| 00:01 | <markp> | i once threatened to make captchas based on strunk and white |
| 00:01 | <Philip`> | You could extend it to something like "d^2x/dy^2 - 5 x = 0; y = ?" if you want to raise the barrier a little |
| 00:06 | <Dashiva> | Make them write alt text for the captcha |
| 00:08 | <Philip`> | Make them write alt text for one of the 3000 unlabelled holiday photos that the blog's owner uploaded to Flickr |
| 00:09 | <Dashiva> | I wonder if anyone has contacted Flickr for statistics :) |
| 00:09 | <markp> | actually i've thought about using mechanical turk for adding alt text |
| 00:10 | <markp> | it's a skill, it can be taught, individuals could get good at it and get paid for it |
| 00:10 | <webben> | One could certainly use something like Google Image Labeller as some sort of CAPTCHA. reCAPTCHA uses captcha responses to OCR books. |
| 00:10 | <markp> | recaptcha++ |
| 00:11 | <Dashiva> | Any captcha can be broken by an ample supply of cheap human labor, after all |
| 00:11 | <webben> | throw Google Image Labeller together with something like http://research.microsoft.com/asirra/ |
| 00:12 | <markp> | seriously though, imagine if flickr (yahoo) sponsored an ongoing project to add really good alt text to images as people uploaded them |
| 00:12 | <markp> | photo owners could be notified when their images got tagged and could accept/reject/edit them |
| 00:13 | <webben> | markp: Yep that's one (of many ways) Flickr could start producing alt text. |
| 00:13 | <Hixie> | do blind people care? |
| 00:14 | <webben> | Hixie: Some do. Some don't. Bit like other people really. |
| 00:14 | <markp> | there's more to flickr than images |
| 00:14 | <Hixie> | i don't understand why anyone would want photos described. they're boring enough when you can see them. |
| 00:14 | <webben> | flickr turns photos into social objects |
| 00:14 | <markp> | people add annotations, tags, comments |
| 00:15 | <markp> | all of which are text |
| 00:15 | <Hixie> | yeah i never understood the point of the tags and annotations either |
| 00:15 | <Hixie> | the captions and comments make sense |
| 00:15 | <Hixie> | if you're showing pictures to your friends |
| 00:15 | Hixie | shrugs |
| 00:16 | <markp> | anything that increases the amount of accessible content in the world is a good thing |
| 00:16 | <markp> | the street will find its own uses for it |
| 00:16 | <Hixie> | i don't really see how alt text increases the accessibility of a photo |
| 00:16 | <Hixie> | it's like lyrics for a song |
| 00:17 | <Hixie> | ok, so you can see the lyrics |
| 00:17 | <Hixie> | still doesn't give you the song |
| 00:17 | <webben> | I think people get too hung up on the irreducibility of experiences. |
| 00:17 | <kingryan> | Hixie: but if you can't hear the song, but everyone's talking about it, wouldn't it be nice to have the context of the lyrics? |
| 00:17 | <webben> | Often the song, the image are less important (or only as important) as they social meaning and context anyhow |
| 00:17 | markp | quietly hums "we didn't start the fire" |
| 00:18 | <Hixie> | kingryan: i have a "disability" in that i cannot recognise fashionable clothing from unfashionable clothing. My solution is not to have people describe to me what is fashionable and what isn't. My solution is to not care. |
| 00:18 | Hixie | shrugs again |
| 00:19 | <Philip`> | I would guess the solution of having a female decide what you should wear is a more common one for that problem :-) |
| 00:20 | <Hixie> | my girlfriend actually has the same "disability" as i do. our solution is to get her gay boyfriend to help us pick clothes when necessary. :-) |
| 00:20 | <othermaciej_> | tags are pretty good for photos, though they don't seem as nice for most other things |
| 00:21 | <Hixie> | othermaciej: oh i use tags a ton on iPhoto. I just don't see the point in tags for other people's photos. |
| 00:21 | <webben> | othermaciej: Do you not find tags useful for bookmarking? |
| 00:22 | <Hixie> | othermaciej: but then, i don't really care about photos that aren't relevant to me, so... |
| 00:22 | <mpt> | Hixie, do you ever use Google Image Search? |
| 00:23 | <Hixie> | mpt: yeah, that's a great example of where tagging isn't necessary. |
| 00:24 | <Hixie> | even in iPhoto, image tagging is really just a workaround for the computer's inability to recognise content in the photo |
| 00:24 | <Hixie> | i don't enjoy manually tagging photos, i shouldn't have to |
| 00:24 | <webben> | Good example of the blind caring about the social context of photography was a recent blindtech question: a couple of blind parents looking for a photo sharing site that didn't require a visual CAPTCHA to register, so they could upload some photos of their newborn |
| 00:24 | <Hixie> | i should just be able to ask the computer for every photo in the collection with a cat in it |
| 00:24 | <Hixie> | or every photo that has an element in focus |
| 00:24 | <Hixie> | (or doesn't) |
| 00:24 | <mpt> | How would you know that it has a cat in it, if "cat" is in the tags but not in the title/description? |
| 00:24 | <webben> | Hixie: you want Riya and Polar Rose ;) |
| 00:25 | <webben> | http://www.riya.com/ and http://www.polarrose.com/ |
| 00:25 | <othermaciej> | Hixie: they are interesting in flickr because (a) it makes searching work better than just captions and titles would, given that AI image anaysis does not exist yet; |
| 00:26 | <othermaciej> | Hixie: and (b) because they provide a categorization scheme, so I can easily see "all of my friends' pictures of dogs" or the like |
| 00:26 | <Hixie> | webben: again, those are their own photos, and the alt text wouldn't really help insofar as i can tell -- how could text possibly convey anything useful? |
| 00:26 | <othermaciej> | webben: I don't find boomkarking useful, so I'm not a good person to ask |
| 00:26 | <Hixie> | mpt: image analysis |
| 00:26 | <webben> | Hixie: which photo is which, primarily. |
| 00:26 | <mpt> | ha ha |
| 00:26 | <Hixie> | mpt: (same way a human can tell) |
| 00:26 | <Hixie> | webben: that's a title |
| 00:26 | <webben> | othermaciej: ah okay :) |
| 00:27 | <Hixie> | othermaciej: again i'm just talking about my personal preference, but personally, i have never found search on flickr to be useful |
| 00:27 | <Hixie> | the image on my blog being a great example of how it's not useful :-) |
| 00:28 | <webben> | Hixie: Titles don't necessarily identify what images actually are. (They might be a pun etc... ) |
| 00:28 | <webben> | also, you can display images without titles |
| 00:28 | <webben> | in which case you'd need alt to tell them apart |
| 00:28 | <Hixie> | mpt: but, text that identifies the image is useful to everyone, not just if you don't have images. |
| 00:28 | <Hixie> | er |
| 00:28 | <webben> | (e.g. Flickr photostreams) |
| 00:28 | <Hixie> | s/mpt/webben/ |
| 00:29 | <Hixie> | note that i'm not arguing against including alt text here |
| 00:29 | <webben> | Hixie: It's more useful to people who can't see the image, since to some degree you can tell images apart just by looking at them. |
| 00:29 | <Hixie> | i'm just saying i don't understand why anyone would care for pure alt text (as opposed to text that is shown near the picture even when you do have images) |
| 00:29 | <Hixie> | webben: *shrug* |
| 00:29 | <mpt> | I thought we were talking about tags |
| 00:29 | <mpt> | My apologies if we weren't |
| 00:30 | <webben> | Hixie: Yeah. I can't really see why you wouldn't want an image described. |
| 00:30 | <mpt> | tags would be lousy alt= text |
| 00:30 | <Hixie> | mpt: regarding tags, all i was saying was that they are an annoying and hopefully temporary workaround for searching photos. |
| 00:30 | <webben> | Then again, my best friend recently finished writing about a third of the catalogue for an eighteenth-century art exhibition. So I've got well-used to reading long descriptions ;) |
| 00:31 | <othermaciej> | Hixie: I find it useful for ego search :-) |
| 00:31 | <Philip`> | Hixie: Where "temporary" means "until we've solved AI"? :-) |
| 00:31 | <Hixie> | webben: that's another example of how the alt text for a photo should be visible (and isn't "alt text") |
| 00:31 | <othermaciej> | Hixie: or finding pictures of specific other people |
| 00:31 | <webben> | (It's one of the reasons I think that Lady of Shalott example isn't that good. In practice, images of the Lady are all interestingly different.) |
| 00:31 | <Hixie> | Philip`: right |
| 00:31 | <Hixie> | Philip`: a specific form of ai, image analysis, which is actually quite advanced |
| 00:31 | <Hixie> | Philip`: i wouldn't call it AI really |
| 00:32 | <webben> | Hixie: Well, not really. It's just that analysis of an image mixes in description of it. |
| 00:32 | <Hixie> | webben: the example in the spec is purely decorative, imho |
| 00:32 | <webben> | Strictly, you'd need yet more description to understand the image. |
| 00:32 | <webben> | I suppose it's partly a matter of print vs online conventions. In print, decorative paintings tend to be credited. |
| 00:33 | <Hixie> | webben: i don't honestly see that any amount of description can explain a photo or painting. |
| 00:33 | <Philip`> | People seem to still have problems doing image analysis to detect a human face in an image, when it's not facing straight at the camera |
| 00:33 | <webben> | Hixie: I think that's like saying: I don't see how a photo can represent the Grand Canyon. |
| 00:33 | <Hixie> | webben: indeed |
| 00:33 | <webben> | experiences are not reducible in their entirety |
| 00:33 | <Hixie> | webben: i agree |
| 00:33 | <webben> | but they are reducible enough that the reduction may be useful |
| 00:34 | <Hixie> | webben: but that reduction need not be hidden from those who can see the unreduced form |
| 00:34 | <Hixie> | webben: (especially since if it's not hidden from them, they can improve it, wiki style) |
| 00:35 | <webben> | Hixie: Well that's true. (And a strong argument for allowing authors to label explicitly visible text as an alternative/complement.) But not all designers will want all that text. |
| 00:35 | <Hixie> | webben: not all designers of what? |
| 00:35 | <webben> | pages |
| 00:35 | <Hixie> | webben: i thought we were talking about photo gallery pages like flickr |
| 00:35 | <webben> | Also, there's no reason why sighted people shouldn't have access to long descriptions on another page |
| 00:36 | <Hixie> | (or at least, sites where the image is the point) |
| 00:36 | <webben> | Hixie: Oh. I thought we were talking about alt's for photos generally. |
| 00:36 | <Hixie> | webben: another page is equivalent to hiding the text in the alt attribute, which i content is bad. |
| 00:36 | <Hixie> | contend |
| 00:37 | <webben> | Hixie: Well, not hiding the text can also be bad, because in practice when you write text alongside an image you risk assuming the image. |
| 00:37 | <othermaciej> | I think in the case of photo galleries, most text that would be useful to someone who can't consume the image would be useful to those who can as well |
| 00:37 | <webben> | (and hence failing to reduce the experience at all) |
| 00:37 | <webben> | othermaciej: probably. But not when e.g. in a slideshow. |
| 00:38 | <othermaciej> | and offering the ability to add visible text will probably in practice lead to better text than the ability to add normally hidden text |
| 00:38 | <Hixie> | webben: the rusk that the text will assume the image is far less imho than the risk that the text will be omitted altogether. |
| 00:38 | <Hixie> | risk |
| 00:38 | <Hixie> | in a slideshow you wouldn't show the text at all |
| 00:38 | <webben> | Hixie: well exactly, but you would want the alternative text. |
| 00:38 | <Hixie> | you also wouldn't expect a blind person to use a slideshow |
| 00:38 | <webben> | i don't see why not |
| 00:39 | <webben> | it's just a sequential progress through images |
| 00:39 | <Hixie> | it's a specifically visual representation! |
| 00:39 | <othermaciej> | flickr's slideshow feature doesn't show any of the text |
| 00:39 | <webben> | it's a sequence is all |
| 00:39 | <othermaciej> | the photostream is also a sequence |
| 00:39 | <webben> | yep |
| 00:39 | <Hixie> | the semantic might be a sequence, but the presentation is a visual one |
| 00:40 | <webben> | partly |
| 00:40 | <Hixie> | now another stylesheet might be aural, sure |
| 00:40 | <Hixie> | but then it's not a slideshow. |
| 00:40 | <othermaciej> | the reason for a slideshow to exist is that showing the images in large format on a single page is not convenient for visual consumption |
| 00:40 | <webben> | e.g. a lot of slideshows have an audio component |
| 00:40 | Hixie | doesn't understand the point webben is trying to make |
| 00:40 | <othermaciej> | lots of text at once vs. one bit of text at a time does not have the same kind of difference in usability |
| 00:40 | <webben> | othermaciej: well it can also be about the order, depending on the sort of slideshow. |
| 00:41 | <Hixie> | anyway |
| 00:41 | <Hixie> | i'm surprised nobody has yet mailed public-html quoting this and telling DanC that i'm clearly showing lack of care for the disabled and should be banned from the group |
| 00:42 | <Hixie> | the complainers are clearly not running at their usual speed today |
| 00:44 | <kingryan> | where's andy mabbett when you need him? |
| 00:47 | <kingryan> | ok, maybe that's only funny in the microformats channel |
| 00:48 | hober | laughed |
| 01:04 | <Philip`> | Lachy: About "IE 5.x [is] so insignificant, [it's] not even worth looking at": http://www.thecounter.com/stats/2007/August/browser.php suggests it's about as (in)significant as all versions of Opera combined |
| 01:05 | <webben> | Philip`: One possible issue with that is that they may not be efficiently filtering out bots that claim to be IE 5.5. |
| 01:05 | <webben> | (never mind opera masquerading as other browsers) |
| 02:42 | <Lachy> | Philip`, IE5.x is also obsolete, Opera isn't. |
| 02:44 | <othermaciej> | I like to look at this chart: http://marketshare.hitslink.com/report.aspx?qprid=6 |
| 02:44 | <othermaciej> | but there are other considerations like upward vs. downward trend, and share in particular markets |
| 02:45 | <othermaciej> | note, it shows Opera 9.x as significantly higher than IE 5.5 |
| 02:46 | <othermaciej> | I find the "thecounter" stats dubious |
| 02:46 | <othermaciej> | their 10% "Netscape comp." category seems too large even for an "other" total, let alone an actual different browser |
| 02:49 | <othermaciej> | http://marketshare.hitslink.com/report.aspx?qprid=0 seems pretty in line with the general view of the top 4 browser engines |
| 02:53 | <othermaciej> | incidentally wikipedia seems to agree with me that this particular usage share survey is the most credible: http://en.wikipedia.org/wiki/Comparison_of_web_browsers |
| 02:57 | <Lachy> | zcorpan, see RFC 3676, which obsoletes 2646. I think the delsp parameter is what you need when splitting long words, though I particularly wouldn't worry about it for our needs |
| 03:05 | <MikeSmith> | othermaciej - interesting to see how high up the Nintendo Wii is in the "Operating System Market Share" chart there |
| 03:05 | <MikeSmith> | http://marketshare.hitslink.com/report.aspx?qprid=2 |
| 03:07 | <MikeSmith> | and how up up the browser charts Opera Mini is, as well as other mobile browsers |
| 03:08 | <MikeSmith> | especially that they now have more share than particular versions of desktop browsers |
| 03:11 | <karlUshi> | http://www.xitimonitor.com/en-us/browsers-barometer/ie7-and-firefox-2-july-2007/index-1-2-3-103.html |
| 04:12 | <Lachy> | Hixie, do you have PHP4 or 5 running on status.whatwg.org? |
| 04:25 | <Hixie> | Lachy: yes |
| 04:25 | <Hixie> | (dunno which one) |
| 04:26 | <Lachy> | hmm. the response header tells me: Server: Apache/2.0.54 (Unix) PHP/4.4.7 mod_ssl/2.0.54 OpenSSL/0.9.7e mod_fastcgi/2.4.2 DAV/2 SVN/1.4.2 |
| 09:09 | <Lachy> | zcorpan, yt? |
| 09:12 | <Lachy> | zcorpan, I completely rewrote the emailing code. made it object oriented, dealt with the word wrapping and fixed a bunch of other stuff |
| 10:30 | <jwalden> | there's a lunar eclipse right now if you're in CA |
| 10:31 | <Lachy> | also if you're in eastern Aus |
| 10:35 | <krijnh> | Also if you're in NL |
| 11:34 | <zcorpan> | Lachy: i did some changes to the php yesterday, but utf8_wordwrap() still needs a 4th argument |
| 11:34 | <Lachy> | utf8_wordwrap doesn't work well at all, I did some testing with it and it's not needed |
| 11:34 | <zcorpan> | aha |
| 11:34 | <zcorpan> | ok |
| 11:35 | <zcorpan> | wordwrap() is ok? |
| 11:35 | <Lachy> | the existing wordwrap function seems to work reasonably well |
| 11:35 | <zcorpan> | good then |
| 11:35 | <Lachy> | there's only one bug that I found, but it's never likely to occur |
| 11:35 | <zcorpan> | split inside a multibyte character? |
| 11:36 | <Lachy> | if you have a long line longer than 990 characters, without any white space, and the bytes 990-991 are part of a multibyte char |
| 11:36 | <zcorpan> | right |
| 11:36 | <zcorpan> | the url will be ascii only |
| 11:36 | <Lachy> | yeah |
| 11:36 | <zcorpan> | and the rest won't contain that long lines |
| 11:37 | <Lachy> | since it gets urlencoded |
| 11:37 | <zcorpan> | yeah even that |
| 11:37 | <zcorpan> | ok |
| 11:53 | Lachy | is going to write some twitter posting code |
| 13:56 | <Lachy> | PHP is annoying, at least the way it's configured on my server. I need to use constructors and destructors in PHP5, but the curl functions only work when I use PHP4 :-( |
| 21:53 | <zcorpan> | Lachy_: the DelSp parameter was not added to handle lines longer than 998 characters. i'm still not sure what generating agents are supposed to do with such lines |
| 22:09 | <markp> | kingryan: i've checked in revision 962 of html5lib |
| 22:09 | <kingryan> | markp: I see that |
| 22:10 | <kingryan> | I'll updated the ruby version this evening |
| 22:10 | <markp> | ok |