| 00:04 | <jgraham> | Philip`: That final answer looks familiar |
| 00:10 | <Philip`> | Oops, I think I gave Robert Burns the wrong idea about how <input usemap> is currently supported |
| 00:11 | <Philip`> | (I'm not sure why he doesn't just try the example in Firefox and see that it works like it's meant to...) |
| 00:18 | <Hixie> | ok so if p=0.0001878 (fraction of pages with a <td<> element [sic] in my sample) |
| 00:19 | <Hixie> | and n=3.5e9 |
| 00:20 | <Hixie> | then the error is 6e6e-7 |
| 00:21 | <Hixie> | which if n is, say, 650000, then that's about +/- 2000 |
| 00:21 | <Hixie> | which is actually smaller than the error margin that my methodology introduces in and of itself, heh |
| 00:22 | <Hixie> | interesting |
| 00:22 | <Hixie> | (my numbers are obtained using some statistical estimators that have, in the case of this particular sample, an error margin of +/- 4000 or so) |
| 00:36 | <Philip`> | I'm not sure how you're going from n=3.5e9 to n=650000 |
| 00:37 | <Hixie> | er, s/n/p/ and s/650000/650000 div 3.5e9/ |
| 00:41 | <Philip`> | Did you miss the sqrt? |
| 00:42 | <Philip`> | With n=3.5e9, p=0.0001878, then n*p*(1-p) =~ 650000, and 2*sqrt(n*p*(1-p)) =~ 1600 |
| 00:42 | <Philip`> | Ah, I think that about makes sense with what you said |
| 00:43 | <Philip`> | (There is an assumption somewhere that the population is infinite, so it might be necessary to be a bit careful if the sample size is a significant fraction of the actual population, but I have no idea how relevant that is) |
| 00:46 | mpt | is confused by the aspect ratio restriction for <img> |
| 00:47 | <mpt> | "If both [width and height] attributes are specified, then the ratio of the specified width to the specified height must be the same as the ratio of the logical width to the logical height in the image file." |
| 00:48 | <mpt> | Does that mean in <https://translations.launchpad.net/ubuntu>; we're doing something wrong? (pretending for the moment that that page is HTML 5) |
| 00:51 | <Hixie> | Philip`: sorry i wasn't clear |
| 00:51 | <Hixie> | Philip`: let me try again |
| 00:51 | <Hixie> | Philip`: if i have N=3.5e9 (number of pages examined) |
| 00:51 | <Hixie> | Philip`: and r=650000 (number of pages with an element <tr<> |
| 00:52 | <Hixie> | Philip`: then p=0.000186 ish |
| 00:54 | <Hixie> | Philip`: and the error (95% confidence the real population mean r_real is within this much of r) is 2*sqrt(n*(r/n)*(1-r/n)) |
| 00:54 | <Hixie> | right? |
| 00:54 | <Hixie> | which is about 2000 (to 1sf) |
| 01:40 | <Philip`> | Hixie: That sounds highly plausible to me |
| 01:41 | <Hixie> | ok |
| 01:41 | <Hixie> | so that error is about half of the error that i get just from the statistical estimators that i use in getting my data |
| 01:44 | <Hixie> | btw i think one thing that's being missed in this input/usemap discussion is that the majority (all but three of the 50000+ that have been examined) of uses of input/usemap are either no-ops or actually result in _worse_ user experience in browsers that support usemap |
| 01:45 | <Hixie> | and that therefore browsers are better off not implementing it if they want to compete over their handling of existing pages |
| 01:46 | <Philip`> | They're not better off to a sufficiently significant extent that Mozilla or Opera have bothered changing their implementation because of it, so that doesn't seem like an especially important issue |
| 01:47 | <Hixie> | fair enough |
| 01:48 | <Philip`> | Oops, seems I responded to an old message of Rob Burns after he already said it was wrong... |
| 01:49 | <Philip`> | But I still don't understand "the image map areas should instead submit the form with the coordinates (again, there doesn't have to be a server-side to this; some use-cases do not require a server-side round-trip)" - how do you submit a form without a server side? |
| 01:49 | <Hixie> | don't ask me |
| 01:49 | <Hixie> | i don't have any idea what he's talking about |
| 01:54 | <Philip`> | Maybe you could do an accessible email-based version of Where's Wally, so you have to find and click on him in the image and then it submits the coordinates to a mailto: link, and to be accessible it also has a client-side image map with an <area> per person in the image where the alt text is the name of the person and it submits the email form with the person's coordinates when you choose one |
| 01:55 | Philip` | struggles to think of more useful examples |
| 02:01 | <Hixie> | i have written a script that reads the mbox files for public-html, and scores each participant by giving them 1 point per e-mail sent and 1 point per e-mail that was a reply to their e-mail. |
| 02:02 | <Philip`> | Does that only count direct replies? |
| 02:02 | <Hixie> | yes |
| 02:02 | <Hixie> | it counts them via In-Reply-To headers |
| 02:03 | <mjs> | I wonder4 who scores the highest? |
| 02:04 | <Philip`> | I'd tend to assume most messages only get one or two replies, because all the annoyingly large conversations are annoyingly long rather than wide |
| 02:05 | <Philip`> | in which case the score would be mainly dependent on how many emails the particular individual sends, and not affected much by the replies |
| 02:18 | <Hixie> | so i'm actually relatively surprised as to who scores the worst on this thing |
| 02:19 | <Philip`> | Are you assuming high scores are bad? :-) |
| 02:19 | <Hixie> | yes :-) |
| 02:20 | <Philip`> | Depending on who's at the top, you could say it shows who has blessed us most with their knowledge, experience and ideas |
| 02:21 | <Hixie> | the person at the top is mostly at the top because they have triggered so many replies |
| 02:23 | <mjs> | http://www.zeldman.com/2007/08/15/what-crisis/ |
| 02:23 | <mjs> | who's got the high score? I gotta know |
| 02:24 | <Hixie> | mjs: you |
| 02:25 | <mjs> | I'm not all that surprised, if it's mainly through triggering replies |
| 02:25 | <Hixie> | i think the results would be very different if i only did the last month |
| 02:26 | <mjs> | well that's true |
| 02:27 | <mjs> | I've hardly sent any mail, let alone had lots of replies |
| 02:28 | <Hixie> | bbl |
| 03:09 | <mpt> | Hixie, the definition of alt= for <area> is much vaguer than the definition of alt= for <img> |
| 03:13 | <mpt> | Perhaps the former should reference the latter somehow |
| 03:30 | <Lachy> | Hixie, what's my score in the email count? |
| 03:31 | <Lachy> | and what position do I get? |
| 03:44 | <Hixie> | http://junkyard.damowmow.com/290 |
| 03:46 | <Lachy> | 7th isn't too bad. I correct guessed who would get second, though I would have expected him to get first if mjs hadn't |
| 03:47 | <mjs> | wow, I'm the only one who elicited more responses than emails sent? |
| 03:47 | <mjs> | no, I guess not |
| 03:47 | <mjs> | misread chart at first |
| 03:47 | <Lachy> | it's a shame it doesn't distinguish between productive discussion and useless trolling |
| 03:49 | <mjs> | regarding contribution value of the #2 entry on the list, I present this: http://bugs.webkit.org/show_bug.cgi?id=14945 |
| 03:51 | <mjs> | (but maybe it's just a different guy with the same name?) |
| 03:51 | <Lachy> | Hixie, for alt text in "A key part of the content that has no textual alternative", the spec should probably try to emphasise more that alt text should be provided if possible |
| 03:52 | <Lachy> | "In such cases, the alt attribute must be omitted." should be reprhased to say "... may be omitted, but should be provided if at all possible" or something like that |
| 03:54 | <Lachy> | mjs: it's the same person, he posted about the same topic to public-html |
| 03:54 | <Lachy> | suggesting that XHTML5 should somehow more clearly specify XML parsing |
| 03:55 | <Lachy> | (though I didn't bother paying too much attention to it) |
| 03:57 | <Lachy> | oh, I understand Gregory's point about his photo album now! http://my.opera.com/oedipus/albums/showpic.dml?album=212490&picture=3302055 |
| 03:58 | <Lachy> | he wasn't suggesting that "perception - photography - image interpretation - blindness" is good alt text, he was suggesting that the comments people have provided for him are good descriptions |
| 04:01 | <mjs> | really the relevant alternative isn't good alt text vs no alt text |
| 04:01 | <mjs> | alt="" has always been allowed |
| 04:01 | <mjs> | the question is just whether leaving alt out entirely is in some cases better than alt="" |
| 04:02 | <mpt> | It's better than authoring tools producing automated alt="" for sticking-plaster compliance |
| 04:02 | <mjs> | yes, I think so too |
| 04:03 | <mjs> | but there's incentive effects both ways |
| 04:03 | <mjs> | the HTML4 way made it likely that images with no provided alternative text would be treated the same as decorative images, which is harmful to blind users or users of text-only browsers |
| 04:03 | <mpt> | For the same reason that good transcribers include "[inaudible]" in transcripts, instead of just omitting the words/phrases they didn't hear |
| 04:04 | <mjs> | the (current draft) HTML5 way might do less to encourage thoughtful authors writing by hand to add good alt text |
| 04:04 | <mjs> | since the conformance checker won't remind them |
| 04:04 | <mjs> | that was one potential advantage of noalt |
| 04:04 | <mjs> | you could verify that all your images that lack alt text are doing so on purpose |
| 04:05 | <mpt> | ooh, noalt |
| 04:05 | <mpt> | that's a nifty idea |
| 04:05 | <mpt> | but then would authoring tools produce automated noalt... |
| 04:06 | <mjs> | presumably, yes |
| 04:06 | <mjs> | but that is better than them adding alt="" |
| 04:06 | <mjs> | and it makes it harder to forget to add alt accidentally when authoring by hand |
| 04:07 | <mjs> | I could see the argument either way |
| 04:07 | <mpt> | Maybe there could be an authoring-tool exception, like there is/was for <font> |
| 04:07 | <mjs> | ironically I think Gregory's image gallery shows a valid use case for omitted alt |
| 04:08 | <mpt> | Apologies if I'm unwittingly rehashing suggestions made in public-html@ |
| 04:08 | <mjs> | I think this was mostly discussed on the whatwg list actually |
| 04:08 | <mjs> | I don't think the authoring tool exception is a good concept |
| 04:08 | <mjs> | and doesn't really work well for user-generated content embedded in a hand-authored dynamic page |
| 04:09 | <mpt> | Whenever anyone says "user-generated content" I want to throw something |
| 04:09 | <mpt> | but anyway |
| 04:10 | <mjs> | if you have a better phrase to use for what I'm referring to I would be happy to use that instead |
| 04:10 | <mjs> | I mean to distinguish content created by casual end-users who are doing it incidentally, not professionally or as their primary goal |
| 04:11 | <mpt> | Well, part of the problem with it is that I don't know what you're referring to |
| 04:11 | <mpt> | Photos? |
| 04:11 | <mpt> | Weblog comments? |
| 04:11 | <mjs> | photos posted to a photo sharing site, weblog posts, comments on a review site like yelp... |
| 04:11 | <mpt> | Forums drizzled with emoticons? |
| 04:11 | <mjs> | email messages |
| 04:11 | <mjs> | web forums |
| 04:11 | <mjs> | IM messages |
| 04:11 | <mpt> | So what are you excluding? |
| 04:12 | <mjs> | profesionally developed web sites |
| 04:12 | <mjs> | I should have said "web forum comments" |
| 04:12 | <mpt> | If it's a hand-authored page, I'm not sure how the authoring tool exception kicks in at all |
| 04:12 | <mpt> | oh, unless it's a Weblog comment system that allows images? |
| 04:12 | <mjs> | I would distinguish the web forum itself ("profesionally developed" in some sense), from comments posted by users |
| 04:12 | <mjs> | right |
| 04:12 | <mpt> | but they usually don't anyway, to prevent CSRF |
| 04:13 | <mjs> | but photo sharing sites allow embedding images |
| 04:13 | <mpt> | Maybe "usually" is overstating it |
| 04:13 | <mjs> | and email programs |
| 04:13 | <mpt> | ok, so e-mail programs would use the authoring-tool exception |
| 04:13 | <mpt> | while photo sharing sites would use noalt |
| 04:14 | <mjs> | now I'm not sure what your proposal is |
| 04:14 | <mpt> | That the authoring-tool exception save authoring tools from having to use alt="" or noalt or whatever they think the best default is |
| 04:16 | <mpt> | so that the UA can tell whether any thought has been put into the alternate text |
| 04:17 | <mjs> | if noalt existed I think the right thing would be to treat it the same as missing alt |
| 04:17 | <mjs> | so I don't think that would buy anything |
| 04:17 | <mjs> | the question is just whether noalt is worth it |
| 04:17 | <mpt> | While a photo sharing site knows that the photos are the point of the site, and can therefore apply noalt to all the images. |
| 07:25 | <G0k> | oy all |
| 07:36 | <G0k> | so er...more questions about the server-side events thing |
| 07:38 | <G0k> | when dealing with boolean event attributes, should I follow the pattern specified for Cancelable and Bubbles (i.e. "No" is false but everything else is true) or should it treat "false" and "0" as false also? |
| 07:46 | <Lachy> | G0k: IIRC, the default cancelable and bubbles values are defined for each event in the DOM Events spec, and I think HTML5 gives default values for custom events (though I'd have to check the spec) |
| 07:46 | <G0k> | yeah i got that |
| 07:47 | <G0k> | i just mean how should i parse the fields |
| 07:47 | <G0k> | for example, for the MouseEvent interface |
| 07:47 | <G0k> | if i get, say |
| 07:47 | <G0k> | ctrlKey: false |
| 07:47 | <Lachy> | doesn't the spec define that? |
| 07:47 | <G0k> | well it says |
| 07:47 | <G0k> | "If the specified interface has an attribute that exactly matches the name of the field, and the value of the field can be converted (using the type conversions defined in ECMAScript) to the type of the attribute, then it must be used." |
| 07:47 | <G0k> | but that doesn't really make sense for booleans |
| 07:47 | <G0k> | i don't think any string in ECMAScript evaluates to false |
| 07:48 | <Lachy> | hmm. interesting |
| 07:48 | <G0k> | plus, that's inconsistent |
| 07:48 | <G0k> | "No" doesn't evaluate to false in ECMAScript |
| 07:49 | <Lachy> | you could send an email to one of the lists and ask |
| 07:49 | <Lachy> | it may be a bug in the spec |
| 07:50 | <G0k> | oki |
| 08:58 | takkaria | considers unsubscribing from public-html :/ |
| 09:05 | <G0k> | what'd they do? |
| 09:06 | <takkaria> | just constant sniping and personal attacks and seemingly knee-jerk reactions |
| 09:06 | <takkaria> | to /everything/ |
| 09:07 | <G0k> | yeah so fuck em. revolution up in here |
| 10:16 | <hsivonen> | Hixie: I think the score would illustrate the issue better if you divided it by # of days the person has been on the list. (yeah, I realize that developing the score script isn't your main activity) |
| 10:20 | <Hixie> | hsivonen: hehe, not a bad idea. time from first post to last post? |
| 10:21 | <Hixie> | mpt: see the recent large alt text revamp e-mail i bcc'ed you on |
| 10:21 | <Hixie> | so i'm thinking maybe the lack of an alt="" attribute should also be a sign of a "low quality" page, though maybe that would just make people set it to alt="". (empty) |
| 10:21 | <hsivonen> | Hixie: yes |
| 10:22 | <hsivonen> | Hixie: or time from first post to today |
| 10:25 | <G0k> | should specify that not having an alt attribute means that the UA must not load any banner ads on that page |
| 10:26 | <G0k> | designers never fix anything until you mess with their money |
| 10:28 | <G0k> | i suppose it would just lead to people setting alt attributes like alt="asz89aj9g" but still |
| 10:41 | <Whiskey_M> | G0K, trust me that would be even worse - also what if you have a page with no images on that aren't purely for display. Therefore they shouldn't have any alt |
| 10:41 | <G0k> | see I almost wonder if there should really be like two different tags |
| 10:42 | Whiskey_M | would be intrigued by the stats of pages using primarily alt="" which is AFAIK pretty much a kick back against the measurement of automated accessibility checkers |
| 10:42 | <G0k> | an element for images with semantic meaning, then another element for those without |
| 10:42 | <G0k> | thus people who don't care could go on using img-no-alt for their purely display images |
| 10:43 | <G0k> | then people who cared could give use a new image element which absolutely needed the alt tag |
| 10:45 | <G0k> | er attribute |
| 10:45 | <G0k> | thing |
| 10:46 | Whiskey_M | doesn't see what would necessarily be gained - off to a meeting now, but will be thinking about it |
| 11:21 | <Hixie> | http://junkyard.damowmow.com/291 |
| 11:22 | <Hixie> | lachy and i don't even appear on that chart |
| 11:24 | <Lachy> | that aligns with my expectations better than the last one - it has expected winner! |
| 11:24 | <Hixie> | we come somewhat below the top 20 |
| 11:24 | <Hixie> | 2.59 218 199 0.91 161 Dan Connolly <connolly⊙wo> |
| 11:24 | <Hixie> | 2.50 23 24 1.04 19 John Boyer <boyerj⊙cic> |
| 11:24 | <Hixie> | 2.49 171 211 1.23 154 Lachlan Hunt <lachlan.hunt⊙lia> |
| 11:24 | <Hixie> | 2.44 2 3 1.50 3 Shane McCarron <shane⊙ac> |
| 11:24 | <Hixie> | 2.36 188 191 1.02 161 Ian Hickson <ian⊙hc> |
| 11:24 | <Hixie> | 2.35 2 3 1.50 3 Geoffrey Sneddon <geoffers⊙gc> |
| 11:25 | <Lachy> | how is the score calculated? |
| 11:26 | <Hixie> | (sent+recvd)/days |
| 11:26 | <Hixie> | more or less |
| 11:26 | <Hixie> | it's actually done by seconds |
| 11:27 | <Hixie> | if i merge tina's two accounts she drops off the top 20 too |
| 11:28 | <Lachy> | ok, so the score is the average number of emails sent per day |
| 11:29 | <Hixie> | right |
| 11:29 | <Hixie> | sent and received |
| 11:29 | <Lachy> | yeah |
| 11:30 | <Lachy> | so Robert sends 6.63 emails per day |
| 11:31 | <Lachy> | I wonder if there's some correlation between the number of posts sent+received per day and the quality of the post |
| 11:34 | <hsivonen> | Hixie: thanks for the stats |
| 11:34 | <Hixie> | are philip⊙zdcu and excors⊙gc the same person? |
| 11:34 | <Hixie> | or are those two different philips? |
| 11:35 | <Lachy> | Hixie: I think they're the same person, we're fairly sure there's only 2 Philip Taylors |
| 11:36 | <Hixie> | really? i thought there were three. |
| 11:38 | <zcorpan_> | http://www.w3.org/2000/09/dbwg/details?group=40318&public=1 lists 2 |
| 11:38 | <Lachy> | excors⊙gc is Philip` |
| 11:38 | <Hixie> | ok |
| 11:38 | <hsivonen> | zaynar is Philip` too |
| 11:38 | <Hixie> | ok |
| 11:38 | <Hixie> | excellent |
| 11:38 | <Lachy> | @zaynor is the same as P.Taylor⊙RAU |
| 11:38 | <Hixie> | no need to special case the philips in my script! |
| 11:38 | <hsivonen> | Hixie: I believe rhul and turnbridge (sp?) are one as well |
| 11:39 | <hsivonen> | Lachy: huh? |
| 11:39 | <Hixie> | wait, wait, you both just contradicted yourselves |
| 11:39 | <Lachy> | oh, maybe I'm mistaken |
| 11:39 | <Lachy> | let me check... |
| 11:39 | <Hixie> | i have: |
| 11:39 | <Hixie> | Philip TAYLOR <Philip-and-LeKhanh⊙RO> |
| 11:39 | <Hixie> | Philip Taylor (Webmaster) <P.Taylor⊙RAU> |
| 11:39 | <Hixie> | Philip Taylor <excors⊙gc> |
| 11:39 | <Hixie> | Philip Taylor <philip⊙zdcu> |
| 11:40 | <hsivonen> | Hixie: my belief is the the first two are one person and the second two are another person |
| 11:40 | <Hixie> | my script will turn those into three, collapsing the bottom two |
| 11:40 | <Hixie> | ok |
| 11:40 | <Hixie> | so i do need to special case them |
| 11:41 | <Lachy> | I thought there were only 3 email addresses for the Philips, so I assumed the @zaynar one had to be the same as the other |
| 11:42 | <Lachy> | Philip`: yt? Care to settle this mystery about the last 2 email addresses? |
| 11:43 | <Philip`> | The last two are me, and the first two are not me |
| 11:43 | <Philip`> | (I don't remember having sent to the list with any other email address too) |
| 11:48 | <Lachy> | Philip`: you have about 22 times by my count of my local archive |
| 11:49 | <Lachy> | oh, wait, I misread your last statement. |
| 11:49 | <Philip`> | I meant I don't remember ever using any other than those two |
| 11:50 | <Lachy> | yeah, I realise that now, I thought it said you didn't remember using the latter email address |
| 11:56 | <Hixie> | wow, a lot of people use mutliple addresses |
| 11:57 | Hixie | adds generic name collapsing to this script |
| 11:58 | <Hixie> | http://junkyard.damowmow.com/291 updated |
| 11:58 | zcorpan_ | has used zcorpan@hotmail and zcorpan@gmail on whatwg, and zcorpan@gmail and simonp⊙oc on public-html |
| 11:59 | <Hixie> | yeah my script collapses those all together now |
| 12:01 | <hsivonen> | zcorpan_: you have an admirable skill of staying out of the ratholes on public-html |
| 12:04 | <Lachy> | Hixie, can you plot the volume of email from each user over time on a graph? |
| 12:05 | <Hixie> | i could |
| 12:05 | <Hixie> | not gonna :-) |
| 12:05 | <zcorpan_> | hsivonen: thanks :) |
| 12:05 | <Hixie> | if you know perl i'm happy to send you my script and you can repurpose it |
| 12:06 | <Lachy> | dammit Hixie! I demand that you do my research for me! :-) |
| 12:06 | <Lachy> | unfortunately I don't know perl |
| 12:06 | <Hixie> | :-) |
| 12:07 | <Philip`> | Perl is easy to learn - it's just like Python without all the words |
| 12:07 | <hsivonen> | perl gives you a hangover. python doesn't |
| 12:08 | <Hixie> | perl is just misunderstood |
| 12:08 | <Hixie> | mostly because it's hard to understand, but still |
| 12:09 | <hsivonen> | by hangover I mean that the day after when one reads one's own perl code, one goes wtf |
| 12:20 | <Hixie> | btw spec is updated |
| 12:25 | <Hixie> | with new alt text advice |
| 12:30 | <Lachy> | Hixie: I think the new text balances the issue nicely now |
| 12:33 | <Philip`> | "Note that the following would be a very bad use of alternative text:" - why "very bad"? It doesn't really hurt much, except wasting a couple of seconds of someone's time when they listen to the image, so it only seems a little bit bad |
| 12:35 | Lachy | sets up message filter to clearly identify the sensible Philip from the other one |
| 12:35 | <zcorpan_> | Hixie: how does missing alt now integrate with <figure> semantics in the fallback case? |
| 12:36 | <Lachy> | (Philip`: now messages from you get marked with a gold star in my inbox :-)) |
| 12:37 | <Lachy> | Hixie: maybe s/very bad/redundant/ in that paragraph |
| 12:37 | <Philip`> | (Lachy: I feel very special) |
| 12:38 | <Philip`> | (though I remember I used to like shiny silver stars more than gold stars, at least when I was 5) |
| 12:39 | gsnedders | tries to remember who Philip` is |
| 12:40 | <gsnedders> | which Philip is he? |
| 12:41 | <Philip`> | I'm hope that I'm the sensible one |
| 12:41 | <Philip`> | s/e/ing/ |
| 12:41 | <zcorpan_> | the one who wrote the spec splitter, the 8000 tokenization tests, the tokenization diagrams, the canvas test suite |
| 12:41 | <zcorpan_> | and canvex |
| 12:42 | <gsnedders> | zcorpan_: yes, but on the mailing list, which? :P |
| 12:42 | <gsnedders> | the title case one? |
| 12:42 | <Philip`> | and three tokenisers :-) |
| 12:42 | <zcorpan_> | gsnedders: the other than Philip TAYLOR / Philip Taylor (Webmaster) |
| 12:42 | <gsnedders> | zcorpan_: yeah. that's what I thought. |
| 12:43 | <gsnedders> | silly common names… |
| 12:43 | <hsivonen> | use case for namespaces! |
| 12:43 | <Lachy> | Philip`: sorry, Thunderbird only gives me gold stars |
| 12:43 | <Hixie> | Lachy/zcorpan_: send mail, it's way past my bed time |
| 12:43 | <Hixie> | nn |
| 12:44 | <Lachy> | Hixie: you gonna wake up for the telcon? |
| 12:48 | <Lachy> | hsivonen: they're email addresses are the namespaces. The problem is they both break compatibility by using 2 each, and thus expect each reader to be aware of both! |
| 12:49 | <Philip`> | Readers should just apply message analysis heuristics to determine which addresses are equivalent |
| 12:50 | <Lachy> | that's what I did, it's just confusing because it takes reading the first few sentences to know |
| 12:50 | <hsivonen> | Philip`: that's how I figured there are two persons |
| 12:50 | <zcorpan_> | oops, perhaps i should have changed the subject line for that email... oh well |
| 12:50 | <Philip`> | Also, if you see two of the email addresses replying to each other, they're either distinct people or are one person who is crazy |
| 12:51 | <Philip`> | Anyway, I don't have any problem telling who each message came from :-p |
| 12:52 | <Lachy> | what about in the case of Robert Burns, who frequently responds to himself from the same email address? |
| 12:52 | <Lachy> | zcorpan_: I don't understand the question in your email |
| 12:53 | <zcorpan_> | Lachy: "If the embedded content cannot be used, then, for the purposes of establishing what the figure element represents:" |
| 12:54 | <Philip`> | Hmm, I meant (but didn't say) "replying" as in "responding and discussing an issue from different sides", rather than as in just having Reply-To headers that might be followup messages instead of responses |
| 12:56 | <zcorpan_> | Lachy: the figure section assumes that there will be fallback content (which might be empty), but no alt means that there is no fallback content |
| 12:56 | <Lachy> | yeah, I just read that section |
| 12:58 | <Philip`> | (That reduces to the message analysis heuristics case again, so I guess it's not much help) |
| 13:00 | <zcorpan_> | thinking about it, <embed>, <video> and <audio> don't have fallback content either |
| 14:14 | <zcorpan_> | Philip`: do you have any suggestions for how to create lots of testcases for serializing html fragments? i already have a framework to run test cases, now i just need to create them |
| 14:18 | <zcorpan_> | Philip`: the format looks like http://simon.html5.org/temp/zon/tests.dat |
| 14:19 | <Philip`> | The way I did it with the tokeniser was to create a load of short (mostly 1 character) strings, run the tokeniser algorithm on all of them, then remove the ones that were redundant (i.e. resulted in exactly the same tokeniser state), then made a load more slightly-longer strings by appending short strings to each of the ones from the previous stage |
| 14:19 | <Philip`> | ...and then repeated until the strings were sufficiently long that they covered every tokeniser state |
| 14:20 | <Philip`> | (and then did another half-step so that it (almost) covered every transition out of every state) |
| 14:20 | <zcorpan_> | aha. interesting |
| 14:21 | <Philip`> | That requires the ability to run the tokeniser algorithm and see what state it ends up in - I have no idea if you can do the same with HTML serialisation |
| 14:25 | <Philip`> | (Also, it's only really testing that specific implementation of the algorithm - that works alright for the tokeniser since everyone implements it the same way, but it doesn't work in more general cases) |
| 14:28 | <zcorpan_> | hmm, the algorithm doesn't require any specific order for attributes |
| 14:28 | <zcorpan_> | which might be good for performance optimazation but not good for writing test cases |
| 14:56 | <hsivonen> | we are developing a *Web* spec, right? how are local drives part of the Web? |
| 14:58 | <Philip`> | I've seen quite a few file:/// links on web pages, though I'm not sure if that counts |
| 14:58 | <Lachy> | hsivonen: Google should just index file:// URIs too, and then we could survey local hard drives too ;-) |
| 14:59 | <Lachy> | we could survey intranets too. I'm sure there are plenty that don't have enough security and would let a bot in to fetch the pages |
| 14:59 | <Philip`> | <a href="file:///etc/passwd">SomeUniqueWord</a> then search Google and view the cached copy to see Googlebot's passwords |
| 15:00 | <Lachy> | LOL |
| 15:03 | <zcorpan_> | Philip`: actually, Support Existing Content applies to <input usemap>, it's just that in order to support existing content you have to remove support for the feature... :) |
| 15:06 | <Philip`> | I'm not really convinced by that as a reason, since the number of affected pages seems to be small enough that Mozilla and Opera haven't bothered removing the feature |
| 15:07 | <zcorpan_> | yet |
| 15:07 | <zcorpan_> | we didn't know it was a problem in supporting it |
| 15:07 | <zcorpan_> | :) |
| 15:08 | <zcorpan_> | but it's not critical |
| 15:09 | <Lachy> | I like Rob's claim that there were no examples presented that break! :-) |
| 15:10 | <Lachy> | in fact, if it were to be implemented as he imagines, it would break any of the sites that use <input usemap> as a hyperlink image map (though, they should use <img> anyway) |
| 15:12 | <zcorpan_> | i thought Hixie specifically tried to find valid uses of <input usemap>, and filtered out obvious misuse |
| 15:23 | <Lachy> | zcorpan_: it depends on your definition of misuse. Hixie filtered out areas with no href on the grounds that it's completely useless in all existing UAs, whereas Rob was to include those and exlude those with href, because that's how his imaginary implementation works |
| 15:24 | <Lachy> | s/was/wants/ |
| 15:25 | <Lachy> | but what Rob should be looking for, is pages that fulfil the use case using alternative techniques. But he's refusing to do any work |
| 15:26 | <zcorpan_> | right |
| 15:41 | Philip` | wonders when to give up |
| 15:51 | <Lachy> | Philip`: just give up now. I have. He's clearly not interested in resolving the use case issue, and there's no point discussing it any further |
| 16:43 | <Lachy> | can anyone make sense of William Loughborough's recent post in the baby steps thread? |
| 16:48 | <hober> | It didn't make the slightest bit of sense to me. |
| 16:52 | <Philip`> | I have problems parsing his first sentence-paragraph |
| 16:52 | <Lachy> | I just carefully re-read it, looking up obscure words I'd never seen before like 'engendered' and 'deleterious', he seems to be talking about the harm of eliminating the alt attribute and replacing it with something else |
| 16:52 | <Philip`> | (I can't find any way to read it that makes grammatical sense) |
| 16:53 | <Lachy> | nor can I. I mostly did a sort of heuristic analysis and constructed the meaning from a few key words |
| 16:54 | <hsivonen> | it is totally counter-productive that people dramaticize making the alt attribute omissible in some cases as the removal of the possibility to use the attribute |
| 16:54 | <Philip`> | Seems easier to just ignore it :-) |
| 16:54 | <Philip`> | (though maybe less fun, depending on what kind of fun you like having) |
| 16:56 | <Lachy> | it's too stressful to deal with misunderstandings. If someone wants to jump to conclusions about us dropping the alt attribute, good for them. We know we're not, we're trying to improve it as a whole |
| 16:58 | <Philip`> | It seems the intention is to remove bad features, which nobody really minds (like all the presentational features, and frames, and @name, and script@charset, etc), but some accessibility features are bad features, and people focus on the fact that they're accessibility features rather than that they're bad features |
| 17:00 | <Lachy> | that's because all accessibility features are automatically *good*, regardless of how ineffective they are |
| 17:11 | <Lachy> | I've been thinking it might be a good idea to explain in the spec that alt text is dependent upon the context in which it's used, rather than just the image itself. |
| 17:11 | <Lachy> | so it's possible for the same image to be used in different contexts, and have completely different alt text |
| 17:12 | <mpt> | oh yes |
| 17:12 | Lachy | will send mail |
| 17:12 | <mpt> | I tried to explain that to the Wikipedia developers once |
| 17:12 | Lachy | wonders whether I should send to public-html or whatwg? |
| 17:12 | <mpt> | They were wondering if the alt text for an image could be kept on the image's page |
| 17:12 | <zcorpan_> | http://simon.html5.org/temp/zon/001.htm -- safari does surprisingly well |
| 17:14 | <Lachy> | it would be clearer if you listed the tests that pass as well. I looked and thought it had failed |
| 17:14 | <Lachy> | or gave some indication about how many tests there are and how many it passed, listing the failed ones below |
| 17:17 | <Lachy> | mpt: do you know of an example I could list in my email? |
| 17:17 | <Philip`> | http://som.uthscsa.edu/ is weird - it looks like it's really trying to sensibly use <td usemap>, but that doesn't work at all in any browser I've tried |
| 17:18 | <Philip`> | I guess they must have started with a <img usemap> and then changed it to a <td background=... usemap> |
| 17:21 | <zcorpan_> | Lachy: hmm... is that good enough? :) |
| 17:23 | <Lachy> | oh I see it now. I didn't notice it at first, cause it was just below the bottom of the viewport in Firefox |
| 17:24 | <mpt> | Lachy, unfortunately not, I couldn't find one at the time |
| 17:24 | <Lachy> | could you add a summary that says "Passed: 20, Failed 5" (or equivalent) |
| 17:24 | <mpt> | Lachy, however, Wikipedia's "This image is used in these articles:" feature might be useful |
| 17:25 | <zcorpan_> | Lachy: now have "Finished processing. 3 of 25 tests passed." |
| 17:25 | <Lachy> | yeah, that's good |
| 17:29 | <Philip`> | I can't find any real examples, but I can imagine a page that has a fancy logo in the header (alt text being the textual part of it, like the organisation name), where the page is about the history of the organisation and includes images of various logos that it has had (including the same one it's currently using in its header, with alt text being a description of its appearance) |
| 17:34 | <Lachy> | like Google's list of special event logos? |
| 17:34 | <Lachy> | good example! |
| 17:34 | <Lachy> | http://www.google.com/holidaylogos.html |
| 17:35 | <gsnedders> | what's worse to come across: the HTML WG, or two hunters on legendary when the rest of the UNSC forces are dead? |
| 17:35 | <gsnedders> | (and when you're out of practice) |
| 17:35 | <Lachy> | what's UNSC? |
| 17:36 | <Lachy> | and what does "two hunters on legendary" mean? |
| 17:36 | <gsnedders> | Halo. |
| 17:36 | <Philip`> | http://www.google.com/stickers.html - top-left logo should presumably have alt="Google", other logos should perhaps have "Small Google logo on white background" so people who want to use it on their own site can find the appropriate one |
| 17:37 | <Lachy> | I'll include both examples |
| 17:37 | <gsnedders> | Lachy: http://en.wikipedia.org/wiki/United_Nations_Space_Command and http://en.wikipedia.org/wiki/The_Covenant_%28Halo%29#Hunters |
| 17:37 | <Lachy> | although, perrhaps not, if the table included row headers that describeed the white, grey and black alternatives |
| 17:37 | <gsnedders> | Lachy: and legendary is the hardest difficulty in Halo (and Marathon, IIRC) |
| 17:37 | <Philip`> | The holiday logos might be interesting to consider when they're used on the main search page - the alt text should probably still just say "Google", but it'd be nice to have a longdesc or something so people can find out more about the particular image if they're interested |
| 17:38 | <Lachy> | indeed, it should say google, but they use the alt text given in that page |
| 17:38 | <Lachy> | it should have the extra info in the title attributwe |
| 18:08 | <Philip`> | Hmm, dangerously close to getting a "+1" on public-html... |
| 19:28 | Philip` | wonders what the OOXML namespaces are |
| 19:29 | <Philip`> | The five hundred page Primer document is full of examples like '<revisions xmlns="..." xmlns:r="...">' and I can't find anywhere saying what you actually write in place of '...' |
| 19:42 | <Philip`> | Oh, the five thousand page specification does mention the namespaces in various places, so that's alright |
| 19:43 | <Lachy> | Philip`: did you download the entire 5000 page spec? |
| 19:43 | <G0k> | well is "..." a valid namespace? |
| 19:44 | <Lachy> | G0k: no, it needs to be a URI |
| 19:44 | <Lachy> | (though, in practice, "..." would work since consumers simply do string comparison. They never need to resolve the URI |
| 19:45 | <Philip`> | Lachy: It's only 33MB |
| 19:46 | <Lachy> | ok |
| 19:46 | <hsivonen> | Lachy: true for just about any API except XOM |
| 19:46 | <Lachy> | what does XOM do specially? |
| 19:46 | <hsivonen> | Lachy: XOM is designed to barf if the string does not look like a URI |
| 19:47 | <Philip`> | "<w:r><w:t>This sentence needs to be long enough to cause some kind of line br</w:t><w:softHyphen/><w:t>eaking.</w:t></w:r>" - isn't there, like, a Unicode character for soft hyphens? |
| 19:47 | <Lachy> | what's its criteria for looking like a URI? |
| 19:47 | <hsivonen> | Lachy: dunno. I haven't checked |
| 19:48 | <Philip`> | (Their examples of line-breaking are all rather broken because the PDF page width is wider than their "long enough" sentences so it doesn't actually demonstrate any line-breaking) |
| 19:51 | <zcorpan_> | Philip`: there is, U+00AD |
| 19:52 | <hsivonen> | Lachy: https://xom.dev.java.net/source/browse/xom/src/nu/xom/URIUtil.java?rev=1.26&view=markup |
| 19:52 | <G0k> | http://www.oreillynet.com/xml/blog/images/OpenXMLSchemas3.pdf |
| 19:52 | Philip` | wonders why they use the term "high ANSI" for all Unicode characters that are not ASCII, "complex script" or "East Asian" |
| 19:52 | <G0k> | is that list of namespaces at the end what you're looking for? |
| 19:52 | <hsivonen> | Lachy: doesn't appear to be comprehensive |
| 19:52 | <Lachy> | hsivonen: thanks |
| 19:53 | <Philip`> | zcorpan_: I guess the more relevant question is, why <w:softHyphen/> instead of ­? |
| 19:54 | <Philip`> | G0k: Ah, thanks, looks like it is |
| 19:56 | <zcorpan_> | Philip`: ­ doesn't have enough angle brackets? :) |
| 19:57 | <G0k> | think we should just remove all non-tag stuff from our formats |
| 19:58 | <G0k> | <word value="hello"/><comma /><space /> <word value="world"/> |
| 19:59 | <Philip`> | It's not like they're that fond of angle brackets - they use one tag with a comma-separated lists of font names in an attribute, instead of something properly XMLish like <fonts><font name="1"/><font name="2"/></fonts> |
| 20:01 | <hsivonen> | http://about.validator.nu/htmlparser/ |
| 20:01 | <hsivonen> | does that look like something that could be announced? |
| 20:02 | <hsivonen> | nope |
| 20:02 | <Philip`> | ODF says to use Unicode characters for soft/non-breaking hyphens |
| 20:02 | <Philip`> | but it does have a <text:tab/> element |
| 20:02 | <Philip`> | Oh, it also has <text:s/> for a space character |
| 20:02 | <Philip`> | (and <text:s text:c="42"/> for multiple space characters) |
| 20:03 | hsivonen | fixed the download link |
| 20:04 | <Philip`> | I guess XML isn't so great at marking up text documents when you want to have complex typography like whitespace |
| 20:05 | <hendry> | is there some spec that specifies img {display none;} shouldn't download the image? |
| 20:06 | <Philip`> | hsivonen: http://about.validator.nu/htmlparser/apidocs/nu/validator/htmlparser/sax/HtmlParser.html says "Version: $Id$" which doesn't sound terribly useful |
| 20:06 | <Philip`> | hendry: I believe HTML5 says they must download the image |
| 20:07 | <hsivonen> | Philip`: hmm. template dating from CVS days. how do I do -kkv with Subclipse? |
| 20:07 | <Philip`> | (or at least it says they must download the image when the image element is created, regardless of whether it's invisible via CSS) |
| 20:07 | <Philip`> | (but Opera doesn't do that, if I remember correctly) |
| 20:08 | <Philip`> | hsivonen: I have no idea with Subclipse, but you'd have to set the file's "svn:keywords" property to "Id" (for each affected file) |
| 20:08 | <hsivonen> | Philip`: thanks |
| 20:09 | <hendry> | Philip`: {display none;} seems to be a strategy for not downloading imgs on network expensive environments like handhelds/mobiles |
| 20:09 | <Philip`> | hsivonen: Is htmlparser-1.0.zip meant to contain a __MACOSX directory containing only hidden files? |
| 20:09 | <hsivonen> | Philip`: artifact of OSX |
| 20:10 | <Philip`> | I guessed that from the name ;-) |
| 20:10 | <Philip`> | (but have no idea what it's containing) |
| 20:11 | <hsivonen> | Philip`: HFS resource fork and metadata emulation cruft |
| 20:11 | <Philip`> | hendry: Do you know if browser other than Opera acts that way? |
| 20:11 | <Philip`> | *if any browser |
| 20:15 | <hendry> | Philip`: no, but I think it's a good idea if certain imgs can be marked not to download for the "@media handheld" |
| 20:15 | <Philip`> | hsivonen: The XSLT4HTML5 example appears to run correctly, so that seems good :-) |
| 20:15 | <hsivonen> | Philip`: good |
| 20:16 | <hsivonen> | proof by implementation that XSLT can be used with HTML5 |
| 20:19 | Philip` | tries sorting the HTML5 spec |
| 20:19 | <Philip`> | "Exception in thread "main" org.xml.sax.SAXParseException: Start tag ?h1? has a non-NCName name." - oh, that didn't work :-( |
| 20:20 | <hsivonen> | whoa |
| 20:21 | <hsivonen> | Philip`: thanks |
| 20:21 | <hendry> | """When the src attribute is set, the user agent must immediately begin to download the specified resource, unless the user agent cannot support images, or its support for images has been disabled.""" |
| 20:23 | <hendry> | its support for images has been disabled. == img {display none;} ?? |
| 20:24 | <hsivonen> | forgot the Digit production... |
| 20:24 | <takkaria> | hendry: no, as in the user has told the browser to not display images |
| 20:24 | <takkaria> | setting an image to display:none does not amount to diabling image support |
| 20:24 | <takkaria> | since e.g. alt text isn't displayed |
| 20:24 | <hendry> | that's really binary when it comes to images. |
| 20:25 | <hendry> | though I guess it could prevent odd things happening. |
| 20:30 | <hsivonen> | Philip`: fixed |
| 20:31 | <Philip`> | "Exception in thread "main" org.xml.sax.SAXParseException: This document is not mappable to XML 1.0 without data loss to ?--? in a comment." |
| 20:31 | <Philip`> | (I assume that's intentional) |
| 20:32 | <hsivonen> | Philip`: that is correct, isn't it |
| 20:35 | <Philip`> | Hmm... Your validator complains about it on line 41550, column 5 |
| 20:35 | <Philip`> | which is |
| 20:35 | <Philip`> | <!-- XXX alt="": Define that either the src="" is shown (as an image) |
| 20:35 | <Philip`> | or the alt="" is shown (inline) but should not ever have both at |
| 20:35 | <Philip`> | once. --> |
| 20:35 | <Philip`> | I'm not quite sure where the "Consecutive hyphens did not terminate a comment." comes from in that section |
| 20:35 | <Philip`> | Oh |
| 20:35 | <Philip`> | There's a <!-- just before that |
| 20:36 | <hsivonen> | whew |
| 20:37 | <Philip`> | If I fix that, it all seems to work correctly |
| 20:38 | <hsivonen> | Philip`: thanks |
| 20:41 | <Philip`> | If I pass a load of arbitrary HTML through it, I get "Exception in thread "main" org.xml.sax.SAXParseException: Attribute name ?type"block"? is not an NCName." which seems entirely true |
| 20:45 | Philip` | can't think of any other easy ways to try to break it |
| 22:51 | tndH | is really fed up of hearing the phrase "fight the good fight" in a web standards context |
| 22:51 | <G0k> | yeah we gotta fight dirty |
| 22:51 | <gavin_> | we need to fight the good fight and get rid of that phrase |
| 22:53 | <Philip`> | We could all be friends and stop fighting |
| 22:53 | <G0k> | what's the fun in that? |
| 22:53 | <Philip`> | Well, maybe we could just act like we're friends, which leaves open the opportunity for backstabbing people when they're not careful |
| 22:54 | <G0k> | aha excellent |
| 23:03 | <karlUshi> | 1. You don't talk about fight club. |
| 23:03 | <karlUshi> | 2. You don't talk about fight club. |
| 23:04 | Philip` | has probably been playing Defcon too much, where the only reason you ever ally with someone is so you can drop nukes on them from behind when they are busy elsewhere |
| 23:05 | <Hixie> | i'm amused that none of the people discussing the KDE example in the spec have noticed that i didn't make it up |
| 23:05 | <Hixie> | it's lifted straight from wikipedia |
| 23:06 | <takkaria> | Hixie: you're missing an "attribute" in the new img text for the no-alt section: "In such cases, the alt attribute may be omitted, but the alt should be included" |
| 23:07 | <G0k> | uh. "may be omitted, but should be include" isn't that like...repetitive? |
| 23:09 | <G0k> | not that there's anything wrong with that |
| 23:09 | <Hixie> | takkaria: fixed, thanks |
| 23:10 | <Hixie> | G0k: in this case, the words "may" and "should" have very specific meaning that make this contrived phrasing useful |
| 23:10 | <G0k> | fair enough |
| 23:11 | <G0k> | Hixie: not to bug but did you get my other note on event-source stuff? |
| 23:13 | <Hixie> | yeah, my mail server is having issues and i can't send my reply |
| 23:13 | <Hixie> | but it's written |
| 23:13 | <G0k> | neato |
| 23:15 | <G0k> | sorry to keep bugging about that...i have an implementation that i don't want to commit until it's pretty |
| 23:15 | <Hixie> | ah |
| 23:15 | <Hixie> | ok |
| 23:15 | <Hixie> | coo |
| 23:15 | <Hixie> | l |
| 23:15 | <G0k> | don't want to start of broken |
| 23:15 | <G0k> | *off |
| 23:17 | <zcorpan_> | was it good ideas and bad code that was successful in building communities? :) |
| 23:27 | <Hixie> | the people working on the html5 status thing are charlvn and zcorpan_ right? |
| 23:30 | <zcorpan_> | yeah |
| 23:30 | <zcorpan_> | though i haven't done anything on that for quite some time now |
| 23:31 | <zcorpan_> | charlvn had set up some php backend iirc |
| 23:31 | <gsnedders> | ergh. PHP is horrid. |
| 23:31 | <zcorpan_> | but i don't know what is supposed to be passed on to the php script |
| 23:32 | <gsnedders> | (sorry, I've just been paid to do two weeks of work in PHP) |
| 23:32 | <G0k> | php is the VB of the web generation |
| 23:32 | <Philip`> | Writing CGI scripts in C is fun |
| 23:33 | <Philip`> | http://canvex.lazyilluminati.com/misc/dom-viewer/reflect.cgi?%3Cblink%3EHello%20world - I don't know if I should worry that it's the ultimate XSS hole |
| 23:34 | <G0k> | heh |
| 23:34 | <Hixie> | zcorpan_: yeah i need to find someone to take the ball on that and run with it |
| 23:35 | <zcorpan_> | yep, would be good |
| 23:35 | <G0k> | what's the status thing supposed to do? |
| 23:36 | <zcorpan_> | mark sections as being "work in progress", etc |
| 23:37 | <Philip`> | Could any of it be rolled into the spec-splitter script, or does it all need to work in the single-page version too? |
| 23:38 | <zcorpan_> | dunno |
| 23:39 | <zcorpan_> | i think right now only the interface to update the markers is actually missing |
| 23:41 | <Hixie> | yeah i really want something that i can just have as a floating toolbar over the spec itself that i can just click on a section and then say "this is stable" or "known issue here" and include links to feedback (like blogs) that i have to deal with, as well as linking to test cases and listing what implementations support a particular section, etc |
| 23:43 | <G0k> | would be really cool if it like..scanned the mailing lists and automagically added links to messages which quoted sections |
| 23:43 | <G0k> | how crrrazy would that be? |
| 23:45 | <Philip`> | That doesn't sound incredibly crazy |
| 23:45 | <Philip`> | at least if "quoted sections" means "posted a link to a section", rather than quoting the actual content |
| 23:45 | <G0k> | sure that |
| 23:46 | Philip` | already has a script that matches up quoted bits of the spec within a paragraph, though |
| 23:46 | <Philip`> | (to generate the annotations for http://canvex.lazyilluminati.com/tests/tests/spec.html ) |
| 23:48 | <Hixie> | i'd be happy to have cray cool stuff too, but i recommend starting small :-) |
| 23:49 | Philip` | needs to get around to finishing rejigging the spec-splitter at some point... |