| 03:09 | <Hixie> | wtf is "next column" in 3.15.11.1.9.1 ?? |
| 03:09 | <Hixie> | i really have to pay more attention when writing up these algorithms |
| 05:36 | <Hixie> | i found tens of thousands of documents last modified in 1990 that claim to have class=MsoNormal |
| 05:36 | <Hixie> | sigh |
| 05:57 | <duryodhan> | where? you googled? |
| 05:58 | <Hixie> | wow, the sheer number of pages with screwed-up markup is amazing |
| 05:58 | <Hixie> | duryodhan: part of my research |
| 05:58 | <duryodhan> | ohh ok |
| 05:58 | <duryodhan> | the sheer number of screwed up markup ... heh |
| 05:59 | <duryodhan> | take today ... I am a developer who wants to make a site ... |
| 05:59 | <duryodhan> | what should I do? |
| 05:59 | <duryodhan> | XHTML5, XHTML 1,2 , HTML4 , HTML Forms. WebForms 1.0 , XForms , use DOM to check forms or use XForms .... |
| 05:59 | <duryodhan> | ? |
| 06:00 | <duryodhan> | I mean, probably a decade later many of these will become obselete |
| 06:00 | <duryodhan> | and someone will look on them as docs with screwed up markup ... |
| 06:00 | <Hixie> | today? just use HTML4. |
| 06:00 | <duryodhan> | ok |
| 06:00 | <duryodhan> | today is a little .... |
| 06:00 | <Hixie> | XHTML5 isn't done yet, XHTML 1 isn't supported by IE, XHTML2 isn't done yet, XForms doesn't work in browsers |
| 06:01 | <duryodhan> | why do we have web forms as well as XForms ? |
| 06:01 | <Hixie> | dunno, ask the xforms guys |
| 06:01 | <Hixie> | it's pretty common for groups of people to try to invent new replacement technologies |
| 06:01 | <duryodhan> | (XForms using something like orbeon) |
| 06:01 | <Hixie> | sometimes new technologies take off |
| 06:02 | <Hixie> | (most often they don't) |
| 06:02 | <duryodhan> | heh... haven't you guys made web forms ? |
| 06:02 | <Hixie> | web forms is just a fancy name for what html4 does |
| 06:02 | <Hixie> | web forms 2 is just the next revision of html4 forms |
| 06:02 | <Hixie> | it's part of html5 |
| 06:02 | <duryodhan> | k |
| 06:02 | <Hixie> | the name "web forms 2" is likely to die a peaceful death |
| 06:02 | <duryodhan> | heh, isn't that a whatwg spec ? webforms 2? |
| 06:03 | <Lachy> | duryodhan: yes, but it will be integrated into the larger HTML5 spec in the future |
| 06:04 | <Lachy> | assuming the xforms guys don't get in the way and try to push their xforms transitional instead |
| 06:04 | <duryodhan> | my point is ... with so many specs ... many docs will still be again with "screwed up markup" in a decade |
| 06:04 | <duryodhan> | so we really haven't learnt from our past ... |
| 06:04 | <Lachy> | screwed up markup isn't caused by having many specs |
| 06:04 | <duryodhan> | XForms guys are putting webforms 2 as a Xforms-minimal or something ... |
| 06:05 | <Lachy> | they're working on "XForms Transitional", I believe |
| 06:05 | <duryodhan> | yeah |
| 06:05 | <Hixie> | so i found over 100000 html files with a last-modified date of 1988. |
| 06:06 | <Hixie> | wtf |
| 06:06 | <duryodhan> | well I was looking at developing a way for forms to have digital signatures ... and I still don't know a good way :( |
| 06:06 | <jruderman> | Hixie: time travel. how many from 1987? |
| 06:06 | <duryodhan> | there are just so many possibilities .. |
| 06:07 | <Hixie> | jruderman: around the same |
| 06:07 | <Lachy> | that's really weird, I find it hard to believe there are that many servers out there with clocks set so wrongly |
| 06:07 | <Hixie> | jruderman: i even found hundreds from 1662. |
| 06:07 | <jruderman> | that's impressive |
| 06:08 | <Hixie> | yes. |
| 06:08 | <duryodhan> | Hixie: 1662? :O |
| 06:08 | <Lachy> | I'm sure the internet archive will be very interested in those historical documents :-) |
| 06:09 | <Lachy> | Found anything from BCE? |
| 06:10 | <duryodhan> | Lachy: they were all used for writing "the da vinci code " |
| 06:10 | <Hixie> | my methodology was to look for the first 4 digit number in the last-modified field that was not preceded or suceeded by numbers or a + |
| 06:10 | <Hixie> | so i couldn't find anything BCE |
| 06:10 | <Hixie> | wouldn't surprise me htough |
| 06:10 | <Lachy> | what if it was preceded by a - > |
| 06:10 | <Lachy> | ? |
| 06:11 | <Hixie> | i'd take it but ignore the - or > |
| 06:12 | <Lachy> | I meant to write "what if it was preceded by a '-'?" Ignore the '>' |
| 06:12 | <Hixie> | i treated -s like spaces |
| 06:12 | <Hixie> | even more impressive is the thousands of files from dates greater than 2010 |
| 06:12 | <Lachy> | oh, then you might have counted timezone offsets as years |
| 06:13 | <Hixie> | i got a million or more from 2099 |
| 06:13 | <Hixie> | and almost 200,000 from max time_t |
| 06:13 | <Lachy> | so is the conclusion that we can draw from this, that the last modified date is completely useless? |
| 06:15 | <Lachy> | well, I suppose it's still somewhat useful for setting the If-last-modified-since HTTP header |
| 06:16 | <Hixie> | also a lot from 2099 and 2100 (like, over a million and over half a million respectively) |
| 06:18 | <Lachy> | is there any correlation between dates the use the correct format and those that have plausible dates? |
| 06:20 | <Lachy> | my hypothesis would be that servers that are configured to send the right date format are more likely to be configured with more accurate dates, and the others are just broken and unreliable. |
| 06:21 | <Hixie> | i dunno |
| 06:21 | <Hixie> | there were a LOT of different formats |
| 06:21 | <Hixie> | like, thousands |
| 06:21 | <Hixie> | the spec allows three |
| 06:21 | <Hixie> | which maps to about 10 actual formats |
| 06:21 | <Lachy> | which spec? |
| 06:21 | <Hixie> | http |
| 06:21 | <Lachy> | ok |
| 06:21 | <Lachy> | I thought it only allowed one. |
| 06:21 | <Hixie> | nope |
| 06:22 | <Hixie> | it defines three formats, all retarded |
| 06:23 | <Lachy> | oh, I see |
| 06:23 | <Lachy> | Sun, 06 Nov 1994 08:49:37 GMT ; RFC 822, updated by RFC 1123 |
| 06:23 | <Lachy> | Sunday, 06-Nov-94 08:49:37 GMT ; RFC 850, obsoleted by RFC 1036 |
| 06:23 | <Lachy> | Sun Nov 6 08:49:37 1994 ; ANSI C's asctime() format |
| 06:24 | <Hixie> | yeah |
| 06:24 | <Hixie> | so the second one, i ignored |
| 06:25 | <Hixie> | http://junkyard.damowmow.com/284 is an older version of this data btw |
| 06:25 | <Lachy> | the first is at least somewhat sensible |
| 06:26 | <Lachy> | the last one is just a really strange order |
| 06:26 | <Hixie> | the junkyard one doesn't exclude the + character |
| 06:27 | <Hixie> | hence all the low numbers |
| 06:27 | <Hixie> | and the peaks at 100s |
| 06:28 | <Lachy> | how do you explain the peaks at dates like 1428? |
| 06:29 | <Lachy> | 1969-70 can be explained because that's the epoch |
| 06:30 | <Hixie> | yeah 2038 can be explained too |
| 06:30 | <Lachy> | yeah, the max 32bit time |
| 06:30 | <Hixie> | #### only means thousands, so it could just be one misconfigured site |
| 06:31 | <Hixie> | 2250 i don't get |
| 06:33 | <Lachy> | so how many does ########## represent (the 2007 value) |
| 06:35 | <Lachy> | I'm surprised there aren't peaks at years like 0030, 0130, 0230, etc. for timezone offsets |
| 06:35 | <Hixie> | # = one order of magnitude |
| 06:36 | <Hixie> | there aren't that many :30 TZ offsets |
| 06:39 | <Lachy> | ok |
| 06:41 | <Hixie> | ########## is in the billions |
| 06:42 | <Lachy> | aargh! It really annoys me how some people conflate making content accessibile with providing fallback to those without the necessary software |
| 06:44 | <Hixie> | the most annoying thing for me in public-html is the way most people jump to a solution rather than determining the problem |
| 06:46 | <Lachy> | yeah, that too. I tried getting people to focus on the problem months ago, and it didn't really work then, and still not working now |
| 06:47 | <Lachy> | like in the whole headers="" debate, I tried to talk about how we could make tables accessible without needing headers, and basically got accused of ignoring the needs of the accessibility community |
| 06:47 | <Hixie> | yeah |
| 06:48 | <Hixie> | it's ridiculous |
| 06:48 | <Lachy> | although, Henri seemed to get a really good response from Aaron (I believe) that showed significant improvement |
| 06:50 | <Lachy> | this one ttp://www.w3.org/mid/4680E4F5.6080903⊙mn and http://www.w3.org/mid/4680FE26.9090802⊙mn |
| 06:51 | <Hixie> | yeah |
| 06:51 | <Hixie> | let's hope people go more in that direction |
| 06:53 | <Hixie> | what year was cellpadding="" invented? |
| 06:54 | <Hixie> | i have about 250,000 documents labelled 1990, and about 250,000 documents labelled 1990 that have an element with a cellpadding="" attribute |
| 06:55 | Hixie | dismisses the 1990 data |
| 06:58 | <Hixie> | the sheer number of different doctypes is insane |
| 06:59 | <Lachy> | Hixie: HTML tables were invented around 1995-96 and published in http://www.ietf.org/rfc/rfc1942.txt |
| 06:59 | <Lachy> | that includes cellpadding |
| 07:00 | <Hixie> | yeah so basically anything before 1995 is statistically insignificant |
| 07:00 | <Hixie> | pity |
| 07:00 | <Hixie> | not surprising though |
| 07:42 | <Hixie> | wow |
| 07:42 | <Hixie> | limited quirks was in the 0%-2% range until 2004, then it jumped to 11%, 13%, 20% in 2006 |
| 07:47 | <Lachy> | what do you mean by limited quirks? |
| 08:05 | <Hixie> | Lachy: "almost standards" |
| 08:05 | <Lachy> | ok |
| 08:06 | <Lachy> | I wonder if that's because tools like Dreamweaver started producing reasonable code with transitional DOCTYPEs around that time |
| 08:08 | <Lachy> | actually, dreamweaver was doing that in 2002 when they released Dreamweaver MX |
| 08:16 | <Hixie> | it started around 1999, with xhtml |
| 08:53 | <Hixie> | year over year, the most popular class names are very variable |
| 08:55 | <Hixie> | oh nm |
| 09:04 | <Hixie> | hmm, <font> is dropping in usage |
| 09:04 | <Hixie> | that's encouraging |
| 09:04 | <Hixie> | <table> isn't |
| 09:05 | <Hixie> | my 2000 data is borked |
| 09:05 | <Hixie> | probably skewed by one site or something |
| 13:22 | Philip` | kind of dislikes it when the spec has exactly the same paragraph repeated in two different places, since his spec<->testcase annotation script uses regexps within paragraphs to identify the right sentence for each test and gets confused by duplicates :-( |
| 13:35 | <zcorpan> | http://diveintomark.org/archives/2007/06/30/irony style="" (quote from a t-shirt with red text) |
| 13:43 | <mpt> | webben_, in my experience, Mozilla bug commentators saying "X is available in an extension, therefore it shouldn't be in the base software". is a common mistake caused by underestimating the difficulty of finding+installing extensions. It's not particularly striking. |
| 13:46 | <mpt> | (Personally I think longdesc= is a waste of time for a browser to support, but not because there are extensions that support it.) |
| 13:48 | <Lachy> | the fact that ATs already support it is the strongest reason to include it, but in practice, I think it has failed |
| 13:51 | <Dashiva> | I thought longdesc seemed like a good case for a microformat |
| 13:52 | <Lachy> | Dashiva: do you mean <a rel=longdesc>? |
| 13:53 | <Dashiva> | No, more that the use case seems so limited, it would make more sense to let the group actually using it decide how they want it, and keep it out of the main spec |
| 13:53 | <Dashiva> | This is orthogonal to fallback/alt content for images, though |
| 13:57 | <Lachy> | oh well, the legal stick of accessibility has been waived again :-/ |
| 13:57 | <Lachy> | why is it that when accessibility advocates can't come up with a rational argument, they always fall back to the legal stick? |
| 14:00 | <Dashiva> | Well, maybe they realize there are no carrots available? |
| 14:17 | <mpt> | If the Web had smell-o-vision, would accessibility advocates fight for longdescs of odors on behalf of those with no sense of smell? |
| 14:19 | <Lachy> | A perfume site that made use of smell-o-vision would probably provide a description of the smell anyway for all users, so they can know what it's like before sampling. |
| 14:20 | <Dashiva> | But that's just like now |
| 14:20 | <Dashiva> | Everyone is disabled with respect to smelling on the web |
| 14:20 | <Dashiva> | Why don't we have accessible smells? |
| 14:25 | <Philip`> | Are there any free non-patent-encumbered media formats for odours? |
| 14:27 | <Philip`> | It would be a pain if you had to use multiple encoders (encodours?) since the common browsers all support different formats :-( |
| 14:36 | <Lachy> | it looks like Sony already has a patent on one form of the technology http://theredactor.blogspot.com/2005/04/birth-of-smellovision.html |
| 14:37 | <Lachy> | http://blog.teledyn.com/node/2286 |
| 14:38 | <Lachy> | of course, it's another case of the US patent office granting another invalid patent for a non-existent invention |
| 14:41 | <zcorpan> | you can patent things that haven't been invented? |
| 14:41 | <Lachy> | apparently Sony can |
| 14:41 | <zcorpan> | why am i not surprised |
| 14:42 | <Lachy> | I really do not get the whole <picture> idea. It provides nothing more than <object> does, and has absolutely no browser support |
| 14:43 | <zcorpan> | i think it's the case of http://ln.hixie.ch/?start=1108984991&count=1 |
| 14:44 | <Philip`> | <object> having browser support is a problem if that support contradicts the requirements for supporting images, and can't be changed (due to compatibility concerns), whereas a new element like <picture> doesn't have that problem |
| 14:45 | <Philip`> | (since <picture> could work eventually, while <object> could never work) |
| 14:45 | <Philip`> | (I don't know if the existing <object> support does have that problem in practice, though) |
| 14:46 | <Lachy> | <object> can work, and does work in some UAs already. However, I would like to know how well object fallback works with screen readers when used to embed images |
| 14:46 | <Lachy> | webben_: yt? |
| 15:16 | <webben_> | Lachy: yt? |
| 15:16 | <Lachy> | yo, is it you who knows all about screen readers and stuff? |
| 15:16 | <webben_> | I certainly wouldn't say all. :) |
| 15:17 | <Lachy> | well, more than I do :-) |
| 15:17 | <webben_> | Lachy: do you have a testcase? |
| 15:17 | <Lachy> | I can make one |
| 15:18 | <webben_> | Lachy: I suspect it's basically dependent on how browsers handle object fallback. |
| 15:19 | <webben_> | but I can test with VO and Window-Eyes easily enough |
| 15:19 | webben_ | unfortunately doesn't have a copy of JAWS |
| 15:19 | <Lachy> | make a page with this and let me know if they read it <object data="http://ln.hixie.ch/resources/images/astrophy/original"; type="image/png">Fallback Content</object> |
| 15:22 | <Lachy> | and can you also see how well they do with the flash and fallback content on this page http://www.3m.com/intl/au/office/TakeCommand/ |
| 15:28 | <webben_> | Lachy: well, the object alternative in http://www.benjaminhawkeslewis.com/www/test-cases/object-fallback.html doesn't work with latest WebKit + VO |
| 15:33 | <Lachy> | you could try <noembed> as well |
| 15:33 | <webben_> | Lachy: but the problem is things are embedded |
| 15:34 | <Lachy> | yeah, true |
| 15:34 | <webben_> | that's the difference between fallback and alternative/descriptive content |
| 15:34 | <webben_> | I tried embed alt ... that didn't work in VO either |
| 15:34 | <webben_> | it read the picture as "Frame 1" |
| 15:35 | <Lachy> | webkit probably doesn't support <embed alt>. Opera's build in screen reader might, or maybe Opera with Windows Eyes (if possible) |
| 15:36 | <webben_> | Lachy: Opera doesn't work with screen readers ATM. Supposedly that will be fixed in a forthcoming release. |
| 15:36 | <webben_> | I think the 3m site is way too complex to function as a testcase |
| 15:38 | <webben_> | yeah webkit doesn't support embed alt ... feeble. |
| 15:38 | <Lachy> | I just wanted to know if that flash was accessible |
| 15:38 | <webben_> | might try and submit a patch for that |
| 15:38 | <webben_> | Lachy: oh right |
| 15:39 | <webben_> | if we want an example of supposedly accessible flash, there's jkrowling's site |
| 15:39 | <Lachy> | I built that 3m site, but I didn't make the flash |
| 15:39 | <webben_> | ah |
| 15:43 | <Lachy> | Opera Voice will read the object fallback |
| 15:44 | <Lachy> | but it won't read the embed alt |
| 15:46 | <webben_> | which is silly, as flash accessibility won't work from Opera |
| 15:47 | <webben_> | hmm actually maybe it's not that silly |
| 15:47 | <webben_> | if you're embedding video you may still be able to watch an auto-playing video |
| 15:48 | <Lachy> | Opera wouldn't read the fallback for the Flash, only the image |
| 15:55 | <Lachy> | why did JK Rowling make the accessibility enabled version completely separate? The english version looks no different from the accessible english version |
| 15:56 | <webben_> | Lachy: I don't know. There are a lot of things I don't understand about how that site was put together. |
| 15:56 | <webben_> | src="http://ln.hixie.ch/resources/images/astrophy/original"; is too confusing for IE |
| 15:56 | <webben_> | as is data="http://ln.hixie.ch/resources/images/astrophy/original"; |
| 15:57 | <webben_> | it throws up a warning that it wants to download Microsoft HTML Viewer (WTF?) |
| 15:57 | <webben_> | presumably it thinks it's an HTML include |
| 15:59 | <Philip`> | Try <object src="http://ln.hixie.ch/resources/images/astrophy/original"; type="image/png">Fallback Content</object> in the DOM viewer in IE - it's great fun, with the kind of 'fun' that only IE can manage |
| 16:00 | <webben_> | funie |
| 16:01 | <Philip`> | (It gives me <OBJECT type=text/html data=data:application/x-oleobject;base64,IGkzJfkDzxGP0ACqAGhvEzwhRE9DVFlQRSBIVE1MIFBVQkxJQyAiLS8vVzNDLy9EVEQgSFRNTCA0LjAgVHJhbnNpdGlvbmFsLy9FTiI+DQo8SFRNTD48SEVBRD4NCjxNRVRBIGh0dHAtZXF1aXY9Q29udGVudC1UeXBlIGNvbnRlbnQ9InRleHQvaHRtbDsgY2hhcnNldD13aW5kb3dzLTEyNTIiPjwvSEVBRD4NCjxCT0RZPg0KPFA+Jm5ic3A7PC9QPjwvQk9EWT48L0hUTUw+DQo= src="http://ln.hixie.ch/resources/images/astrophy/original">; in IE6, and something a bit short |
| 16:01 | <webben_> | hmm I just see #text with IE7 |
| 16:02 | <Philip`> | ...a bit shorter in IE7) |
| 16:02 | <webben_> | I think we need an image with a file extension to test this in IE |
| 16:03 | <webben_> | (IE seems highly reliant on file extensions) |
| 16:04 | <Philip`> | IE only does that x-oleobject for me when type="..." is a recognised MIME type (e.g. text/html, image/png, application/xml, not application/xhtml+xml, etc) |
| 16:04 | <zcorpan> | here's such an image: http://simon.html5.org/valid-html5.png |
| 16:04 | <webben_> | thanks |
| 16:05 | <webben_> | now it recognizes the embed as an image |
| 16:05 | <webben_> | still trips up on object though |
| 16:06 | <Philip`> | http://hsivonen.iki.fi/validator/html5/?doc=http%3A%2F%2Fsimon.html5.org%2Fvalid-html5.png - that filename is wrong - it's not valid HTML5 |
| 16:07 | <zcorpan> | -_- |
| 16:09 | <zcorpan> | funny how innerHTML in ie remembers whether an attribute had an = or not |
| 16:10 | <zcorpan> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20html%3E%0D%0A%3Cp%20foo%3E |
| 16:10 | <zcorpan> | vs http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20html%3E%0D%0A%3Cp%20foo%3D%3E |
| 16:11 | <Philip`> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20html%3E%0D%0A%3Cp%20class%3E%3Cp%20class%3D%3E |
| 16:13 | <zcorpan> | interesting |
| 16:13 | <Philip`> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20html%3E%0D%0A%3Cp%20class%3E%3Cp%20class%3D%3E%0D%0A%3Cp%20style%3E%3Cp%20style%3D%3E%0D%0A%3Cp%20id%3E%3Cp%20id%3D%3E%0D%0A%3Cp%20foo%3E%3Cp%20foo%3D%3E%0D%0A |
| 16:13 | <Philip`> | so that's at least four different behaviours |
| 16:14 | <zcorpan> | try disabled |
| 16:14 | <Philip`> | Oh, but I only get those four behaviours in IE6, not IE7... |
| 16:14 | <Philip`> | Oh, yes I do |
| 16:14 | Philip` | was stupid and looking at the wrong part of the page |
| 16:15 | <Philip`> | Okay, five different behaviours |
| 16:17 | <Philip`> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20html%3E%0D%0A%3Cp%20class%3E%3Cp%20class%3D%3E%3Cp%20class%3D%3D%3E%0D%0A%3Cp%20style%3E%3Cp%20style%3D%3E%3Cp%20style%3D%3D%3E%0D%0A%3Cp%20id%3E%3Cp%20id%3D%3E%3Cp%20id%3D%3D%3E%0D%0A%3Cp%20foo%3E%3Cp%20foo%3D%3E%3Cp%20foo%3D%3D%3E%0D%0A%3Cp%20disabled%3E%3Cp%20disabled%3D%3E%3Cp%20disabled%3D%3D%3E%0D%0A |
| 16:18 | <zcorpan> | <p => :) |
| 16:24 | <zcorpan> | <p=> |
| 16:26 | <zcorpan> | wow, that last one is really weird |
| 16:27 | <Philip`> | Isn't that just the same as <foo>? |
| 16:27 | <zcorpan> | oh wait, i was testing in opera |
| 16:28 | <zcorpan> | <p=> is really weird in opera |
| 16:28 | <zcorpan> | yes, in ie it's the same as <foo> |
| 16:29 | <Philip`> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20HTML%3E%0D%0A%3Cp%3E%3Cp%3D%3Ex%3C/p%3Ex%0D%0A |
| 16:29 | <Philip`> | Opera's innerHTML says "<!DOCTYPE HTML><html><BODY><P><p>x<px</html>" |
| 16:30 | <Philip`> | Oh, but that's exactly the same as for parsing <p><x>x</p>x |
| 16:33 | <Philip`> | http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C%21DOCTYPE%20HTML%3E%0D%0A%3Cp%3E%3C%21%3Ex%3C/p%3Ex%0D%0A innerHTML is peculiar in Opera too |
| 16:34 | <zcorpan> | i guess the weird part is that the = is not in the dom |
| 16:36 | <Philip`> | Ah, okay |
| 16:37 | <Philip`> | Looks like all the attributes on <p= p=p> disappear too |
| 16:37 | <Philip`> | but not on <p== p=p> |
| 16:38 | <zcorpan> | not all attributes |
| 16:39 | <Philip`> | Oh, only the first one |
| 16:39 | <zcorpan> | the first is equivalent to <p="p=p"> |
| 16:39 | <zcorpan> | seems like the tokenizer thinks it's an attribute |
| 16:40 | <zcorpan> | </p=> closes it |
| 16:45 | <Philip`> | It's kind of worrying that I can't find one single canvas feature whose tests all pass in all of Firefox, Opera and Safari |
| 16:47 | <Dashiva> | getContext? |
| 16:48 | <Philip`> | "There is only one CanvasRenderingContext2D object per canvas, so calling the getContext() method with the 2d argument a second time must return the same object." - but getContext('2d') === getContext('2d') fails in Opera |
| 16:49 | <Philip`> | (at least 9.2 - I'm hoping some issues will be fixed a bit in 9.5...) |
| 16:49 | <Philip`> | (and if they aren't, I probably ought to start submitting bug reports) |
| 16:52 | <annevk> | You probably should |
| 16:52 | <annevk> | Although we do have fixed lots of <canvas> bugs (per my measure of "lots of") |
| 16:55 | <Philip`> | Are all those fixes going to be in the public 9.5 builds (rather than 10.0 or whatever)? |
| 16:59 | <annevk> | yeah |
| 16:59 | <annevk> | some might be in the Wii version already |
| 16:59 | <annevk> | or in Opera Mini 4 (although I'm not 100% sure how that will function with <canvas>) |
| 17:02 | annevk | wonders how much we should care about RFC3986 |
| 17:02 | <annevk> | URL5 |
| 17:04 | <annevk> | Lachy, debugging is an important use case for making style= conforming |
| 17:04 | <Philip`> | Ooh, neat, the Opera Mini 4 simulator works |
| 17:04 | <Lachy> | annevk: why? |
| 17:04 | <Philip`> | Too bad it returns BGRA instead of RGBA in getPixelData, so all the automated test results are broken... |
| 17:05 | <annevk> | Lachy, for instance your debugging tool might display an error if you use incorrect markup |
| 17:05 | <annevk> | Philip`, whoa, that sounds like a major bug |
| 17:05 | <Lachy> | maybe |
| 17:05 | <Philip`> | (Actually, I don't know if it's getPixelData or getPixel) |
| 17:06 | <annevk> | Philip`, hmm, we did add some support for getPixelData and putImageData |
| 17:08 | Philip` | tries testing it properly |
| 17:11 | <Philip`> | Oops, getPixelData isn't very good to test, since it's not spelt that way... |
| 17:12 | <annevk> | heh |
| 17:13 | <annevk> | funny that I just copied your sentence for getPixelData but got the other method name right... |
| 17:13 | <Philip`> | Hmm, getPixel looks nicely broken too |
| 17:18 | <Philip`> | http://canvex.lazyilluminati.com/misc/pixel.html |
| 17:19 | <Philip`> | Hmm, now Opera Mini gives me java.lang.IllegalStateException |
| 17:19 | <annevk> | they are in different order here |
| 17:21 | <Philip`> | http://canvex.lazyilluminati.com/misc/operamini.png |
| 17:21 | <annevk> | yeah, get the same here |
| 17:21 | <annevk> | what's the correct order anyway? |
| 17:22 | <annevk> | nm |
| 17:22 | <Philip`> | It should say 12, 34, 56, 127 (or 128) in both cases |
| 17:22 | <Philip`> | (Well, the getPixel case should say 12, 34, 56, 0.5) |
| 17:22 | <Philip`> | (and it shouldn't have a load of random junk at the end, I'm guessing) |
| 17:23 | <Philip`> | Firefox gets it wrong and says 6, 17, 28, 127 because it does premultiplied alpha |
| 17:23 | <annevk> | ouch |
| 17:24 | <Philip`> | Also I don't know whether getImageData(...).data is meant to be an actual JS array, or if it's allowed to be another object (like that CanvasPixelArray in Opera) |
| 17:31 | <Philip`> | It's intriguing how Opera Mini 4 fails on http://canvex.lazyilluminati.com/tests/tests/toDataURL.complexcolours.html too - all the transparent colours are correct, but the solid one is missing... |
| 17:32 | Philip` | goes to fix up his test system to cope with that getImageData bug |
| 17:34 | <hsivonen> | Hixie: does it work for you if I post my comments on the parsing algorithm piecemeal to public-html and say that they are my "detailed review"? |
| 17:44 | <Philip`> | I also like how Opera Mini shows JPEG artifacts on the <canvas> output :-) |
| 17:49 | <annevk> | Philip`, hmm, the solid color missing could indicate that it expects a float for the alpha value |
| 17:49 | annevk | ponders |
| 17:53 | <Philip`> | Ah, it looks like that is the case |
| 17:53 | <Philip`> | (which is odd since it works fine in Opera 9.2) |
| 17:54 | <annevk> | maybe it was a fix in the CSS parser |
| 17:54 | annevk | ponders |
| 17:55 | <Philip`> | CSS3 Color says the alpha value is a <number>, and 1 is a <number>, as far as I can see |
| 17:55 | <annevk> | k |
| 17:57 | Philip` | wishes the Opera Mini simulator had a bigger screen :-p |
| 18:00 | <annevk> | I filed bugs for the getImageData and rgba(... 1) |
| 18:01 | Philip` | tries to find other regressions |
| 18:01 | <annevk> | much appreciated :) |
| 18:01 | annevk | notes that getImageData is not a regression but a pretty serious feature bug... |
| 18:02 | <hsivonen> | what's the purpose of the document.write/noscript arrangement at http://demo.opera-mini.net/fourzerobeta/ ? |
| 18:03 | hsivonen | notes that the foremost Java applet use case (demoing Opera Mini) uses <applet> |
| 18:03 | <annevk> | I've not much against <applet> |
| 18:05 | <hsivonen> | Hixie: one thing that might make sense for research is counting <applet> vs. <object> with Java traits |
| 18:06 | <hsivonen> | my non-researched hunch is that if one embeds an applet, <applet> makes more sense than fighting the <object> windmills |
| 18:07 | <annevk> | keeping it in makes some sense |
| 18:07 | <annevk> | but I'm not really that interested in writing Java applets so I haven't done much about it |
| 18:07 | <hsivonen> | It is totally unclear to me what problem *not keeping it* solves |
| 18:08 | <annevk> | language simplicity is one argument |
| 18:08 | <hsivonen> | it is a rather weak argument if it needs to be specced anyway and <object> is hairier than <applet> |
| 18:09 | <annevk> | well, in theory it's <object data=applet.java> |
| 18:09 | <annevk> | and that actually works in most browsers just not in IE |
| 18:11 | <Philip`> | Hmph, now I just get java.io.IOException when trying to run all the tests... |
| 18:12 | <annevk> | https://bugs.opera.com/ |
| 18:12 | annevk | has to go |
| 18:12 | <Philip`> | Ooh, it worked that time |
| 18:14 | <Philip`> | though it doesn't like me pressing the buttons to record manual results :-( |
| 18:24 | <Hixie> | hsivonen: sure |
| 18:31 | <hsivonen> | Hixie: ok |
| 19:00 | <Philip`> | Hmm, I can get Opera Mini to either run the test cases, or to submit a form, but can't get it to ever do both at the same time, and there's no way I'm going to manually copy out the test results... |
| 19:08 | <Philip`> | Aha, looks like Opera Mini doesn't like setInterval at all |
| 19:15 | <webben_> | Lachy: wrt to 3M, AFAICT WIndow-Eyes doesn't even see the Flash object because of the SWFObject insertion; investigation of the same object shows that it isn't exposing any information to MSAA; but captions for video may not use a MSAA interface |
| 19:16 | <webben_> | can't seem to get Window-Eyes to notice the fallback content for object either |
| 19:16 | <webben_> | (though maybe it would if i disabled flash and js etc.) |
| 19:17 | <webben_> | this may of course all be my Window-Eyes incompetence though :) |
| 19:25 | <Philip`> | Opera Mini has a rather odd image that looks kind of like splattered blood on its "Internal server error: Failed to transcode URL" page |
| 19:28 | hsivonen | tries to shift the kind of signal public-html gets to the whatwg-style direction :-) |
| 19:35 | <webben_> | Lachy: "Providing alternative content in the same page or linking to it." ... That's not as a good a solution as progressive enhancement. |
| 19:35 | <webben_> | traditionally alternative sites have a) lagged behind the main page in terms of content and b) not been as "accessible" as they think they are |
| 19:35 | <webben_> | (wrt http://lists.w3.org/Archives/Public/public-html/2007Jul/0027.html) |
| 19:46 | Philip` | finally gets Opera Mini to sort of almost cooperate |
| 19:47 | <Philip`> | http://canvex.lazyilluminati.com/tests/20070701/tests/results.html - Desktop 9.2 vs Mini 9.5 (with a few missing results because Mini doesn't cooperate enough) |
| 19:49 | <Philip`> | Regressions that I can see: The CSS-colour-outputting code (getPixel, reading fillStyle, etc) writes a load of garbage characters (yay, buffer overflow) instead of the alpha value |
| 19:49 | <Philip`> | Interpolation of gradients with alpha doesn't work - it doesn't draw the gradient at all |
| 19:50 | <Philip`> | ...and I can't find anything else (except the rgba(..., 1) thing, and one case with radial gradients but radial gradients are horribly broken anyway so that doesn't matter) |
| 19:52 | Philip` | thinks he needs to split up his tests so there's only ~100 on a page at once, because otherwise Opera Mini and WebKit (and Opera to a limited extent) get pretty unhappy |
| 19:55 | <Philip`> | Oh, sorry, looks like that gradient-alpha interpolation thing is just the broken CSS parsing again |
| 20:07 | <hsivonen> | hmm. <pre>\n vs. <pre><!-- -->\n |
| 20:09 | <hsivonen> | grr. they are different |
