| 00:00 | <TabAtkins> | jamesr_: Hmm, I didn't know about -moz-image-rendering. What does it do, control the interpolation when you scale the iamge, or something? |
| 00:00 | <Hixie> | bga_: so just generate the pages ahead of time, one for each language |
| 00:00 | <jamesr_> | TabAtkins: that's what it does |
| 00:00 | <jamesr_> | mozilla didn't actually prefix image-rendering, they have a value that they prefix |
| 00:01 | <jamesr_> | i think it came un-prefixed from IE |
| 00:01 | <TabAtkins> | Ah, kk. Well, I should be defining it in Image Values, then. |
| 00:01 | <bga_> | i understand |
| 00:01 | <bga_> | ok |
| 00:02 | <TabAtkins> | jamesr_: bbiab |
| 00:02 | <bga_> | thanks Hixie |
| 00:02 | <Hixie> | np |
| 00:02 | <Hixie> | sorry i can't give you a better answer :-) |
| 00:56 | <TabAtkins> | jamesr_: Ah, I see. image-rendering comes from SVG, that's why it's non-prefixed. The only innovation is applying it to the HTML <img>. |
| 00:57 | <jamesr_> | TabAtkins: and <canvas> and <video> |
| 00:57 | <TabAtkins> | Yeah. |
| 00:58 | <jamesr_> | or background-image on any element |
| 00:58 | <TabAtkins> | So you just wanted a less shittier name than crisp-edges? |
| 00:58 | <jamesr_> | yeah |
| 00:58 | <jamesr_> | and to see if anyone else is interested in it |
| 00:58 | <jamesr_> | i'd propose nearestNeighbor |
| 01:00 | <TabAtkins> | That goes against the current trend in the names, which is to express *intent* and letting that map to some particular algorithm that respects the intent. |
| 01:02 | <heycam> | "optimizeSpeed - Indicates that the user agent shall emphasize rendering speed over quality. The user agent should use a resampling algorithm which achieves the goal of fast rendering, with the requirement that the resampling algorithm shall be at least as good as nearest neighbor resampling." |
| 01:02 | <heycam> | although i wouldn't be surprised if nearest neighbour might be slower than resampling on gpus or something |
| 01:06 | <Hixie> | ok the magic disappearing scheme in the chrome url bar is starting to piss me off |
| 01:06 | <Hixie> | every time i try to copy/paste a domain name i get a scheme along with it |
| 01:07 | <TabAtkins> | And? |
| 01:07 | <Hixie> | i don't want a scheme, i want the domain name! |
| 01:07 | <TabAtkins> | Where does the distinction matter? |
| 01:07 | <Hixie> | when i try to paste the domain name somewhere that wants the domain name, i have to go back and edit the scheme out |
| 01:07 | <Hixie> | every time |
| 01:07 | <TabAtkins> | Where are you pasting things that take a domain name but don't understand schemes? |
| 01:07 | <Hixie> | e.g. if i want to edit ~/damowmow.com/playground/htmlwg/chart.html |
| 01:08 | <Hixie> | i end up getting ~/http://damowmow.com/playground/htmlwg/chart.html |
| 01:08 | <Hixie> | and have to go and edit out the http:// bit |
| 01:08 | <TabAtkins> | Oh. That seems like a weird edge case. |
| 01:08 | <Hixie> | happens to me at least a dozen times a day |
| 01:08 | <Hixie> | but i agree i'm an edge case |
| 01:09 | <Hixie> | not saying it should change |
| 01:09 | <Hixie> | just that it's frustrating for me :-) |
| 01:09 | <TabAtkins> | Solution: copy all but the first letter of the domain name, then type the first letter before pasting. |
| 01:09 | <TabAtkins> | Should be roughly equivalent in difficulty to just copying the domain without the scheme in a non-Chrome browser. |
| 01:10 | <Hixie> | i generally copy it by double clicking the first word and dragging to the end |
| 01:10 | <Hixie> | doing it more precisely takes more effort |
| 01:10 | <TabAtkins> | Dude, you click? |
| 01:10 | <TabAtkins> | ctr+l, home, right, shift+end, ctrl+c |
| 01:11 | <Hixie> | no home key on this keyboard |
| 01:11 | <TabAtkins> | Your keyboard's broken. |
| 01:11 | <Hixie> | no argument from me there |
| 01:11 | TabAtkins | can't do any serious text editting without home/end. |
| 01:11 | <Hixie> | i use ctrl+a ctrl+e on qwerty keyboads |
| 01:12 | <Hixie> | i do use home/end on my dvorak keyboard though |
| 01:12 | <Hixie> | don't have that here though |
| 01:14 | <jamesr_> | heycam: that's kind of my problem with optimizeSpeed/optimizeQuality |
| 01:14 | <jamesr_> | heycam: the speed/quality tradeoff is pretty different on the GPU |
| 01:14 | <TabAtkins> | Which is why the intent-based keywords are better. ^_^ |
| 01:14 | <jamesr_> | but nearestNeighbor is asking for a certain visual appearance, not some perf hint |
| 01:14 | <heycam> | right |
| 01:14 | <heycam> | i'd be happy with nearestNeighbor for exactly that reason |
| 01:14 | <heycam> | that's the specific behaviour you want |
| 01:15 | <heycam> | nearestNeighbor/nearestNeighbour :) |
| 01:15 | <TabAtkins> | I think not. ^_^ |
| 01:15 | <jamesr_> | the american spelling is less bytes on the wire :) |
| 01:15 | <TabAtkins> | The web is written in American. |
| 01:15 | <heycam> | yeah, yeah |
| 01:15 | <TabAtkins> | 'Murrican, I mean. |
| 01:15 | <heycam> | but look at the X11 colo[u]r name synonyms! |
| 01:16 | <TabAtkins> | Don't get me started on X11 colors. |
| 01:19 | <heycam> | relax, peachpuff :D |
| 01:20 | <Hixie> | in the last 24 hours, the htmlwg just reached it's highest number of open issues since we started using an issue tracker |
| 01:21 | <Hixie> | http://damowmow.com/playground/htmlwg/chart.html |
| 01:22 | <jamesr_> | what are they doin' over there? |
| 01:28 | <TabAtkins> | Hrm. The "default" size of replaced elements is, what, 300px by 200px? |
| 01:28 | <jamesr_> | 300x150 iirc |
| 01:28 | <TabAtkins> | kk. |
| 01:38 | <TabAtkins> | Hmm. I'm about to define the element-ref() function in CSS, which allows the functionality Moz currently offers with -moz-element() and mozSetImageElement(), where you can use elements that aren't even in the document as image sources. |
| 01:39 | <TabAtkins> | I'm wondering which part makes sense to specify in CSS, and which part makes sense to specify in HTML. |
| 01:41 | <TabAtkins> | CSS always only applies to a document, so I don't know how sensical it is to talk about the ability to create a canvas in js and use it as an image without adding it to a document. |
| 01:48 | <TabAtkins> | Hixie: I think I'll just write it as "the host language may define a way to associate elements with idents for the purpose of this function", and then give you some text for HTML5. Sound all right? |
| 02:11 | <heycam> | why is the default replaced element size not the golden ratio, that is the question |
| 02:11 | <jcranmer> | because that's not sufficiently arbitrary |
| 07:24 | <cardona507> | isn't there a version of the spec that is a little easier on the browser? |
| 07:25 | <cardona507> | i'm forgetting the url |
| 07:25 | <Rik`> | cardona507: http://www.whatwg.org/html5 |
| 07:25 | <cardona507> | gracias |
| 08:27 | <MikeSmith> | I wonder if anybody here has ever set up a mercurial mirror of a subversion repo and might have advice on how best to go about it |
| 08:28 | <MikeSmith> | I found it pretty easy to set up git mirrors for subversion repos, but am finding it not quite as easy to figure out if/how to do it for mercurial |
| 08:33 | <hsivonen> | MikeSmith: there's a one-time import tool at least |
| 08:34 | <hsivonen> | MikeSmith: is this for validator stuff or something else? |
| 08:34 | <MikeSmith> | for validator stuff |
| 08:34 | <MikeSmith> | was going to try to see if I could set up some mirrors at bitbucket |
| 08:37 | <MikeSmith> | it looks like the Atlassian folks have added a lot of features lately and it has some better feature parity with github now |
| 08:37 | <jgraham> | MikeSmith: hgsvn ? |
| 08:37 | <MikeSmith> | jgraham: seems to be deprecated |
| 08:38 | <jgraham> | Oh, well I guess you follow the pointer to hgsubversion then |
| 08:38 | <MikeSmith> | yeah |
| 08:39 | <jgraham> | Looks quite like git-svn really |
| 08:50 | <hsivonen> | MikeSmith: I'd be ok with moving repos other than the syntax repo to bitbucket without mirroring |
| 08:50 | <hsivonen> | MikeSmith: i.e. making bitbucket become the new repo of record |
| 08:51 | <hsivonen> | MikeSmith: for the syntax repo, it's necessary to discuss with fantasai |
| 08:51 | <MikeSmith> | ok |
| 08:51 | <hsivonen> | MikeSmith: assuming the a migration tool that attempts to preserve history is used |
| 08:51 | <MikeSmith> | yep |
| 08:53 | <MikeSmith> | so I will try to get the others set up at bitbucket today |
| 08:53 | <MikeSmith> | OK if I e-mail fantasai to ask about the syntax repo? |
| 08:53 | <MikeSmith> | (and CC you of course) |
| 08:54 | <hsivonen> | MikeSmith: OK |
| 08:54 | <MikeSmith> | OK, thanks |
| 08:55 | <hsivonen> | I'm http://bitbucket.org/hsivonen |
| 08:55 | <hsivonen> | has bitbucket been always owned by atlassian? |
| 08:56 | <hsivonen> | looks like it's a recent purchase |
| 09:10 | <MikeSmith> | hsivonen: yeah, I think Atlassian acquired it only a couple months ago |
| 09:30 | <MikeSmith> | hsivonen: got https://bitbucket.org/validator/validator set up |
| 09:30 | <MikeSmith> | and you should have full admin access to it |
| 09:31 | <MikeSmith> | working on the util module next |
| 09:32 | <MikeSmith> | jgraham: hgsubversion appears to be the way to go |
| 09:33 | <MikeSmith> | dead simple to get things copied over with it |
| 09:33 | <MikeSmith> | though it seems a lot slower than doing similar with git-svn |
| 09:33 | <hsivonen> | MikeSmith: it looks like changeset authorship didn't get migrated |
| 09:33 | <MikeSmith> | no? |
| 09:34 | <hsivonen> | MikeSmith: every changeset is authored by sideshowbarker |
| 09:34 | <hsivonen> | MikeSmith: including ones I made |
| 09:34 | <MikeSmith> | no, that's just the most recent one |
| 09:34 | <MikeSmith> | oh wait |
| 09:34 | <MikeSmith> | that's probably because I muffed it up |
| 09:34 | <hsivonen> | MikeSmith: e.g. https://bitbucket.org/validator/validator/changeset/71849bf8f10b |
| 09:34 | <MikeSmith> | not the importt |
| 09:35 | <MikeSmith> | but there is a way that you can set aliases within bitbucket |
| 09:35 | <MikeSmith> | I think I just set the alias wrong |
| 09:35 | <MikeSmith> | yeah |
| 09:35 | <hsivonen> | it should be possible to retain the changeset authors with the migration tools, since the migration from svn to hg.mozilla.org preserved auhors |
| 09:35 | <MikeSmith> | lemme see if I can fix it right now |
| 09:35 | <MikeSmith> | yeah, it did preserve it for sure |
| 09:36 | <MikeSmith> | just the UI is showing what it shows because I muffed up the bitbucket UI setting for it somehow |
| 09:37 | <MikeSmith> | there we go |
| 09:37 | <MikeSmith> | should work now |
| 09:37 | <hsivonen> | MikeSmith: cool. thanks. |
| 09:37 | <MikeSmith> | there's a settings thing to change the mappings here: |
| 09:37 | <MikeSmith> | https://bitbucket.org/validator/validator/admin/committers |
| 09:38 | <hsivonen> | MikeSmith: btw, after the DoS against bugzilla.validator.nu taking down the script and style sheet that html5.validator.nu uses, I started (again) contemplating serving style.js and script.js from the servlet itself |
| 09:39 | <hsivonen> | MikeSmith: would that work for your deployment? |
| 09:39 | <MikeSmith> | yeah, sure |
| 09:39 | <hsivonen> | ok |
| 09:39 | <MikeSmith> | I can't see any reason why it would not, anyway |
| 09:39 | <MikeSmith> | and if I run into problems, I'll just figure out how to fix them |
| 09:40 | <hsivonen> | my plan is |
| 09:40 | <hsivonen> | 1) Upon startup, load the files into RAM from the jar that has the schemas |
| 09:40 | <hsivonen> | 2) compute crytographic hashes of the file contents |
| 09:40 | <hsivonen> | 3) gzip the files |
| 09:41 | <hsivonen> | 4) Keep around byte arrays for the gzipped contents in RAM |
| 09:41 | <hsivonen> | serve the contents from HASH.js and HASH.css |
| 09:41 | <hsivonen> | 6) Make the HTML point to HASH.js and HASH.css |
| 09:42 | <hsivonen> | 7) set the expiry of HASH.js and HASH.css to a year |
| 09:43 | <wirepair> | DoS? care to explain? |
| 09:44 | <hsivonen> | wirepair: someone kept GETting show_bug.cgi with various bug numbers from bugzilla.validator.nu |
| 09:44 | <wirepair> | ah |
| 09:45 | <hsivonen> | wirepair: this exhausted the resources on the server that was also hosting about.validator.nu, which servers the UI script and style for html5.validator.nu |
| 09:45 | <wirepair> | gotcha |
| 09:45 | <hsivonen> | although html5.validator.nu is on a different server |
| 09:45 | <wirepair> | right |
| 09:45 | <MikeSmith> | hsivonen: sounds workable to me |
| 09:45 | <hsivonen> | MikeSmith: ok |
| 09:46 | <MikeSmith> | I think the only change it would mean for my deployment is that I'd need to point to the HASH.js and HASH.css files in my custom page and form templates |
| 09:46 | <MikeSmith> | as far as I can understand ti |
| 09:48 | <hsivonen> | MikeSmith: don't your templates use the current configurability mechanism for the URLs of those files? |
| 09:48 | <hsivonen> | anyway, this shouldn't be a big change for the templates |
| 09:48 | <MikeSmith> | yeah |
| 09:48 | <MikeSmith> | yeah, true (what you said) |
| 09:49 | <MikeSmith> | they are just using the same mechanism |
| 09:49 | <MikeSmith> | so, anyway, it'd be no problem as far as I can see |
| 09:50 | <MikeSmith> | hsivonen: so if I can get the non-syntax repos set up at bitbucket today, you want me to also go ahead and switch the build script to use them instead? |
| 09:50 | <hsivonen> | MikeSmith: I'm ok with you changing the build script, but I can do it too, if you'd like |
| 09:50 | <MikeSmith> | ok |
| 09:50 | <MikeSmith> | I guess we'll see how long it takes me to get stuff set up |
| 09:51 | <MikeSmith> | but I think I can get them set up within the next hour or so |
| 09:51 | <hsivonen> | ok. (I'm stepping away for a bit to get lunch) |
| 09:51 | <MikeSmith> | ok |
| 09:51 | <MikeSmith> | I will be stepping away at some point for beer |
| 09:52 | <MikeSmith> | after which I may or may not be stepping or stumbling back |
| 09:52 | <MikeSmith> | but anyway, there's no rush I guess |
| 09:56 | <thiessenp> | Is it possible to create a canvas DOM fragment/element, set its width/height (el.style.width=200), and then append it to the body? When I try this my style rules keep getting ignored - seriously cramping my style :> |
| 10:10 | <zcorpan> | looking at http://damowmow.com/playground/htmlwg/chart.html it looks like a chair change is overdue |
| 10:12 | <MikeSmith> | that's an interesting way to interpret things |
| 10:12 | <jgraham> | We had that converstaion like 6 months ago |
| 10:13 | <zcorpan> | oh |
| 10:13 | <MikeSmith> | do UAs process scripts as expected in content served as text/html if you do <script type="text/javascript"> <![CDATA[... unescaped script content ...]]></script> ? |
| 10:14 | <MikeSmith> | that is, if you use CDATA inside the script element in text/html |
| 10:14 | <MikeSmith> | a pseduo-CDATA section |
| 10:14 | <zcorpan> | MikeSmith: the "<![CDATA[" is passed to the script engine and results in a compile time syntax error |
| 10:15 | <MikeSmith> | I see |
| 10:15 | <MikeSmith> | will the rest of the script still be processed? |
| 10:15 | <zcorpan> | no |
| 10:16 | <MikeSmith> | hmm |
| 10:16 | <MikeSmith> | OK |
| 10:16 | <MikeSmith> | I guess I should let dude on the www-validator list know that |
| 10:18 | <zcorpan> | MikeSmith: SVG <script> in text/html supports <![CDATA[ though |
| 10:18 | <MikeSmith> | OK |
| 10:19 | <zcorpan> | and even needs <![CDATA[ unless you fancy escaping your < and & as < and & |
| 10:21 | <MikeSmith> | I definitely don't fancy that |
| 10:21 | <MikeSmith> | I don't really fancy needing to use CDATA sections either |
| 10:22 | <MikeSmith> | but I guess I fancy it more than I would escaping JS code |
| 10:22 | <zcorpan> | there was some discussion about aligning the rules between the two <script>s |
| 10:23 | <zcorpan> | in some bug? |
| 10:38 | <MikeSmith> | zcorpan: yeah, I think I remember something about it but can't remember the discussion having led anywhere |
| 10:39 | <MikeSmith> | validator sources except for the syntax repo are now mirrored at bitbucket |
| 10:39 | <MikeSmith> | https://bitbucket.org/validator |
| 10:40 | <MikeSmith> | hopefully fantasai will be OK with having the syntax repo moved there as well |
| 10:47 | <Philip`> | thiessenp: Set canvas.width=200, not canvas.style.width=200 |
| 10:48 | <MikeSmith> | man, the set of content-blocking mechanisms for users in IE seems… baroque |
| 10:48 | <MikeSmith> | http://blogs.msdn.com/b/ie/archive/2010/11/30/selectively-filtering-content-in-web-browsers.aspx |
| 10:48 | <thiessenp> | Philip': oh huh strange ok - thanks! :D |
| 10:48 | <MikeSmith> | I wonder how normal users are supposed to figure out what knob to turn for what particular type of blocking |
| 10:49 | <MikeSmith> | but I guess that's the case for other browsers as well |
| 10:59 | <annevk> | Hixie, http://search.cpan.org/dist/Net-Twitter/lib/Net/Twitter.pod |
| 11:00 | <annevk> | Hixie, make it happen |
| 11:01 | <jgraham> | Someone fixed it, I thought |
| 11:01 | <annevk> | the latest tweet is from September 1 |
| 11:01 | <annevk> | matjas semi-volunteered to fix it |
| 11:02 | <hsivonen> | :-( http://damowmow.com/playground/htmlwg/chart.html times out in Opera Mini. That is, Opera Mini stops long-running scripts and sends the result to the client before the chart has been fully rendered |
| 11:02 | <annevk> | but I think if Hixie just uses that library it should be fine |
| 11:02 | <hsivonen> | also, SVG would be more appropriate than canvas for charts |
| 11:03 | <hsivonen> | I hate it when I forget to hg add test cases to a patch |
| 11:04 | <annevk> | there's worse things in the world |
| 11:04 | <annevk> | like breaking the tree |
| 11:05 | <hsivonen> | In Mozilla's case, failing to hg add test cases breaks the tree |
| 11:05 | <hsivonen> | but fortunately, in my case only a tryserver push got broken |
| 11:06 | <annevk> | http://kimjongillookingatthings.tumblr.com/ |
| 11:17 | <hsivonen> | can anyone subscribe to public-html-testsuite by sending the usual email to public-html-testsuite-request? |
| 11:19 | <hsivonen> | bah. I guess I'll reply to jgraham without being subscribed and let the list owner perform a moderator action |
| 11:21 | <jgraham> | hsivonen: Anyone should be able to subscribe in the normal way, I think |
| 11:21 | <hsivonen> | jgraham: if so, the system is really slow today |
| 11:35 | <MikeSmith> | hsivonen: mailing-list DB says you are already subscribed to public-html-testsuite |
| 12:03 | <thiessenp> | I feel obligated to ask (out of curiosity) is their an archive or repository of lo gs/talks specific to HTML5 inclusion-vs-separation of WAI-ARIA? (I help out with ARIA :) |
| 12:05 | <annevk> | other than searching I wouldn't really know |
| 12:06 | <annevk> | there's a task force mailing list of some kind too |
| 12:07 | <thiessenp> | I'll search around - thanks |
| 12:19 | <MikeSmith> | hsivonen: it occurs to me that we could also set up a mirror of the jing-trang code on bitbucket if we cared to |
| 12:20 | <hsivonen> | MikeSmith: does the importer software really handle export properly, too? |
| 12:20 | <MikeSmith> | dunno what you mean by export… |
| 12:21 | <MikeSmith> | you mean export from Bitbucket? |
| 12:21 | <hsivonen> | that is, if I commit to the jing-trang bitbucket mirror, will it end up in the main jing-trang svn automagically? |
| 12:21 | <MikeSmith> | ah, no |
| 12:21 | <MikeSmith> | nope |
| 12:21 | <MikeSmith> | that'd be the downside, I guess |
| 12:22 | <hsivonen> | what value would a bitbucket mirror of jing-trang add other than pull-only users not having to deal with installing svn? |
| 12:22 | <MikeSmith> | that'd be it |
| 12:22 | <hsivonen> | I see |
| 12:22 | <MikeSmith> | that's the only value I had in mind |
| 12:22 | <MikeSmith> | just eliminating a dependency |
| 12:23 | <MikeSmith> | but I guess most users who have mercurial installed are very likely the have subversion installed also |
| 12:23 | <thiessenp> | (yup) |
| 12:23 | <hsivonen> | if build.py still supported pulling from upsteam svn, I'd be ok with providing an option to pull using hg instead, but I'd continue using svn |
| 12:24 | <hsivonen> | of course, I'd prefer the jing-trang project upsteam converting the Google Code project from svn to hg |
| 12:24 | <MikeSmith> | yeah |
| 12:24 | <MikeSmith> | indeed |
| 12:24 | <MikeSmith> | I guess maybe that's the better thing to try to get done |
| 12:25 | <hsivonen> | earlier, it seemed that Linus was just flaming the svn guys, but I'm inclined to think he was right |
| 12:25 | <hsivonen> | "CVS done right" isn't a good goal |
| 12:25 | <MikeSmith> | heh |
| 12:25 | <MikeSmith> | yeah |
| 12:25 | <MikeSmith> | I think history has proven that |
| 12:26 | <MikeSmith> | and based on my own experiences, I think think subversion deserves every ounce of grief heaped upon it |
| 12:26 | <jgraham> | hsivonen: an interesting difference between Opera and Mozilla is that we don't require 100% pass rate for a test to be added to our regression tracking system |
| 12:26 | <jgraham> | Therefore I don't think designing on the assumption that everyone will require that makes sense |
| 12:27 | <hsivonen> | jgraham: interesting |
| 12:27 | <hsivonen> | jgraham: we manually flag failing assertions in the test itself |
| 12:27 | <hsivonen> | by using todo() instead of ok() and todo_is() instead of is() |
| 12:27 | <jgraham> | Yes, that obviosuly doesn't work for shared tests |
| 12:28 | <hsivonen> | jgraham: and if known-to-fail assertions start accidentally passing, the tree is considered broken |
| 12:28 | <hsivonen> | jgraham: we probably wouldn't import a given test until it passed |
| 12:29 | <hsivonen> | that is, I expect we wouldn't just pull the testsuite wholesale without manual vetting |
| 12:29 | <jgraham> | That seems strange and wrong |
| 12:29 | <thiessenp> | :) |
| 12:29 | <jgraham> | Does Mozilla use any non-in-house tests at the moment? |
| 12:30 | <hsivonen> | jgraham: yes, but they've been mochitestified |
| 12:30 | <hsivonen> | jgraham: e.g. the DOM test suite has bogus parts modified |
| 12:30 | <jgraham> | Oh, wow |
| 12:30 | <hsivonen> | jgraham: and for html5lib tree builder tests, we have a file that lists the tests we don't pass |
| 12:30 | <hsivonen> | it's a short list |
| 12:31 | <hsivonen> | I'd expect us to maintain some kind of exclusion manifest for HTML WG tests |
| 12:31 | <MikeSmith> | hsivonen: btw, if you have a few minutes to review my patch for un-assbackwards-ing the error reporting for src-less script elements, the patch is at http://bugzilla.validator.nu/attachment.cgi?id=187 (and sent you by e-mail too, but I get the feeling your inbox is even more overloaded than mine is…) |
| 12:31 | <hsivonen> | MikeSmith: looking |
| 12:31 | <MikeSmith> | thanks |
| 12:33 | <jgraham> | Also, since I seem to be getting a, I think unjustified, reputation for optimising for percentages, I will point out the minutes of the last two HTML test telecons where I suggested that we stop publishing percentages |
| 12:34 | <hsivonen> | MikeSmith: looks good |
| 12:35 | <MikeSmith> | thanks, I'll go ahead and commit it |
| 12:35 | <jgraham> | My desire to have Tests and asserts seperate is based on the conceptual difference between necessary and sufficient conditions |
| 12:35 | <Philip`> | I put "@moz-todo" markers in my canvas tests, which cause the mochitestification process to output todo() instead of ok() etc |
| 12:35 | <hsivonen> | jgraham: I didn't think percentages were your fault |
| 12:35 | <Philip`> | (Not a very scalable solution) |
| 12:36 | <jgraham> | hsivonen: You and sicking both suggested that the Test/assert difference is motivated by a desire to make percentages work well |
| 12:36 | <hsivonen> | jgraham: oh has sicking commented as well? I didn't notice. |
| 12:36 | <jgraham> | On IRC last night |
| 12:36 | <hsivonen> | I see |
| 12:37 | <hsivonen> | Opera Mini usually doesn't show the last bits in the IRC logs, so I tend to miss late night IRC chatter when checking the logs in the bus in the morning |
| 12:37 | <jgraham> | So far Maciej has suggested using something closer to WebKit tests and you and sicking have suggested cloning Mochitests |
| 12:38 | <hsivonen> | jgraham: not surprising :-/ |
| 12:38 | <annevk> | also boring |
| 12:39 | <annevk> | I'd rather this was sorted out already so I could release some more stuff |
| 12:39 | <annevk> | (after converting appropriately) |
| 12:39 | <jgraham> | hsivonen: FWIW I consider depending on onerror a non-starter |
| 12:39 | <hsivonen> | jgraham: :-( |
| 12:40 | <annevk> | I would be fine with depending on onerror btw |
| 12:40 | <annevk> | It's something that will be fixed way before we're done |
| 12:40 | <hsivonen> | jgraham: why is depending on onerror a non-starter? who doesn't implement it? |
| 12:40 | <jgraham> | Apart from anything else it doesn't actually solve the problem because a single exception stops the whole file |
| 12:41 | <jgraham> | hsivonen: We dont' and webkit don't |
| 12:41 | <jgraham> | Maciej had the same opinion :) |
| 12:41 | <hsivonen> | jgraham: aren't Opera and WebKit supposed to be the fastest to add stuff? ;-) |
| 12:42 | <jgraham> | hsivonen: Even if we add it, it doesn't solve the problem |
| 12:42 | <hsivonen> | jgraham: only if you want more fine-grained data than test file passed/failed |
| 12:42 | <hsivonen> | jgraham: I think individual asserts are only interesting for finding out why the test file failed when it does |
| 12:43 | <jgraham> | Which we do if you can add thousands of tests in a file |
| 12:43 | <jgraham> | I totally disagree |
| 12:43 | <hsivonen> | worked for me with the whole html5lib test suite running as one mochitest |
| 12:43 | <jgraham> | It is more or less a requirement for us that we can see which subtests passed and which failed |
| 12:43 | <jgraham> | and that we get those results consistently even when we regress unexpected things |
| 12:44 | <hsivonen> | for us, you don't get to push until you've sorted out the test failures |
| 12:45 | <hsivonen> | so the process is 1) edit code, 2) push to try, 3) observe test failures, 4) change code or tests as appropriate until the tests don't fail |
| 12:45 | <hsivonen> | 5) push to m-c |
| 12:46 | <gsnedders> | hsivonen: For us, it's just there have to be no regressions from pass to fail. |
| 12:46 | <hsivonen> | where with HTML WG tests step #4 could involve editing an exclusion manifest |
| 12:46 | <jgraham> | Our process is similar except that we can cope with long-lived failures. That is not all tests must pass. But we need to get a consistent set of results from each file |
| 12:46 | <hsivonen> | isn't that equivalent to having a manifest of what failures to ignore as long as the failures don't abort the test? |
| 12:46 | <jgraham> | and we don't have to hack the test files to deal with that |
| 12:47 | <hsivonen> | if once in a while you have a failure that aborts the whole tests, it's easier to fix it right away than to create abstraction for it |
| 12:49 | <jgraham> | In general it is better to avoid such things |
| 12:50 | <jgraham> | Because it is not easy to tell that a file that used to give 25 results and now gives 5 results is a failure |
| 12:51 | <hsivonen> | does that mode of failure happen often enough to justify making tests harder to write? |
| 12:51 | <jgraham> | It seems to happen reasonably often |
| 12:51 | <gsnedders> | Yes, certainly. |
| 12:51 | <hsivonen> | (I realize I live in a glass house that's randomly orange) |
| 12:52 | <jgraham> | I really want to make tests easier to write, but it feels like the requirement to make them easier than almost any other testing framework is onerous |
| 12:52 | <hsivonen> | I just want them to be as easy as mochitests |
| 12:52 | <jgraham> | that is there are few test frameworks that will allow you to write a test without defining a function |
| 12:53 | <gsnedders> | Ours is certainly fairly orange too, but adding stuff with a design that we know to have caused issues before seems bad. |
| 12:53 | <gsnedders> | Because, uh, it'll likely lead to more randomness that'll just need to be fixed at a later date |
| 12:53 | <jgraham> | As far as I can tell MochiTests are conceptually broken in a way that is particularly bad for Opera |
| 12:57 | <hsivonen> | hey. maybe I've already been subscribed to public-html-testsuite |
| 12:57 | <hsivonen> | and that's why my attempt to subscribe didn't work |
| 13:05 | <hsivonen> | Does NeedsRadar on bugs.webkit.org in practice mean that Apple considers the bug a Safari release blocker? |
| 13:18 | <annevk> | Minefield consistently crashes on http://tc.labs.opera.com/apis/XMLHttpRequest/testrunner.htm |
| 13:19 | <annevk> | I filed a bug report |
| 13:19 | <annevk> | euh crash report |
| 13:19 | <annevk> | anything else? |
| 13:19 | <annevk> | (it also seems to score poorly, less than half are passing) |
| 13:19 | <annevk> | Opera meanwhile stops running after a while... |
| 13:19 | <annevk> | bah |
| 13:20 | <hsivonen> | annevk: doesn't crash for me |
| 13:20 | <annevk> | on Mac OS |
| 13:20 | <hsivonen> | annevk: Mac OS X 10.6.x in my case, too |
| 13:21 | <hsivonen> | annevk: so you didn't file a bug yet? |
| 13:21 | <annevk> | I don't have the latest it seems |
| 13:22 | <annevk> | when I go to about Minefield it indicates it is downloading something |
| 13:22 | <hsivonen> | looks like I'm not permitted to search crash reports by URL |
| 13:23 | <annevk> | no longer crashes in latest |
| 13:23 | <annevk> | sorry for the noise |
| 13:58 | <annevk> | "unacceptable to many of our corporate customers" |
| 13:58 | annevk | amused |
| 13:58 | <hsivonen> | annevk: where's the quote from? |
| 13:59 | <annevk> | http://www.w3.org/Bugs/Public/show_bug.cgi?id=11449 |
| 13:59 | <annevk> | another person confusing UI and submission formats |
| 13:59 | <jgraham> | Resolved:Missing the point |
| 13:59 | <hsivonen> | "Missing the point" were the exact words I had in mind |
| 14:00 | <moo_> | those imperialist bastards - time to push metric system and ISO dates forward :) |
| 14:02 | <charlvn> | by "contributor"? |
| 14:02 | <annevk> | whoa, charlvn is back? |
| 14:02 | <hsivonen> | I'm going to a front end developer meeting today. It's good to hear what the users of the stuff we spec think. |
| 14:02 | <annevk> | welcome back charlvn |
| 14:02 | <charlvn> | annevk: was i gone? :) |
| 14:03 | <charlvn> | thanks :) |
| 14:03 | <hsivonen> | charlvn: "contributor" is a generic address for people who haven't logged in |
| 14:03 | <annevk> | somewhere between now and like three years ago? |
| 14:03 | <charlvn> | hsivonen: ah ok thanks |
| 14:03 | <annevk> | at least, haven't seen you around |
| 14:03 | <charlvn> | annevk: lol i have just been lurking :) |
| 14:03 | <zcorpan> | hey charlvn |
| 14:03 | <charlvn> | hi zcorpan :) |
| 14:04 | <zcorpan> | wassup |
| 14:05 | <charlvn> | not much, just lurking (or i guess, not at the moment) :) |
| 14:11 | <annevk> | I guess at some point we should have some CSS to control the UI |
| 14:11 | <annevk> | bug has a point in that browser locale is not really enough |
| 14:14 | <hsivonen> | it's still amusing that "unacceptable to our corporate customers" is used as an argument amplifier |
| 14:20 | <annevk> | yeah, reminds me of the CDF WG |
| 14:20 | <annevk> | and IBM |
| 14:30 | <Rik`> | http://www.opera.com/ |
| 14:31 | <Rik`> | this is really cool, congrats ! |
| 14:31 | <annevk> | hmm, I wasn't invited |
| 14:34 | <annevk> | heh, didn't realize it was interactive until now |
| 14:35 | <Rik`> | just waiting for someone to decrypt the No subject in Drafts |
| 14:35 | <Rik`> | Great Opportunity in Sent is good :) |
| 14:37 | <annevk> | FW: Love the show |
| 14:37 | <annevk> | hahaha |
| 14:37 | <annevk> | oh god our marketing team |
| 14:40 | <jgraham> | Ah, I hoped the new tab button would work |
| 14:41 | <annevk> | yeah, was a whee bit disappointed by that too |
| 14:43 | <Rik`> | I'm disappointed that there is no "150 million Opera users worldwide" in the Sent folder. It should be there since the out of office reply is in received |
| 14:44 | <Rik`> | plus, 8 days to send an out of office reply is weird |
| 14:45 | <annevk> | hours, you mean? |
| 14:46 | <annevk> | was the first thing I checked too |
| 14:46 | <Rik`> | "On Mon, 22 Nov 2010 09:11, Jon von Tetzchner wrote:" |
| 14:47 | <Rik`> | "2010/12/01 13:37" |
| 14:47 | <Rik`> | looks like 8 days to me |
| 14:47 | <Rik`> | 9 actually |
| 14:48 | <annevk> | o_O |
| 14:48 | annevk | was just checking the time |
| 15:00 | hsivonen | wonders how 150 million users was measured |
| 15:00 | <zcorpan> | hmm, can't figure out what to do with the ones and zeros in Drafts |
| 15:00 | <zcorpan> | i guess i'm being slow |
| 15:00 | <hsivonen> | are they active daily users of the desktop version as determined by update pings? |
| 15:00 | <hsivonen> | or does it include Mini and Mobile users? |
| 15:00 | <hsivonen> | or is the metric something other than ADU? |
| 15:01 | <miketaylr> | zcorpan: i can spoil it for you if you like :) |
| 15:01 | <Rik`> | zcorpan: my timeline suggests something around OperaCodeCracker |
| 15:02 | <miketaylr> | binary -> ______ -> _______ |
| 15:03 | <zcorpan> | oh, i tried to decode as _______ but that also gave garbage. i guess i used a bad ______ decoder |
| 15:03 | <miketaylr> | ;P |
| 15:05 | <hsivonen> | solved! |
| 15:05 | <Philip`> | "Pongratulations! Lou cracked the code." |
| 15:05 | <Philip`> | I wonder who Lou is |
| 15:06 | <hsivonen> | Philip`: that's an interesting distortion |
| 15:06 | <hsivonen> | I didn't get the distortion |
| 15:06 | Philip` | just did a "tr [a-z] [n-za-m]" which wasn't quite enough |
| 15:10 | <jgraham> | Oh, solved |
| 15:10 | <jgraham> | Much good it will do me though |
| 15:17 | jgraham | wonders if the effects of nerd-sniping our own engineering department were carefully considered |
| 15:18 | <jcranmer> | jgraham: what could go wrong? |
| 15:19 | <jgraham> | Well some curious developer could be reading it on their mobile on the way home, start tring to solve it and get run over by a bus |
| 15:19 | <jgraham> | For example |
| 15:20 | <jcranmer> | said the actress to the bishop |
| 15:20 | <jcranmer> | nah, doesn't work as well |
| 15:55 | <annevk> | hahaha |
| 15:55 | <karlcow> | http://rebuildingtheweb.com/en/html5-shortcomings/ |
| 15:55 | <annevk> | maybe someone should point the TAG to http://lastweekinhtml5.blogspot.com/ |
| 15:56 | <annevk> | Vlad Alexander has been trolling HTML5 since before it was called that |
| 15:57 | <karlcow> | hehe annevk |
| 15:57 | <Ms2ger> | We should create a registry of known web standards trolls |
| 15:58 | <Ms2ger> | At IANA, perhaps? |
| 15:58 | <karlcow> | Ms2ger: I will be sure that each of us would be part of it at a point in time. Registry is not the good thing. |
| 15:58 | <karlcow> | a trollometer would be better |
| 15:58 | <paul_irish> | i think i successfully prevented this post from being published on Smashing Magazing |
| 15:59 | <karlcow> | but then it would create a race to become the highest troller :/ |
| 15:59 | <annevk> | problem? |
| 16:01 | <karlcow> | trollometer, yes |
| 16:01 | karlcow | is not a big fan of jackass and race to bottom |
| 16:02 | <beowulf> | paul_irish: climategate! |
| 16:02 | <nimbupani> | trolling html5 since 1995? |
| 16:03 | <jgraham> | It's OK if the registery was at IANA no one would ever get registered |
| 16:03 | <jgraham> | Unless, perhaps, they came with their own RFC certifying their trollishness |
| 16:04 | <Ms2ger> | I believe an axiomatic proof of trolling should be enough. |
| 16:04 | <annevk> | karlcow, as in http://memegenerator.net/Troll-Face/ImageMacro/877773/Troll-Face-problem.jpg ;) |
| 16:05 | <karlcow> | heh |
| 16:05 | <jgraham> | Ms2ger: So there would be exactly one troll in the registry? |
| 16:05 | <Ms2ger> | Possibly |
| 16:08 | <Lachy> | that article is so blatantly written to make HTML5 look worse than it is. |
| 16:09 | <Lachy> | it seems to think that just because something is technically permitted in HTML, given its inherrent flexibility, that it is somehow implicitly encouraged. Like using heading elements out of numerical order, even though the spec advices otherwise. |
| 16:09 | <Lachy> | *advises |
| 16:13 | <zcorpan> | HTMLTidy barks on the following valid HTML4: <table><tbody><tr><td> Hello </table> <p>LOL</p> |
| 16:13 | <zcorpan> | (iirc) |
| 16:14 | <zcorpan> | no need to use new html5 features to demonstrate brokenness in HTMLTidy |
| 16:17 | <jgraham> | But it is nice rhetoric anyway. After all how many people will actually test that |
| 16:17 | <jgraham> | ? |
| 16:24 | <annevk> | hsivonen, Gecko does not always override iso-8859-1 with windows-1252? |
| 17:24 | <annevk> | heh, script insertion was annoying seven years ago -- http://ln.hixie.ch/?start=1079103487&order=-1&count=1 |
| 17:25 | Ms2ger | had that post open |
| 17:26 | <MikeSmith> | +1 to Ms2ger suggestion of creating an IANA registry of known web standards trolls |
| 17:29 | <jgraham> | Disappointingly, it seems like a stretch to classify organisations that fail to keep their registries up to date as "trolls" |
| 17:30 | <MikeSmith> | no, that's just plain shirking of actual responsibilities |
| 17:31 | <MikeSmith> | clearly we need to first identify some Designated Experts on trolling |
| 17:31 | <MikeSmith> | that part should not be so hard at least |
| 17:53 | <TabAtkins> | Ms2ger: It's really not worthwhile responding to Shelley's comments on bugs, especially ones that are so clearly trolling. Just ignore them. |
| 20:09 | <AryehGregor> | It's pleasant to actually agree with Shelley Powers and disagree with TabAtkins for once. http://www.w3.org/Bugs/Public/show_bug.cgi?id=11449 |
| 20:14 | <Dashiva> | Why? |
| 20:16 | <Dashiva> | This seems like a perfect example of something the browser should control |
| 20:16 | <Dashiva> | The author will be some ignorant twit who thinks everyone in the whole world uses the exact same system as himself, and hardcode that |
| 20:17 | <AryehGregor> | For the same reason authors are allowed to make their website yellow on neon green. |
| 20:18 | <AryehGregor> | There's no reason a user couldn't override the author-stated preference, or that some browsers ignore the author-stated preference in some or all cases, so the final control is with the user as usual. |
| 20:18 | <Dashiva> | That's not an argument for providing author control in the first place |
| 20:18 | <AryehGregor> | Generally speaking, authors should be allowed to control the appearance of anything that falls within their page if they want, barring security concerns. |
| 20:19 | <AryehGregor> | Authors are going to want to match the rest of their site. If their application is hardcoded to output dates in twelve-hour time, for instance, they might want it to accept them in twelve-hour time too for consistency. |
| 20:19 | <AryehGregor> | Particularly if the inputs and outputs are mixed together on the same page. |
| 20:20 | <AryehGregor> | This is even more important for things that are ambiguous, like mm/dd/yy vs. dd/mm/yy. One site should use the same convention cross-site. |
| 20:20 | <Dashiva> | So you want to allow it to accomodate another abusive use case (hardcoding a date format the user doesn't necessarily want) |
| 20:20 | <AryehGregor> | Users don't necessarily want their browser's default date format. How many know how to change it? |
| 20:21 | <AryehGregor> | Consistency is good too. |
| 20:22 | <AryehGregor> | If a British person is using an American site and is asked for a date in xx/xx/xx format, they'll probably assume it's dd/mm/yy. This should not change if the author uses HTML5 vs. custom controls. |
| 20:22 | <AryehGregor> | Users won't be able to distinguish. |
| 20:23 | <AryehGregor> | Plus, pragmatically, if HTML5 native controls aren't styleable enough, authors will be pushed to use scripted controls instead, which is worse from every perspective. |
| 20:23 | <Dashiva> | Part of the point of native date controls is that they won't have to guess in the first place |
| 20:23 | <Dashiva> | The date picker will clearly distinguish between date and month |
| 20:23 | <AryehGregor> | (These arguments are essentially the same as I made for browsers auto-rendering <date>.) |
| 20:23 | <AryehGregor> | It might, or might not. Have you seen WebKit's "date picker"? :) |
| 20:23 | <AryehGregor> | (yeah, I know, WebKit HTML5 forms are just completely broken) |
| 20:24 | <Dashiva> | Yes, they really messed that one up |
| 20:24 | <Dashiva> | At least Opera had the excuse of being a "legacy implementation" |
| 20:24 | <AryehGregor> | Opera's implementation is pretty terrible, but still vastly better than WebKit's. |
| 20:25 | <AryehGregor> | But basically, sites should be consistent. An American site should use American conventions consistently, not American conventions where the content is author-written but some arbitrary convention where using native inputs (except where necessary for security). |
| 20:25 | <AryehGregor> | Anything that's visually part of the site rather than the UI should be primarily controlled by the author. |
| 20:37 | <Hixie> | TabAtkins: sounds good to me; give me some context also when you send me text (if you haven't yet) |
| 20:58 | <TabAtkins> | Hixie: Will do. I'm writing the text now. I'm quite certain it will be inadequate, but it should be clear enough to write the *proper* text from. |
| 20:58 | <TabAtkins> | I don't know the proper way to do idl/etc. |
| 20:58 | <Hixie> | heh |
| 20:58 | <Hixie> | no worries |
| 20:59 | <Hixie> | all i need is the hook in the other spec |
| 20:59 | <TabAtkins> | kk, that'll definitely be present - it's really simple. |
| 20:59 | <Hixie> | cool |
| 20:59 | <TabAtkins> | HTML5 just needs a single function on Document, and some language about which elements can be associated while out of the document. |
| 20:59 | <Hixie> | what are we actually doing here btw? :-) |
| 20:59 | <TabAtkins> | I'll show you in a few minutes when I finish this text and push it. ^_^ |
| 20:59 | <Hixie> | k :-) |
| 21:13 | <TabAtkins> | Hixie: http://dev.w3.org/csswg/css3-images/#element-reference |
| 21:14 | <TabAtkins> | HTML will just need to define the mapping mechanism for the element-ref() function. |
| 21:14 | Hixie | wonders what legitimate use cases there are for element() that don't involve abusing the semantics of the html page |
| 21:14 | <TabAtkins> | Reflections, for one. |
| 21:15 | <Hixie> | wouldn't you want reflections to reflect whatever the element's rendering is, not what an element referenced by ID's rendering is? |
| 21:15 | <TabAtkins> | #foo::after { content: element(#foo); transform: reflectY(-1); /* or whatever*/ filter: some-mask(); } |
| 21:16 | <TabAtkins> | Another is, frex, slideshow previews for the next/prev slide. |
| 21:17 | <TabAtkins> | Thumbnails in general, actually. |
| 21:17 | <Hixie> | seems like img::after { content: element; transform: reflectY(-1); /* or whatever*/ filter: some-mask(); } would be infinitely more useful |
| 21:17 | <TabAtkins> | Well, here, I'll let mozHacks speak for me: http://hacks.mozilla.org/2010/08/mozelement/ |
| 21:17 | <othermaciej> | I think Hixie is wondering about non-self reference use cases |
| 21:17 | <othermaciej> | live thumbnails are one |
| 21:17 | <othermaciej> | e.g. for a slideshow program |
| 21:17 | <othermaciej> | er, presentation program |
| 21:17 | <othermaciej> | thumbnails of slides at the side, live interactive large version in the main content |
| 21:17 | <Hixie> | seems like that would be better done with XBL, but failing the existence of XBL, that it would be content, not style |
| 21:18 | <Hixie> | i mean, any time that the style sheet must be involved for the page to be usabel is a bit dubious, imho |
| 21:18 | <Hixie> | which it would seem is the case here |
| 21:18 | <TabAtkins> | Hixie: I don't understand how XBL would solve the problem of "render a tiny version of some other element, without duplicating the element". |
| 21:25 | <Hixie> | xbl would solve the problem of "i need to have a pager widget that shows thumbnails of these elements", presumably using element-ref() and so forth |
| 21:25 | <Hixie> | the blog post you linked to fails to answer the question |
| 21:25 | <Hixie> | it just has a lot of use cases i consider abuses of the platform :-) |
| 21:25 | <Hixie> | the SVG ones make more sense |
| 21:25 | <Hixie> | but really the SVG should be embedded in the CSS, not the HTML |
| 21:25 | <Hixie> | anyway, for element-ref(), you need some sort of hook for me to spec. I recommend making the hook <dfn> and styled bold, so that later editors are less likely to change the wording out from under me :-) |
| 21:25 | <TabAtkins> | Hixie: What do you mean? At the moment I'm saying "The host language for the document may expose a method of associating elements with identifiers for the purpose of this function; CSS does not define how this occurs." |
| 21:26 | <TabAtkins> | The intention is that HTML defines a js function to do the mapping. Do I need to expose anything further on the CSS side? |
| 21:26 | <Hixie> | something like "When an element is said to have *a specified element-ref() identifier*, the user agent must..." or some such |
| 21:27 | <TabAtkins> | Ah, kk. One sec. |
| 21:31 | <Hixie> | right, you don't give me a hook :-) |
| 21:31 | <Hixie> | you just say that there is one |
| 21:31 | <Hixie> | some text that can unambiguously be used in both specs so that people know that there is a link |
| 21:31 | <Hixie> | lunch, bbiab |
| 21:31 | <TabAtkins> | Done. "specified element-ref() identifier" it is. |
| 21:32 | <TabAtkins> | The sentence "The host language for the document may expose a method of giving elements a specified element-ref() identifier associating the element with an identifier" isn't sufficient? |
| 21:36 | <roc> | TabAtkins: are you going to respond to my last email on the element() thread, then? |
| 21:36 | <TabAtkins> | roc: Yup. |
| 21:37 | <roc> | I wonder what Hixie means by "abuses of the platform" |
| 21:38 | <TabAtkins> | Presumably things like adding an empty <div> for the sole purpose of receiving an element() background. |
| 21:42 | <roc> | You can use an <img> if that makes you feel better |
| 21:42 | <TabAtkins> | + the content property. |
| 21:43 | <roc> | allowing element() as a value for the content property would make sense |
| 21:43 | <TabAtkins> | Well, that part's automatic. 'content' accepts images as a value, making it act like a replaced element. |
| 22:04 | <TabAtkins> | roc: Where was that email? It's not in my inbox, and I can't find it through searching right now. |
| 22:10 | <TabAtkins> | roc: Never mind, finally found it by just searching for messages you've sent. |
| 22:10 | <roc> | now I can't find it :-) |
| 22:10 | <TabAtkins> | Sent to WHATWG on Nov 23. |
| 22:11 | <roc> | ta |
| 22:12 | <roc> | so, what do you think about my proposal there? :-) |
| 22:12 | <TabAtkins> | I didn't like it, which is why I came up with something else. ^_^ |
| 22:14 | <TabAtkins> | what out-of-document elements are accepted by setImageElement currently? Just <img>, <canvas>, and <video>? |
| 22:14 | <roc> | yes |
| 22:14 | <TabAtkins> | Should we allow other elements, with an argument for a "viewport" width and height to render them in? |
| 22:14 | <roc> | well, it accepts all of them, but only those will render if they're not in a document |
| 22:14 | <roc> | note that allowing setImageElement to refer to nodes in a document is useful |
| 22:15 | <TabAtkins> | Yes, definitely. |
| 22:15 | <TabAtkins> | That's not in question. I'm just wondering about the out-of-document cases. |
| 22:15 | <roc> | trying to render arbitrary elements not in a document is a real pain |
| 22:15 | <roc> | you'll have to construct a presentation for them somehow |
| 22:15 | <roc> | not just define the viewport, but also the medium |
| 22:15 | <TabAtkins> | All right, no problem. We can just disallow them. |
| 22:15 | <TabAtkins> | Ah, true. |
| 22:17 | <roc> | I honestly think that using a selector where a single element is needed is a bad idea, and syntax like first() to resolve the ambiguity would be nice |
| 22:17 | <TabAtkins> | I prefer using element() and elements() to make the distinction. |
| 22:17 | <TabAtkins> | (That's what I've currently got in the Positioned Layout draft.) |
| 22:18 | <roc> | what if someone wants to use the last element that matches a selector? |
| 22:18 | <TabAtkins> | Shrug? |
| 22:18 | <roc> | why is first() bad, really? |
| 22:18 | <TabAtkins> | Having to immediately nest a function feels icky. |
| 22:18 | <TabAtkins> | I don't even particularly like nesting rgba() colors in gradients, but that's tolerable. |
| 22:19 | <TabAtkins> | "background-image: element(first(selector));", though... |
| 22:19 | <roc> | it's only for the non-ID selector case, which isn't even a big use case |
| 22:19 | <roc> | also |
| 22:19 | <TabAtkins> | Not a big use-case for images, no. Somewhat larger use-case for Positioned Layout. |
| 22:20 | <roc> | element(self) is needed |
| 22:20 | <TabAtkins> | I agree. |
| 22:20 | <TabAtkins> | I think that Selectors should define :self or :this or something, which, when the selector is run in the context of an element, matches that element. (Matches nothing when just run in the context of a document.) |
| 22:21 | <roc> | what are the large use-cases for Positioned Layout where a selector matches multiple elements and you want to only get the first one? |
| 22:21 | <TabAtkins> | And presumably matches the superior parent of a pseudoelement. |
| 22:21 | <Hixie> | roc: i mean doing something in html which would look broken if you removed the css, for instance |
| 22:21 | <Hixie> | TabAtkins: lgtm |
| 22:21 | <roc> | why shoehorn 'self' into selectors, if it's only going to work in element() anyway? |
| 22:22 | <TabAtkins> | roc: It should work in element-scoped querySelector, too. |
| 22:22 | <Hixie> | TabAtkins: so actually... should this be in html? it seems like it might make more sense in cssom. I can add it for now and see if anne wants to take it, if you like. |
| 22:23 | <TabAtkins> | roc: re matching multiple elements and only getting the first for Positioned Layout, hm, not sure actually. Thinking on my use-cases, they're all for elements(). |
| 22:23 | <TabAtkins> | Hixie: Yeah, sure, either one makes sense. It's just not appropriate for CSS itself, since the mapping is host-language specific. |
| 22:23 | <Hixie> | there are other host languages than the DOM? :-) |
| 22:24 | <TabAtkins> | Theoretically? Sure. We like to maintain the fiction that CSS is used outside of HTML and SVG. |
| 22:24 | <Hixie> | fair enough |
| 22:24 | <Hixie> | i guess websrt in a way is one :-) |
| 22:24 | <TabAtkins> | Ah, that's right! I keep forgetting to use that as an example. |
| 22:24 | <roc> | I don't even know of any use-cases for element() images where we want a non-ID selector |
| 22:24 | <roc> | so maybe you should just make elements() take a selector and element() take a bare <ident> or #id selector |
| 22:24 | <TabAtkins> | roc: I'll think on it a bit, but you may be right. |
| 22:25 | <TabAtkins> | And if so, then yeah, I'll just merge the two functions. |
| 22:25 | <TabAtkins> | elements() could always take an "nth" argument as well, after all. |
| 22:26 | <TabAtkins> | Hixie: Make the hook "specified element reference identifier", to leave my options open. I'll change the draft in a bit to reflect that. |
| 22:27 | <Hixie> | k |
| 22:27 | <Hixie> | that's a bit generic |
| 22:27 | <Hixie> | i mean, it even describes id="" |
| 22:27 | <TabAtkins> | Insert "CSS" after "specified". |
| 22:27 | <Hixie> | k |
| 22:28 | <jamesr_> | roc: how do you implement that? you redraw the element at the new position, not reuse the old rasterization right? |
| 22:29 | <roc> | Hixie: actually I think all the examples in the hack.mozilla.org post would fall back usefully with CSS disabled, or could be easily made to |
| 22:30 | <Hixie> | TabAtkins: so i went to add soemthing to Window, but realised it's more of a document-specific thing, so i went to add it to HTMLDocument, something like document.setRef(element, ref), but then i realised that the document was superfluous and what really matters is the element, so i went to HTMLElement and tried element.setRef(ref) but then realised that that is basically an attribute so went to add element.ref = ref but then i realised... that's the same as .id |
| 22:30 | <Hixie> | TabAtkins: so... do i just define the element's ID as being the "specificed CSS element reference identifier"? that seems redundant. |
| 22:30 | <roc> | for example the slide thumbnails should be an <ol> with each list item containing the slide title, being a link to the slide itself |
| 22:30 | <TabAtkins> | You want to add it to Document. That way you can associate out-of-document script-created elements with the CSS of a page. |
| 22:31 | <roc> | CSS adds the thumbnails as decoration |
| 22:31 | <roc> | jamesr_: yes, we redraw the element |
| 22:31 | <Hixie> | roc: if you're using style="", imho you're almost always doing something wrong, or at least, the platform isn't supporting you well enough |
| 22:31 | <TabAtkins> | And it's useful to have it distinct from id, so script, particularly libraries, can swap things around without having to interfere with ids across the document. |
| 22:32 | <jamesr_> | roc: so if you have, say, text in the element then you re-rasterize it potentially differently (i.e. if there's a rotation applied then subpixel AA has to be turned off)? |
| 22:32 | <Hixie> | roc: i think it would make perfect sense to have CSS be able to grab the relevant thumbnail from the target of an in-document link, for example... but having to duplicate the information in the style="" attribute means we (as platform designers) have failed to make it easy enough. |
| 22:32 | <Hixie> | TabAtkins: there are no document-less elements, elements are always associated with a document even when not technically _in_ a document |
| 22:33 | <TabAtkins> | Okay, but CSS doesn't know about them if they're not in the document tree. |
| 22:33 | <roc> | Hixie: yeah OK, but that starts to make the feature rather elaborate |
| 22:33 | <roc> | jamesr_: that is correct |
| 22:34 | <roc> | however there are many situations where reusing the rasterization would be desirable, so we'll do that in due course |
| 22:34 | <roc> | as an optimization |
| 22:34 | <TabAtkins> | Hixie: Even if they're associated with the document, they're not children of the root element. |
| 22:34 | <Hixie> | roc: elaborate is fine, especially if it will save authors work on the long run. Having to update every single style="" every time you change your mind on exactly how this works is what CSS was designed to save us from |
| 22:34 | <jamesr_> | yeah, so long as you know it'll rasterize the same way no reason not to re-use the old rasterization |
| 22:35 | <Hixie> | TabAtkins: right, so i can just reuse the ID, right? |
| 22:35 | <roc> | in practice for stuff like the slide thumbnails you use script to generate them |
| 22:35 | <Hixie> | TabAtkins: i mean, i can say that the ID is the "specificed CSS element reference identifier" |
| 22:36 | <TabAtkins> | roc: Can you explain the reasoning behind the "currentslide" example further? That's the reason I'm not solely using the id. |
| 22:36 | <roc> | so generating the right style attributes is very little extra work |
| 22:37 | <roc> | now sure, ideally CSS would have some kind of templating system and be able to generate all this extra content without scripting and keep it in sync automatically, but that's some way out |
| 22:37 | <Hixie> | oh, the currentslide one is interesting |
| 22:37 | <Hixie> | that's backwards from what i was imagining |
| 22:37 | <Hixie> | it's not setting the element to an ID, it's setting an ID to an element |
| 22:37 | <Hixie> | s/ID/reference/ |
| 22:37 | <TabAtkins> | Yeah. |
| 22:37 | <Hixie> | ok, that makes sense |
| 22:38 | <TabAtkins> | kk |
| 22:38 | <roc> | TabAtkins: what needs to be explained? |
| 22:38 | <TabAtkins> | roc: Never mind, Hixie found the example and was convinced by it. |
| 22:39 | <Hixie> | can anyone think of another API in the platform somewhere that sets a key to a value? i guess localStorage is one example |
| 22:39 | <Hixie> | should it be possible to read back from this API? |
| 22:39 | <Hixie> | find out what the currentslide is? |
| 22:40 | <TabAtkins> | getRef? |
| 22:40 | <TabAtkins> | Makes sense, yeah. |
| 22:40 | <Hixie> | well if we expose that the way to do it is just to have an attribute that returns a map object / hash / whatever we call it |
| 22:40 | <TabAtkins> | Sure. |
| 22:43 | <TabAtkins> | Hixie, are you okay with sharing the contents of http://lists.w3.org/Archives/Member/w3c-css-wg/2007JanMar/0535.html? |
| 22:43 | <TabAtkins> | (I assume yes.) |
| 22:43 | <Hixie> | sure |
| 22:44 | <Hixie> | what should this API be |
| 22:44 | Philip` | wonders when the Member archives will appear on Wikileaks |
| 22:44 | <Hixie> | document.what.currentslide = element |
| 22:45 | <Hixie> | document.cssElementMap.currentslide = element? |
| 22:45 | <TabAtkins> | That sounds good to me. |
| 22:49 | <TabAtkins> | roc: All right, I agree with you. Removing the element-ref() function now. |
| 22:52 | <Hixie> | does that mean i should stop adding this feature |
| 22:53 | <TabAtkins> | Oh, no. No effect on what you're doing. |
| 22:53 | <TabAtkins> | Just rejiggering the syntax on my side. |
| 22:53 | <Hixie> | k |
| 22:54 | <Hixie> | what reference should I use for this "specified CSS element reference identifier" term btw? |
| 22:55 | <Hixie> | [CSSGENCON]? |
| 22:55 | <TabAtkins> | CSSIMAGES |
| 22:55 | <Hixie> | k |
| 22:56 | <roc> | is there an actual use case for reading back from this map? |
| 22:56 | <Hixie> | well presumably it's not unusual for code to reference the current slide |
| 22:57 | <Hixie> | why have the JS and the browser maintain this mapping separately if the browser can just expose the current setting |
| 22:57 | <Hixie> | it's less likely to go out of sync if there's nothing to sync |
| 22:57 | <roc> | because developers are likely to have a variable 'currentSlide' somewhere and find that more natural then using document.cssElementMap['currentSlide']? |
| 22:58 | <roc> | but OK |
| 22:59 | <TabAtkins> | Dammit Hixie, your UTF8 decoder does not cater to my needs. I have space-separated binary bytes! |
| 22:59 | <Hixie> | doesn't it support that? |
| 22:59 | <Hixie> | what's your input look like |
| 22:59 | <TabAtkins> | 01010101 01000111 01001010 ... |
| 23:00 | <TabAtkins> | And it takes a trivial amount of effort to convert it to one of your accepted binary input forms! |
| 23:03 | <TabAtkins> | Hm, now I've converted to the 0b-prefixed form and selected "Freeform numeric", but it's claiming I didn't provide any bytes. |
| 23:04 | <Hixie> | odd, freeform wfm |
| 23:05 | <Hixie> | anyway i've added "binary" as an option |
| 23:06 | <TabAtkins> | Huh. Same error still, with the "binary" option. How odd. |
| 23:07 | <Hixie> | paste the start of your data? |
| 23:07 | <TabAtkins> | 01010101 01000111 01001010 01101000 01100100 01000111 01010110 |
| 23:07 | <Hixie> | hm |
| 23:07 | <Hixie> | only the first got converted for me |
| 23:07 | <Hixie> | odd |
| 23:08 | <TabAtkins> | Hm, all these bytes are in the low half, so if they're text decoding them as utf8 wont' do anything special. |
| 23:08 | <TabAtkins> | I already got gibberish doing a plain ascii conversion. |
| 23:08 | <TabAtkins> | This looks like base64 maybe, though. |
| 23:09 | <Hixie> | ok i got it working again |
| 23:09 | <TabAtkins> | All right, almost certainly base64. Looks like it's now enciphered. |
| 23:10 | <TabAtkins> | Somebody with the java runtime wanna help me out real quick? |
| 23:11 | <TabAtkins> | (Need to run an applet that'll crack a viginere cipher key.) |
| 23:11 | <TabAtkins> | Wait... this is rot13. |
| 23:12 | <Rik`> | TabAtkins: aren't you decoding some Opera stuff ? :) |
| 23:12 | <TabAtkins> | Yup. |
| 23:13 | <Rik`> | I have a Python one liner on my twitter account if you need :) |
| 23:13 | <TabAtkins> | Finished already. ^_^ |
