| 00:03 | <Lachy> | Hixie, http://hg.gsnedders.com/anolis/ |
| 00:03 | <Hixie> | thanks |
| 00:03 | <Hixie> | but too late |
| 00:03 | <Hixie> | already sent the mail :-) |
| 00:11 | <Hixie> | <object data=alert.html><object data=alert.html></object></object> alert.html: <script>alert('x')</script> |
| 00:11 | <Hixie> | how many alerts? |
| 00:15 | <Hixie> | wow |
| 00:15 | <Hixie> | IE8's <object> actually moves the markup that's in the <object> into an altHtml="" attribute |
| 01:36 | <Philip`> | http://philip.html5.org/misc/spec-links.html - I don't know what that is meant to show, but maybe some information can be extracted |
| 01:36 | <Philip`> | ...except that it shows Opera doesn't support stroke="rgba(...)", so please use something more like Firefox |
| 01:37 | <Philip`> | (Also it shows repaint bugs in Firefox if I scroll rightwards then leftwards since bits of text go missing) |
| 01:38 | <Philip`> | (but never mind that) |
| 01:43 | <Hixie> | nice! |
| 01:44 | <Hixie> | not sure what to make of it, but nice nonetheless :-) |
| 01:46 | <Lachy> | Philip`, why is it missing lines from the Table of Contents to every other section? |
| 01:47 | <Philip`> | Lachy: I intentionally omitted those lines, because they're uninteresting and would probably make it uglier |
| 01:47 | <Lachy> | it doesn't indicate the direction of the links |
| 01:48 | <Philip`> | How could it indicate the directions? |
| 01:49 | <Lachy> | by colouring the lines from one section linking to another, on :hover or something, so that the coloured lines mean a link from that section, and the other lines mean to that section |
| 01:50 | <Philip`> | That sounds kind of complicated |
| 01:50 | <Lachy> | I know, but your talented enough to waste time doing it :-) |
| 01:51 | <Philip`> | If it requires more than two lines of Perl, I probably won't bother :-) |
| 02:00 | <Hixie> | you could make the direction more obvious using a gradient colour |
| 02:00 | <Hixie> | red to black |
| 02:00 | <Hixie> | or some such |
| 02:01 | <Philip`> | Is it trivial to do gradients like that in SVG? |
| 02:01 | <Hixie> | is anything trivial in svg? |
| 02:01 | <Hixie> | other than getting a well-formedness error |
| 02:01 | <Philip`> | This thing has been fairly trivial SVG so far :-) |
| 02:02 | <Hixie> | gradients aren't that hard, and you like to them as fill:url(#id) iirc |
| 02:02 | <Hixie> | dunno if it'll rotate the gradient though |
| 02:04 | <Hixie> | i love how the response to anyone volunteering to do anything here is always more work, btw :-P |
| 02:05 | <Hixie> | i did consider suggesting showing teh direction but i think in general the direction doesn't matter |
| 02:05 | <Hixie> | since what matters it that there is a link at all |
| 02:06 | <Lachy> | the gradients probably won't work out too well because of all the overlapping lines |
| 02:07 | Philip` | just reused an old Python script he had that extracted the links between sections, and then wrote a quick Perl script to translate that into SVG |
| 02:08 | <Philip`> | (The line widths are 1+log(number of links)) |
| 02:11 | <Philip`> | Google's SearchWiki has a lot of very useful comments, e.g. "wikicheese.", "This is a good article about cheese!", "sucks", "i like cheese", "Mmmmm, cheese", "Cheese is great" |
| 02:13 | <blooberry> | philip`: I love this quote: "If it requires more than two lines of Perl, I probably won't bother" That's almost a Perl mantra in the making |
| 07:25 | <annevk2> | http://crisp.tweakblogs.net/blog/1221/tweakers-punt-net-en-internet-explorer-8-compatibility.html o_O |
| 09:56 | <Hish> | Hixie: ping |
| 09:56 | <Hixie> | hey |
| 10:00 | <annevk2> | sound of Standards Suck recording on the Uhuru Peak is actually not that bad, even while it's done with a simple digital camera rather than typical video recording hardware |
| 10:00 | <annevk2> | only 20s though |
| 10:00 | <Hish> | Hixie, I'm trying to change my OpenOffice > html converter from html 4 to 5 by changing all divs to more meaningful tags. So, currently I'm thinking about footnotes, text frames and absolute positioned graphics. For footnotes I'm about to use the display tag. Then I want to wrap graphics and text frames with the figure tag. does this make sense? |
| 10:01 | <Hixie> | display tag? |
| 10:01 | <Hixie> | footnotes have a section in the html5 spec describing three ways to do them depending on exactly what you want |
| 10:01 | <Hixie> | generally you won't be able to do a good translation from a presentational format to html5 without AI (or real I, e.g. with human intervention) |
| 10:02 | <Hish> | sorry, the details tag |
| 10:02 | <Hixie> | no, footnotes aren't like <details> |
| 10:02 | <Hixie> | <details> is for like disclosure triangles in dialogs |
| 10:02 | <Hish> | ah ok. so' |
| 10:02 | <Hish> | so I'll search the footnore section in the spec... |
| 10:02 | <Hixie> | http://www.whatwg.org/specs/web-apps/current-work/#footnotes |
| 10:03 | <Hish> | my typing today doesn't work... |
| 10:04 | <Hish> | yep. your link makes much more sense than details. |
| 10:59 | <Hixie> | is there any reason i shouldn't be replying to table feedback soon? |
| 10:59 | <Hixie> | jgraham? |
| 11:34 | <jgraham> | Hixie: You don't want to spark another long discussion? ;) |
| 11:34 | <jgraham> | (Seriously I don't know of a reason, although Ben i working on his comparison of the current alorithm and other alternatives) |
| 11:35 | <jgraham> | s/i/is/ |
| 12:02 | <Hixie> | ok i'll wait til he's done with his comparison then |
| 13:38 | <BenMillard> | Hixie, jgraham is right that I'm doing a comparison document |
| 13:38 | <jgraham> | BenMillard: I think I fixed the bugs that you found |
| 13:38 | <jgraham> | in the table inspector |
| 13:39 | <BenMillard> | jgraham, cool |
| 13:39 | <BenMillard> | am I right in saying both algorithms need the whole table to be present before they can run? (i.e. they don't run incrementally while the table cells come in over the wire) |
| 13:42 | <jgraham> | BenMillard: In principle I believe either algorithm could be written to run incrementally |
| 13:43 | <jgraham> | Because IIRC given a cell, the headers that apply to that cell are always before it in document order |
| 13:44 | <jgraham> | This would change if ltr/rtl considerations ment that we sometimes had to look to the right of a cell in order to finds its headers |
| 13:45 | <jgraham> | I don't know if bidi issues have been given serious consideration yet (did you look at any rtl tables?) |
| 13:45 | <BenMillard> | jgraham, so you'd apply the header as far as it can go given how much of the table you currently have...then some more of the table comes in so you continue applying those headers until they reach the end of their scope? |
| 13:46 | <jgraham> | Oh and I guess @headers breaks incremental assignment |
| 13:46 | <BenMillard> | jgraham, I've not looked at RTL tables but I assume the source order follows the same principles as in LTR, it's just the rendering is reversed |
| 13:47 | <BenMillard> | jgraham, I guess headers+id would try applying to the cells you currently have, if there are no matches then you keep receiving cells and trying again until you find a match or the table ends? |
| 13:48 | <jgraham> | BenMillard: Well if I were implementing this in a browser, I would invert the whole algorithm so that it would be "given data cell D which of the other cells in the table form the set of headers for C?" |
| 13:48 | <jgraham> | See also the mail from Aaron Leventhal |
| 13:48 | <BenMillard> | yeah, Aaron Leventhal sent feedback about that |
| 13:48 | <Philip`> | Could you invert the whole algorithm without introducing a huge pile of bugs? |
| 13:48 | hsivonen | just replied to Aaron's email |
| 13:49 | <BenMillard> | (the thread is here: http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2008-October/thread.html#16915) |
| 13:50 | <jgraham> | Philip`: I /think/ the algorithm is not hard to invert. But I have not tried |
| 13:50 | <hsivonen> | I think it would make a lot of sense to invert the algorithm |
| 13:51 | <BenMillard> | HTML4 uses the opposite direction for it's auto algorithm |
| 13:51 | <hsivonen> | (this is the third time, IIRC, that implementation feedback calls for the inversion of an algorithm in HTML5, AFAICT) |
| 13:51 | <BenMillard> | but authors write tables from the perspective of "I want this header cell to apply in this direction" |
| 13:52 | <jgraham> | I am vaugely considering implementing the next iteration of the HTML5 algorithm in js in such a way that it adds aria attributes to tables |
| 13:53 | <BenMillard> | Aaron is right that ATs query a cell for its related header cells, but that happens after those associations are made by the UA |
| 13:53 | <jgraham> | So I guess if I do that I can try inverting the algorithm |
| 13:53 | <jgraham> | BenMillard: I guess Aaron wants to avoid precomputing all the information when it may not be needed, which could be a serious performance hit for large tables |
| 13:54 | <hsivonen> | BenMillard: presumably the UA would compute the association on demand for one data cell instead or rerunning the algorithm and remembering the result every time the table is mutated |
| 13:54 | <jgraham> | Also I agree that an authoring guide would be better written from the same point of view as the current spec |
| 13:55 | <BenMillard> | jgraham & hsivonen, my impression is UAs expose everything to the accessibilty APIs once the page has loaded, or while it is being loaded, so that's what I thought header association should do |
| 13:55 | <BenMillard> | but if the AT is expecting to trigger the UA to do the association lookup per-cell, then inverting the algorithm makes sense |
| 13:56 | <jgraham> | (this is probably another problem with having a normaive language reference that is different from the main spec because it would be tempting to have the normative text describing how associaions are made, and the text describing the required UA processing be from different points of view) |
| 13:57 | <BenMillard> | also, from the user videos I've seen, a typical user scenario is skim-reading through a data table at high speed |
| 13:58 | <BenMillard> | so I thought having all the associations ready and waiting was better for the user, as well as more consistent with how such mappings are done elsewhere in the document |
| 13:59 | <hsivonen> | BenMillard: do you have data on how common it is that authors copy and paste tables within a page without making ids unique within the page? |
| 13:59 | <BenMillard> | hsivonen, I haven't seen that but Hixie has |
| 14:00 | <hsivonen> | ok |
| 14:01 | <jgraham> | hsivonen: Aaron seems to disallow arcs beween tables in his proposed rewording |
| 14:01 | <hsivonen> | oops. |
| 14:02 | <BenMillard> | Aaron Leventhal's rewording is this, right? "If there is an element in the document with a corresponding ID (via getElementById) equal to the value of /id/, and it is a header cell in the current table, then assign it to the data cell." |
| 14:04 | <BenMillard> | hsivonen, should your reply be listed at the end of that thread or am I looking in the wrong place? http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2008-October/thread.html#16915 |
| 14:07 | <jgraham> | BenMillard: I think that only considers replies from the same month |
| 14:08 | <BenMillard> | jgraham, oh of course! lost track of the month I'm on :P |
| 14:16 | <BenMillard> | jgraham, does your Smart Span implementation check whether a cell is empty? Do such cells change the way it works? |
| 14:16 | <BenMillard> | s/Smart Span/Smart Headers/ |
| 14:16 | Philip` | likes the craziness of how a Perl database interface module relies on the Lingua::EN::Inflect module |
| 14:17 | <jgraham> | BenMillard: No. |
| 14:18 | <jgraham> | (the HTML5 algorithm does however. It is not hard to add) |
| 14:18 | hsivonen | wonders if the CPAN deps are tree-like or if there are circular deps |
| 14:21 | <Philip`> | hsivonen: I don't think there's anything to stop circular dependencies, except convention and bug reports |
| 14:22 | <Philip`> | (You can always force a module to install, ignoring its dependencies and its failing tests, but it's usually a bug if you have to do so) |
| 14:22 | <jgraham> | Does CPAN no cope with circular deps? |
| 14:22 | <jgraham> | not |
| 14:23 | hsivonen | learns new things (to him) about C++ link errors |
| 14:23 | <hsivonen> | specifically, one can have two .o files that link together into an .a but not into a shared lib |
| 14:23 | <BenMillard> | jgraham, HTML5 says "User agents may remove empty data cells when analyzing data in a table." |
| 14:24 | <hsivonen> | I guess Steve Yegge is right that linker errors are a top thing of the C++ brand |
| 14:24 | <Philip`> | jgraham: No - when you tell it to install a module, it recursively asks whether you want to install its missing dependencies, and then it compiles and then runs the tests (and the tests will fail if you've got missing dependencies) and then installs |
| 14:25 | <jgraham> | Oh. Well that seems like it could be avoided |
| 14:26 | <jgraham> | BenMillard: HTML 5 treats <th>Foo<td><td> as equivalent to <th colspan=3>Foo |
| 14:27 | <jgraham> | but does not do the same vertically |
| 14:27 | <BenMillard> | jgraham, in a table with 3 columns, yes |
| 14:28 | <BenMillard> | but the "may remove empty data cells" thin seems like a bad idea because users would easily become lost in tables with many empty cells |
| 14:30 | <BenMillard> | in contrast, ignoring empty header cells seems like a good thing (such as an empty <th> in the top left corner of a table with both column headers and row headers) |
| 14:33 | <jgraham> | BenMillard: I wasn't sure what the implications of "removing empty data cells" was supposed to be which I guess is a sign that it sould be clarified |
| 14:33 | <jgraham> | s/cells"/cells when analysing data in a table"/ |
| 14:35 | <BenMillard> | jgraham, my impression was it meant UAs don't have to associated header cells with empty data cells? |
| 14:35 | <BenMillard> | s/associated/associate/ |
| 14:36 | <jgraham> | Also I am unsure why the wide cell thing only works if the cell extends to the edge of the table |
| 14:36 | <jgraham> | or raher if they empty cells extend to the edge of the table |
| 14:37 | <jgraham> | BenMillard: Maybe. If so, the "may" makes the algorithm non-deterministic and the scope should probably be limited to data cells that fom part of a wide cell |
| 14:41 | <BenMillard> | jgraham, I think the wide cell thing is for "section headers" which should use <th colspan="width_of_table">Foo</th> but instead used <th>Foo<td><td><td> to reach the edge of the table |
| 14:41 | <BenMillard> | it makes the gamble that a row header won't be given for a row where there was no data to report (which seems fairly safe) |
| 14:42 | <jgraham> | BenMillard: But that could just as easilly happen with a heading that extended across only part of the table, no? |
| 14:42 | <jgraham> | <th>Foo<td><td><th>Bar |
| 14:42 | <BenMillard> | jgraham, indeed. |
| 14:43 | <jgraham> | I don't understand your row header comment |
| 14:45 | <BenMillard> | jgraham, <tr><th>2006<td>5.6<td>8.3<tr><th>2007<td><td><tr><th>2008<td>7.3<td>9.8 |
| 14:45 | <BenMillard> | no data is present in the row for "2007", but it's not a section header |
| 14:46 | <Philip`> | Hmm, Ubuntu (Gutsy) has .atom mapped to "application/atom" in mime.types, which breaks Opera and Firefox because it should have "+xml" |
| 14:47 | <BenMillard> | jgraham, usually a row where no data is available has a character (http://www.avert.org/aofconsent.htm) or a note but they could just as logically be empty |
| 14:47 | <Dashiva> | Philip`: Good thing there's extensive testing to detect things like that |
| 14:48 | <jgraham> | BenMillard: Oh I see. Interesting |
| 14:49 | jgraham | decides to delay trying to formulate an email covering all of this until later |
| 14:54 | <Philip`> | I thought I was being cleverly lazily by using file extensions so I didn't have to look up the Apache documentation on how to override content-types, but sadly I was thwarted |
| 14:54 | <Philip`> | s/lazily/lazy/ |
| 14:54 | <BenMillard> | jgraham, it's worth noting that an AT can announce "Empty" or somesuch for truely empty cells but has to say "Question" or "Dash" if there's a random character, just in case that character is significant |
| 14:55 | <hsivonen> | Philip`: is that an Apache bug, a Debian bug or an Ubuntu bug? |
| 14:55 | <hsivonen> | Philip`: bad brokenness in any case |
| 14:55 | <BenMillard> | jgraham, so it's actually better for users if empty rows are truely empty, which is incompatible with the "wide header cell" thing in HTML5 when such rows have a row header cell |
| 14:55 | <hsivonen> | me wonders how static array of ints can reference a static pointer |
| 14:57 | <hsivonen> | the linker is telling me that the pointer thingy is an undefined symbol referenced from what is a static array of ints |
| 14:58 | <BenMillard> | jgraham, it's also worth noting that it would compatible with some real tables, such as Table 4.1 here, if they used <th> for their header cell: http://www.federalreserve.gov/generalinfo/basel2/docs2005/potentialimpact.htm |
| 14:58 | <BenMillard> | jgraham, although if they retrofitted <th> to their header cells, they could retrofit appropriate colspan at the same time to make these proper section headers. :) |
| 15:02 | <BenMillard> | jgraham, I think implying <td><b> and <td><strong> as aliases of <th> messes up reasonable tables, so we should drop that heuristic. |
| 15:05 | <BenMillard> | jgraham, most frustrating case I've found is <td><b> both for row headers and for important data, meaning the <td><b> heuristic would get the row headers right but would erroneously apply "5.8 mm" as a column header as well: http://flickr.com/photos/adactio/1383011861/meta/in/photostream/ |
| 15:06 | <BenMillard> | jgraham, could the multi-table support only pick the innermost tables? the above page causes "Table 3" to have the layout table around the data table: http://james.html5.org/tables/table_inspector.py?input_type=type_uri&uri=http%3A%2F%2Fflickr.com%2Fphotos%2Fadactio%2F1383011861%2Fmeta%2Fin%2Fphotostream%2F&source=&algorithm=smartheaders |
| 15:07 | <BenMillard> | jgraham, it's available further down at Table 4 and Table 5, so removing Table 3 due to it containing another table seems possible |
| 15:09 | <Philip`> | hsivonen: The file says "This file is part of the "mime-support" package. Please send email (not a bug report) to mime-support⊙pdo if you would like new types and/or extensions to be added." which makes it sound like Debian |
| 15:10 | <hsivonen> | Philip`: I suggest sending email to that address, then |
| 15:10 | <Philip`> | (Maybe it's fixed in newer versions, but I can't update this Ubuntu because apt-get gives errors while trying to download stuff) |
| 15:17 | Philip` | is surprised that code he wrote two years ago still actually runs, without any library incompatibilities or similar problems, having reinstalled it on a more modern system |
| 15:17 | <Philip`> | despite using the Catalyst framework and various crazy modules |
| 15:36 | <hsivonen> | http://twitter.com/asmitter/statuses/1018146131 |
| 16:21 | <Lachy> | http://www.alistapart.com/articles/thisishowthewebgetsregulated |
| 16:22 | <BenMillard> | ALA's URLs get worse with each issue |
| 16:23 | <jgraham> | BenMillard: Tables with descendant tables are no longer shown. If this causes problems in practice, I will revert the change |
| 16:33 | <BenMillard> | jgraham, that seems to work. :) |
| 16:35 | <Lachy> | I completely disagree with Joe Clark's recommendation to use open captioning instead of closed captioning |
| 16:36 | <Philip`> | Why? |
| 16:37 | <Lachy> | what's really needed is a decent, widely supported, standardised closed captioning/subtitling format that solves all the technical problems with the existing mess of formats |
| 16:38 | <Dashiva> | srt! |
| 16:38 | <Lachy> | so that one closed caption/subtitle file will work regardless of the video codec or device |
| 16:38 | <Lachy> | Dashiva, SRT is a very limited format |
| 16:39 | <Philip`> | Lachy: Won't that still have the same problems about unpredictability of fonts and of requiring extra UI to enable the captions? |
| 16:41 | <Lachy> | we also need a better selection of fonts available, which can be done by providing some guidelines about what fonts that media player vendors/device manufacturers should use |
| 16:41 | <Lachy> | and get them all to standardise on using UTF-8 |
| 16:42 | <Lachy> | a decent subtitle format needs to be able to specifiy fonts, colours, position and timing of text |
| 16:44 | <Lachy> | the problem with open captions is that it requies creating 2 or more copies of the same video. One without captions and one or more with captions or subtitles in various languages |
| 16:47 | <BenMillard> | Lachy, http://screenfont.ca/fonts/today/ and http://screenfont.ca/tech/ are Joe Clark's professional opinion on fonts and Unicode in captioning and subtitles (http://screenfont.ca/about/) which seem close to your own ("New fonts", "A single Unicode encoding") |
| 17:48 | <Lachy> | http://lists.w3.org/Archives/Public/www-archive/2008Nov/0028.html |
| 17:48 | <Lachy> | http://lists.w3.org/Archives/Public/www-archive/2008Nov/0029.html |
| 17:50 | <Lachy> | it's interesting to see how much Dean's opinion has changed so dramatically in the past 12 months |
| 18:39 | <yecril71> | I consider a major disadvantage of HTML (and XML also) that locality of identifiers is not supported. |
| 18:40 | <yecril71> | It effectively forces page authors to use GUIDs for identifiers in templates. |
| 18:40 | <yecril71> | And those GUIDs have to be regenerated each time a template is used, |
| 18:41 | <yecril71> | if only there is a chance that a template is used twice on the same page. |
| 18:41 | <yecril71> | The situation is mildly alleviated by the fact that a NAME is local to a FORM. |
| 18:42 | <yecril71> | Not all problems can be handled in this way though. |
| 19:18 | <gsnedders> | Hixie: Why the asking about Anolis? |
| 19:19 | <gsnedders> | Hixie: http://anolis.gsnedders.com fwiw |
| 19:19 | <gsnedders> | (though it was all ready too late when Lachy gave the wrong answer) |
| 20:27 | <gsnedders> | Why do banks use things for security that are entirely forgettable? |
| 20:29 | <Dashiva> | Because they predate the internet? |
| 20:30 | <gsnedders> | I'm expected to give the 5th, 3rd, and 2nd digit (in that order) of "my" passnumber |
| 20:30 | <gsnedders> | Like I can remember it at all |
| 20:33 | <Philip`> | Maybe it's your date of birth |
| 20:35 | <gsnedders> | Philip`: no |
| 20:39 | <jwalden> | easy! 5, 3, 2 |
| 20:41 | <Philip`> | gsnedders: Oh, okay; my passnumber is your date of birth, so I thought it'd be a good guess |
| 20:41 | <gsnedders> | Philip`: Big/little endian? |
| 20:42 | <Philip`> | gsnedders: If I told you that, you'd be able to steal all my money, so I think I'll pass |
| 20:43 | gsnedders | thinks he knows the first two digits |
| 21:15 | <theanxy> | wii Lachy |
| 21:17 | <theanxy> | sorry. |
| 23:07 | <Lachy> | "It is easy to make the case that the general video uploader would find it an undue hardship to caption their videos. Undue hardship is a legitimate exemption in every relevant piece of legislation. So let’s assume that you the video uploader wouldn’t have to do it. Fine. |
| 23:07 | <Lachy> | "But what about your host? Google could afford to do it for you, though they wouldn’t want to. Should they have to?" |
| 23:07 | <Lachy> | http://www.alistapart.com/comments/thisishowthewebgetsregulated?page=1#10 |
| 23:07 | <Lachy> | Clearly, Joe Clark doesn't have a clue about just how much video content is uploaded to YouTube and Google Video |
| 23:08 | <Lachy> | last I heard, it was something like 10 minutes worth of video uploaded every minute |
| 23:09 | <Lachy> | somehow I think captioning 240 hours of video every day would be impossible for any organisation |
| 23:14 | <Philip`> | http://news.bbc.co.uk/1/hi/uk/7730679.stm - "13 hours every minute" |
| 23:14 | <Philip`> | You're off by a few orders of magnitude :-) |
| 23:20 | <Lachy> | woah |
| 23:21 | <Philip`> | They must have quite a few servers doing video transcoding |
| 23:22 | <Philip`> | It would work fine if Google just farmed out the captioning jobs to the Amazon Mechanical Turk |
| 23:23 | <Philip`> | Offer $0.02 to caption a minute of video, and it'll only cost ~$8M/year |
| 23:25 | <Lachy> | yeah, cause captioning 1 minute of video would take someone about 10-15 minutes worth of work for an expert (including transcribing it and syncing up the text, etc.) and someone's time is really only worth $0.10/hour anyway |
| 23:27 | <jgraham> | I'm not sure you could find ~10,000 people to caption video for that much |
| 23:27 | <Philip`> | These are not just "someone" - they're someone who's decided to scrounge for money on Mechanical Turk, so they're not going to have very high expectations |
| 23:28 | <jgraham> | Even if you could pay 100 times that, which would make the cost prohibitive |
| 23:28 | <Philip`> | And Google could cut down the work hugely by detecting videos with pretty much the same audio track as others, since they're mostly songs or clips from TV shows posted a zillion times |
| 23:29 | <Philip`> | and you don't really need to get a thousand people to independently caption Rick Astley if you can figure out that the repeats |
| 23:29 | <Lachy> | Google should just develop speech recognition and produce captions on the fly |
| 23:30 | <Philip`> | Hmm, but those would only be subtitles and not captions |
| 23:31 | <svl> | Google should just heal every possible kind of auditory impairment people could suffer from, so captioning wouldn't be necessary anymore. |
| 23:31 | <Lachy> | except for people who need to keep the volume turned down low for some reason |
| 23:31 | <jgraham> | I wonder what fraction of the content on youtube is duplicates |
| 23:32 | Philip` | has probably already mentioned several times that he played Portal with no speakers or headphones, and its captioning worked pretty nicely |
| 23:33 | <Philip`> | (complete with textual versions of gameplay-relevant sound effects) |
| 23:36 | jgraham | wonders why Rob Sayre thinks that his issue with the parser spec indicates a faliure of process |
| 23:42 | <Philip`> | It seems he is saying that since it is not perfect, that invalidates claims that the way it is written is better than other ways |
| 23:46 | <Lachy> | he must be completely ignoring the fact that in its current state, it's still better than the state of any other prior spec written using alternative methods |
| 23:46 | <Philip`> | http://philip.html5.org/misc/portal-caption.jpg |
| 23:47 | <Philip`> | although, oddly, the audio says "and then there will be" in place of "[garbled]" |
