3c htmlwg started, we had very few people argue from a position of authority rather than using rational discussion and basing their opinions on research |
| 06:48 | <tantek> | interesting |
| 06:48 | <Hixie> | tantek: still now, indeed, the discussions in whatwg's list are more rational than many on public-html (though there has been much progress on that front in public-html, i must admit) |
| 06:48 | <tantek> | well that's hopeful |
| 06:48 | <Hixie> | it might just be that the w3c community has fostered the development of "experts" whose entire careers are positioned around being on working groups and having their way |
| 06:50 | <Lachy> | yeah, Ben Millard on public-html has done some good work on finding table examples for doing research on headers. I wish that had happened when the headers debate started and we were asking for it |
| 06:50 | <Lachy> | but at least it is improving now |
| 06:50 | <Hixie> | yeah |
| 06:50 | <Hixie> | i'm gonna have to look at that issue soon, i have a ton of data i collected in early july about it |
| 06:51 | <Hixie> | i guess i can do that after <img alt> |
| 06:51 | <Lachy> | what needs to be done for alt? |
| 06:56 | <Hixie> | see the spec for my wip revamp of <img alt> |
| 07:00 | <Lachy> | Hixie, another example for a piece of text with an alternate graphical representation would be some company logos |
| 07:00 | <Hixie> | yeah |
| 07:01 | <tantek> | Hixie, yeah I need to write up the table headers usage examples for hCalendar more properly |
| 07:01 | <Lachy> | You're also missing "An iconic representation of the surrounding content" - it's similar to "graphical representation of some of the surrounding text", but not quite the same |
| 07:02 | <tantek> | wow: http://twitter.com/rogerjohansson/statuses/200225882 |
| 07:03 | <Lachy> | see, for example, http://www.webnauts.net/ - the wheelchair icon represents the concept of accessibility, the magnifying glass represents SEO, etc. |
| 07:03 | <Lachy> | (although, the alt text used on that site is ironically inappropriate) |
| 07:07 | <Lachy> | woah, this is surprising too http://twitter.com/rogerjohansson/statuses/132542642 |
| 07:09 | <tantek> | clearly Roger felt bullied whether or not he actually was |
| 07:10 | <othermaciej> | I'm not sure why he felt bullied |
| 07:10 | <tantek> | i think there is a bit of a culture transition clash going on |
| 07:11 | <othermaciej> | I don't see that many messages from him, and none that are in my public-html folder seem to have engendered particularly heated responses |
| 07:11 | <Lachy> | some people, particularly the accessibility advocates, just have difficulty seeing our point of view and object whenever someone questions the conventional wisdom |
| 07:12 | <Lachy> | I'll speak to Roger later and find out what he really thinks |
| 07:12 | <tantek> | standards organizations, and w3c is no exception, often designed things "a priori", depending on the opinion of "experts" to just "pick the right thing" a priori, based on their expertise. |
| 07:13 | <tantek> | the approach that microformats (and apparently WHATWG) is taking with scientifically challenging features and demanding justification through research/evidence is quite new |
| 07:13 | <tantek> | and quite threatening to those that are used to simply getting their way because they say so |
| 07:13 | <tantek> | as in, how dare anyone question their expertise? |
| 07:13 | <tantek> | shouldn't their expertise be sufficient to justify things? |
| 07:15 | <Lachy> | one thing I've also noticed is that some people refuse to look at the issues from the authoring point of view, and only consider the benefits for those with accessibility-related needs |
| 07:15 | <tantek> | yes, i've noticed that tooo |
| 07:16 | <tantek> | it's like, just because you ask, people aren't necessarily going to do it. econ 101, psych 101 etc. |
| 07:16 | <Lachy> | i.e. they advocate features regardless of how easy they are for authors, and refuse to consider authoring incentives beyond legal and moral arguments |
| 07:16 | <othermaciej> | it really depends on what the standards body in question is, as far as past experience goes |
| 07:16 | <tantek> | Lachy, to be fair, *some* advocate features that way |
| 07:17 | <tantek> | painting all accessibility advocates with the same brush is inaccurate |
| 07:17 | <othermaciej> | my understanding is that, for instance, things like the C99 or H.264 had quite a bit of empirical work go into changes |
| 07:17 | <Lachy> | yeah, I realise it's not all, but it is at least a few of the more vocal ones |
| 07:17 | <othermaciej> | go into their design rather |
| 07:17 | <tantek> | yes on both counts |
| 07:18 | <othermaciej> | I have seen a number of people appear to take the stance that disagreeing with their preferred technical approach means you are against human rights |
| 07:18 | <Lachy> | indeed, though I think some recent discussions clarified that a bit |
| 07:19 | <othermaciej> | voicing opinions like that tends to lead to communication breakdown |
| 07:19 | <othermaciej> | John Foliot still seems to stand by this stance |
| 07:19 | <othermaciej> | I say this based on both his comment post on Molly's blog and some private communication after I raised that issue |
| 07:19 | <tantek> | indeed, labeling someone "against human rights" is quite close to Godwin's law. |
| 07:20 | <tantek> | and of course, one should simply call out such ad hominem attacks for what they are |
| 07:21 | Lachy | google's Godwin's law |
| 07:21 | <Lachy> | ah http://en.wikipedia.org/wiki/Godwin's_Law |
| 07:21 | <tantek> | ah yes John Foliot. he's the one that provided the most ad hominem attack fodder |
| 07:22 | <Lachy> | tantek: also the one to most vocally complain about the attitude problems |
| 07:22 | <tantek> | keep pointing out peoples' ad hominem attacks, for those are the true attitude problems |
| 07:22 | <tantek> | if you can't be civil, then you have an attitude problem and don't deserve any sympathy |
| 07:23 | <tantek> | or put more economically, IMHO, uncivil comments are lower priority than civil comments |
| 10:06 | <takkaria> | http://twitter.com/rogerjohansson/statuses/200225882 |
| 10:09 | <takkaria> | oh. wrong window there |
| 11:08 | <hsivonen> | zcorpan: Re: attribute parsing, I don't have anything to add to what Philip` said |
| 11:08 | <zcorpan> | hsivonen: ok |
| 11:09 | hsivonen | suffers from a bad case of sudden summer flu and should probably go offline instead of reading irc logs |
| 11:10 | <hsivonen> | hmm. looks like Roger Johansson removed the twitters mentioned in the IRC log today |
| 11:25 | <zcorpan> | btw, ben millard and me have discussed possible ways to improve the table headers algorithm |
| 11:26 | <zcorpan> | it will probably be forwarded to the list at some point |
| 11:27 | <zcorpan> | the current algorithm is inadequate for many tables |
| 12:01 | <Lachy> | I had a long, rational and very productive discussion with Roger on ICQ a few hours ago. I think we were able to clarify a lot of issues and understand each others perspectives a lot better. |
| 12:01 | <Lachy> | I suspect that's why he removed the twitters |
| 13:51 | <Whiskey_M> | 'lo |
| 13:52 | <zcorpan> | yo |
| 13:52 | <Whiskey_M> | hi Zcorpan, how goes? |
| 13:53 | <zcorpan> | goes well thank you, you? |
| 13:53 | <Whiskey_M> | none too shabby :-) |
| 15:57 | <met_> | http://manwithnoblog.com/2007/08/12/are-we-becoming-complacent/ |
| 16:00 | <Whiskey_M> | read it earlier today |
| 16:00 | <Lachy> | wow, see Joe's comment on molly's article. http://www.molly.com/2007/08/11/dear-w3c-dear-wasp/#comment-611988 |
| 16:02 | <Whiskey_M> | isn't that normal Joe ;-) |
| 16:10 | <Lachy> | possibly, but it's ironic that he said that after my comment explaining that we are not against accessibility features. I just don't understand why he'd come back with a strawman argument about us dropping the alt attribute |
| 16:17 | virtuelv | notes that he respectfully disagrees with fantasai on browser vendor influence over CSS |
| 16:18 | <virtuelv> | from http://fantasai.inkedblade.net/weblog/2007/css-wg-opening-up/ |
| 16:41 | <sirPalook> | Noob (just read the specs). So in <p><tt><b>foo</p>, 'tt' and 'b' are *not* closed, but cloned? |
| 16:45 | <sirPalook> | In the DOM that is |
| 16:46 | <Philip`> | If you have something after that (other than a </b> or </tt> or EOF), then yes |
| 16:46 | <Philip`> | http://james.html5.org/cgi-bin/parsetree/parsetree.py?source=%3Cp%3E%3Ctt%3E%3Cb%3Efoo%3C%2Fp%3Ebar |
| 16:46 | <Lachy> | sirPalook, yes, I believe they are reopened after the P element, or inside any following block level element (the adoption agency algorithm is confusing) |
| 16:47 | <sirPalook> | I'm trying to figure out the differences with SGML, which theoretically sais that every closing tag closes all open tags... |
| 16:47 | <sirPalook> | Yep, thanks. |
| 16:48 | <Lachy> | sirPalook, yes, that's the theory and that's how OpenSP implemented it. Not sure if that's actually defined in the SGML spec, though |
| 16:48 | <sirPalook> | If i understand correctly, cloning happens for "Formatting tags" |
| 16:48 | <Lachy> | yes |
| 16:48 | <sirPalook> | Lachy. Yep I never invested any precious brain cells on SGML either :) |
| 16:51 | Philip` | wonders if it's possible to get worse than quadratic behaviour |
| 16:51 | <Philip`> | (i.e. worse than http://james.html5.org/cgi-bin/parsetree/parsetree.py?source=%3Cp%3E%3Cb%3E%3Cb%3E%3Cb%3E%3Cb%3E%3Cb%3Ex%3Cp%3Ex%3Cp%3Ex%3Cp%3Ex%3Cp%3Ex ) |
| 16:53 | <sirPalook> | nice one. Although I think that DOM sais the "clones" point to the same thing. |
| 16:54 | <sirPalook> | It's interesting what happens with <A NAME=> though |
| 17:00 | <Philip`> | Not quite sure what you mean about <A NAME=> |
| 17:01 | <hendry> | hsivonen: do you have an idea with the element isn't used by UAs anymore really for aiding navigation? |
| 17:01 | <hendry> | hsivonen: do you have an idea with the LINK element isn't used by UAs anymore really for aiding navigation? |
| 17:03 | <hasather_> | hendry: at least Opera uses it |
| 17:03 | Lachy | responded to Joe |
| 17:04 | <Lachy> | ... on molly's blog |
| 17:04 | <Lachy> | I hope this clears up the issue http://www.molly.com/2007/08/11/dear-w3c-dear-wasp/#comment-612166 |
| 17:04 | <Philip`> | Does Opera use anything other than <link rel=next>? |
| 17:05 | <hendry> | i have loads <links on my blog natalian.org. I think I should remove them. |
| 17:05 | <hasather_> | Philip`: yes |
| 17:06 | <hasather_> | Philip`: activate the navigation bar (Appearance > Toolbar >Navigation bar) |
| 17:07 | <Philip`> | hasather_: Aha, thanks - I never noticed that before |
| 17:11 | <hendry> | hasather_: it's useless on natalian.org :) /me wonders if i can get that view on Opera mobile |
| 17:12 | <virtuelv> | hendry: afaik, no |
| 17:13 | <virtuelv> | besides, it only uses known values for rel |
| 17:14 | Philip` | wonders if it's possible to use some kind of heuristics to decide whether a longdesc string is an attempt at writing a URL or is (as seems far more common) a descriptive phrase |
| 17:15 | <Lachy> | I think Hixie was going to attempt something like that, there are ways you could test it |
| 17:15 | <Philip`> | (By "far more common", I mean I looked at about six pages which use longdesc, and most were text instead of links) |
| 17:16 | <Philip`> | If people are going to the effort of writing long descriptions, it seems unhelpful to just ignore them because they never knew it should have been a link instead |
| 17:16 | <Lachy> | check if it begins with http://, search the string for spaces (which would indicate a phrase), attempt to resolve and retrieve the URL, etc |
| 17:18 | <Lachy> | could you just produce a page that lists all the values found in longdesc attributes, and check it manually? |
| 17:18 | <Lachy> | it shouldn't be too hard for someone to visually inspect the list and recognise phrases vs. URLs |
| 17:19 | takkaria | volunteers to do such inspection |
| 17:22 | <Lachy> | Philip`, you could also compare the value with the alt attribute and list the values for alt and longdesc side by side on the result page |
| 17:22 | <Philip`> | I don't have any data about attribute values at the moment - all I've got is the list at http://canvex.lazyilluminati.com/survey/2007-07-17/analyse.cgi/pages/attr/longdesc |
| 17:23 | <Lachy> | well, we can rule out wikipedia, since they use it incorrectly and also redundantly |
| 17:23 | <Philip`> | ...but it shouldn't require significant effort to get some data about alt/longdesc/etc in particular, since they're not that rare |
| 17:24 | <Philip`> | (though I don't have code that handles non-ASCII character encodings at all correctly, which is a pain if you're trying to actually read the text on a page) |
| 17:26 | <Lachy> | schirmer.com has empty longdesc="" |
| 17:27 | <Lachy> | I think Hixie was also going to compare the longdesc attribute with the value of the surrounding <a href="">, if any |
| 17:27 | <Lachy> | there should be some discussion of this research in the IRC logs somewhere |
| 17:28 | <Lachy> | the 4th result in your list, for instance, links to the same page, as does wikipedia |
| 17:29 | <Lachy> | tv.com has <iframe longdesc="">! |
| 17:30 | <takkaria> | shakedown.fi seems to replicate the title text as longdesc |
| 17:31 | <Philip`> | <IMG align="left" alt="Aurajoen soudut" longdesc="Aurajoen soudut" src="images/aurajoki.jpg" title="Aurajoen soudut"> |
| 17:31 | <Lachy> | oh, HTML4 allows longdesc on iframe. I didn't realise that |
| 17:32 | <takkaria> | Philip`: but a fair few of them don't have alt text at all |
| 17:32 | <Philip`> | (See http://canvex.lazyilluminati.com/survey/2007-07-17/analyse.cgi/pages/tagattr/iframe/longdesc vs http://canvex.lazyilluminati.com/survey/2007-07-17/analyse.cgi/pages/tagattr/frame/longdesc vs http://canvex.lazyilluminati.com/survey/2007-07-17/analyse.cgi/pages/tagattr/img/longdesc if you're interested in a specific tag) |
| 17:33 | <takkaria> | heh, a longdesc with the same value as src |
| 17:34 | <takkaria> | stopitnow.com: <img src="images/logo-homer.gif" alt="stopitnowlogo" ... longdesc="images/logo-homer.gif"> |
| 17:38 | <takkaria> | puntoluce.tv uses longdesc on a product image to link to that product's page; most of the others I've tried have either been attempts at keyword spam, or have been links to the image or to the page they're on |
| 17:42 | <takkaria> | ACLU in Iowa seem to use it on a "get adobe reader!" image to point to the download page; I wonder if that's copy-and-pasted from an adobe webpage? |
| 17:50 | <met_> | Lachy, isn't the story about dropping alt based on http://www.whatwg.org/specs/web-apps/current-work/multipage/section-embedded.html#the-img where "alt" isn't marked as required, although bellow is written "The text is given by the alt attribute, which must be present and must have a non-empty value"? |
| 17:51 | <takkaria> | based on those links (on img), longdesc is generally misused and even when it, there's little useful information there |
| 17:52 | <takkaria> | if some more links could be dredged up from somewhere, I'd go through and put together a wiki page on longdesc usage |
| 17:53 | <Lachy> | takkaria, you can use either the whatwg wiki or the HTML WG section of the W3C wiki |
| 17:53 | <takkaria> | I think I prefer the whatwg one, mediawiki's friendlier |
| 17:54 | <takkaria> | I'll go through the links there after a bite to eat |
| 17:54 | takkaria | bbl |
| 17:54 | <Lachy> | met_, look at last alternative titled "A key part of the content that has no textual alternative", that says "no alt" below |
| 17:54 | <Lachy> | that appears to be why it's no longer required |
| 17:56 | <Lachy> | I wasn't aware of that change until tonight, but it's going to make the validation issue interesting. It's useful to know when an alt has been omitted by mistake, so maybe validators would have to starrt issuing warnings instead. |
| 17:57 | <met_> | ah I see |
| 17:58 | <Lachy> | I'm not sure why there is a distinction between no alt and alt="" in the spec |
| 17:59 | <met_> | but I think that examples mentioned there are not good example, I would prefer even alt="screenshot" before blank alt (personal opinion) - maybe find better examples where alt realle has not sense? |
| 18:01 | <Lachy> | hmm. Maybe, I'm not sure what the best approach is with screenshots and alt text. |
| 18:02 | <Lachy> | it might be good to look up some guidelines about alt usage (like Jukka's pages, htmlhelp.com, etc.) |
| 18:02 | <met_> | just looking in czech accessibility book and there is good example: when has image only decorative meaning e.g. decorative illistration there is preferer blank alt, maybe this |
| 18:05 | <met_> | like this http://www.cbsnews.com/sections/eveningnews/main3420.shtml - every article has some small picture, which often has no information and is mainly for decorative purposes, to draw readers attention |
| 18:07 | <Lachy> | yeah, I call those iconic representations |
| 18:08 | <met_> | or here the man and 2 women http://en.navrcholu.cz/ |
| 18:08 | <Lachy> | though, on News sites like that, the pictures often have associated captions |
| 18:08 | <met_> | ok its not in 100% |
| 18:09 | <Lachy> | yeah, but you're right. I already mentioned this exact issue to Hixie earlier today |
| 18:09 | <Lachy> | or yesterday, actually |
| 18:49 | <zcorpan> | hmm. why is the distributed spec out of sync with svn? |
| 18:52 | <zcorpan> | some issue with revisions after r999? |
| 18:53 | <zcorpan> | r999 was 2007-08-10, the spec header says 11 August 2007 |
| 18:53 | <zcorpan> | Hixie: ^ |
| 18:54 | takkaria | starts doing some research on longdesc properly |
| 19:08 | <zcorpan_> | http://methisto.blogspot.com/2007/08/html5-trailer-find-your-hero.html |
| 19:10 | <zcorpan_> | krijnh: in the logs, perhaps you could have a default right border or padding so that the text doesn't jump around when you hover it? |
| 19:29 | <Lachy> | cool trailer :-) |
| 19:36 | <Hixie> | zcorpan_: the draft on whatwg.org is my working copy -- i updated it to let people see my wip on alt="", but it's not done yet |
| 20:35 | <takkaria> | http://wiki.whatwg.org/wiki/Longdesc_usage |
| 20:35 | <takkaria> | if anyone's interested |
| 20:36 | <takkaria> | of the 62 pages that philip gave with <img longdesc="">, I reckon two of them are valid uses according to the spec |
| 20:43 | <Philip`> | Two? That's probably better than td@headers :-) |
| 20:43 | <takkaria> | one of them is explaining what an acessibility icon means |
| 20:44 | <takkaria> | the other one at one point links to a text/plain document telling the user they can download speech software to hear the website |
| 20:45 | <takkaria> | overall, I'd have to say that longdesc is horribly broken |
| 20:45 | <Philip`> | The data does appear to suggest that invisible metadata is bad - authors get no feedback when they do it wrong, and so they almost never get it right |
| 20:46 | <Philip`> | "3 of which look like SEO attempts (13, 18, 48, 61)" - that's four, not three |
| 20:47 | <takkaria> | true |
| 20:47 | takkaria | fixes |
| 20:47 | <takkaria> | if I ever do this again, I'm going to use something other than pen and paper |
| 20:48 | <Philip`> | Perhaps it'd be possible to use html5lib to extract the relevant data from the pages |
| 20:49 | <takkaria> | well, I used a system of curl'ing all the pages to numerically numbered files, then grepping them for "longdesc", then noting down in a column what category they fit into |
| 20:49 | <takkaria> | hooking up html5lib to a "is this right? yes/no" UI would simplify it a lot in future, I imagine |
| 20:49 | <takkaria> | I don't know python or ruby, though, so I'm no help there |
| 20:50 | <Philip`> | An html5grep might be nice |
| 20:50 | <Philip`> | (so you can find elements/attributes/etc without worrying about case-sensitivity or extraneous whitespace or whatever) |
| 20:55 | <Philip`> | http://search.cpan.org/dist/XML-Twig/tools/xml_grep/xml_grep - ah, there's already at least one XML one - all it needs is an HTML5 parser stuck on the front |
| 21:07 | <takkaria> | what should I do with that page? should I post it to public-html or whatwg? |
| 21:17 | Philip` | doesn't know whether the chance of starting a new longdesc discussion on public-html should be considered an opportunity or a danger |
| 21:18 | <takkaria> | I reckon it's probably a danger at the moment |
| 21:18 | <takkaria> | I just wonder what to do with the page now I've made it. :) |
| 21:19 | <Hixie> | i recommend waiting until i post my results and start looking at the topic, and then forwarding it to the relevant list :-) |
| 21:19 | <Philip`> | Maybe you could mention it on the WHATWG list so people know it exists, and when some longdesc discussion starts again on public-html then point to the page since it's useful data |
| 21:20 | <takkaria> | Hixie: I can do that, assuming I'm still subscribed in a couple years' time ^_^ |
| 21:21 | <Hixie> | that's a pretty cool study |
| 21:21 | <Hixie> | were _any_ of them correctly used? |
| 21:22 | <takkaria> | two, sort=of. but while they might have been correct, they weren't useful |
| 21:22 | <takkaria> | I was greatly amused by the number of sites trying to do SEO though |
| 21:23 | <Philip`> | How many were non-correct but would still be useful if a UA acted in a certain way? |
| 21:23 | Hixie | logs in to his work computer to see what his own longdesc study's results were like and finds that his ruby parser study, which he was running over the weekend, completed |
| 21:23 | <Hixie> | sweet |
| 21:26 | <Hixie> | let's see |
| 21:26 | <Hixie> | longdesc |
| 21:26 | <takkaria> | Philip`: er, I'm not sure any |
| 21:26 | <hsivonen> | w many were non-correct but would still be useful if a UA acted in a certain way? |
| 21:26 | <hsivonen> | oops. sorry |
| 21:27 | <Hixie> | so my script found some hundred billion or so longdescs |
| 21:27 | <Hixie> | er |
| 21:27 | <Hixie> | hundred billion or so <img>s rather |
| 21:27 | <Hixie> | of which about a hundred billion had no longdesc="" |
| 21:28 | <Hixie> | of the hundred million or so that _did_ have longdesc and src and in which neither was blank: |
| 21:29 | <Hixie> | about 800 thousand had longdesc = src |
| 21:29 | <Hixie> | about 85 million had longdesc = an ancestor <a href> |
| 21:29 | <Hixie> | about 7 million pointed to the root of another domain (!) |
| 21:30 | takkaria | nods, I got a fair few pointing to the root of another domain |
| 21:30 | <Hixie> | about 12 million had a space in the longdesc |
| 21:30 | Philip` | wonders if the 85 million includes the two million pages of Wikipedia * average number of images per page |
| 21:30 | <Hixie> | basically i got it down to about 8 million potentially useful longdesc=""s that need further study |
| 21:31 | <Hixie> | Philip`: might well |
| 21:31 | <Hixie> | Philip`: wikimedia in general |
| 21:31 | <takkaria> | nice stats |
| 21:32 | <Hixie> | looking at the sample of those 8 million, most seem to be URIs to images |
| 21:32 | <Philip`> | http://en.wikipedia.org/wiki/Image:STS-118_launch_M-cropped.jpg - that one page has five longdescs which match ancestor hrefs |
| 21:33 | <hsivonen> | Hixie: are you going to post a summary of the results to public-html? |
| 21:33 | <Hixie> | the result, i think, though, is that longdesc="" data is so widely abused and corrupt that an accessibility tool would get better usability without supporting it than with supporting it |
| 21:33 | <Hixie> | hsivonen: yeah, when i look at longdesc="" |
| 21:34 | takkaria | nods |
| 21:34 | <takkaria> | ach, I best be off, bbl |
| 21:35 | <Philip`> | Can an accessibility tool get better usability by automatically determining (/guessing) whether a longdesc is valid or not? |
| 21:35 | <Hixie> | hard to say |
| 21:36 | <Hixie> | i don't know how the heuristic would work |
| 21:36 | <hsivonen> | Hixie: comparing with ancestor href='' and doing a HEAD request to see if the resource is another image |
| 21:37 | <Hixie> | look at, e.g.: http://photofile.ru/users/marusyan/?page=2 |
| 21:38 | <Hixie> | several of the longdesc=""s on that page look useful to that heuristic |
| 21:40 | <zcorpan_> | Hixie: ah, ok |
| 21:44 | <Hixie> | http://wiki.whatwg.org/wiki/Talk:Longdesc_usage |
| 21:44 | <Philip`> | The exact heuristic could be left as an area for UA developers to innovate in |
| 21:48 | <Hixie> | i think before we give ATs things to innovate on, we should wait for their products to become usable in the first place. |
| 23:35 | takkaria | loks at those urls |
| 23:58 | <takkaria> | Hixie: I'll have a bash at those urls tomorrow (UK time) |
