| 00:02 | <karlcow> | gsnedders: the best places are #swig, then http://lists.w3.org/Archives/Public/semantic-web/, then you can ask me in private if you want. I can be slow to reply though sometimes :) |
| 00:03 | <gsnedders> | karlcow: #swig on irc.w3 I guess? |
| 00:05 | <smedero> | gsnedders: on freenode I believe |
| 00:05 | <karlcow> | on freenode |
| 00:06 | <gsnedders> | k |
| 00:06 | <gsnedders> | thx |
| 01:12 | <nessy> | Hixie: thanks a lot! |
| 01:12 | <Hixie> | np |
| 02:07 | <rubys> | hixie: no nothing for you to do, other than confirm your intent to attend the meeting in SF next month |
| 02:12 | <annevk2> | why exactly does <input type=image> not allow value=""? |
| 02:12 | annevk2 | missed that |
| 02:12 | <annevk2> | apparently html5-diff does not mention it |
| 02:19 | annevk2 | now vaguely recalls some debate |
| 02:54 | <annevk2> | so Opera does not use value=, Safari does |
| 02:54 | <annevk2> | no idea about other browsers |
| 03:14 | <annevk2> | i.e. <input type=image name=test value=test> becomes ?test.x=2&test.y=2&test=test |
| 04:16 | <annevk2> | I guess the fact that validator.nu is less than helpful for <input type=text alt=test> is already known? |
| 04:29 | <annevk2> | Hixie, I would remind you that EventSource.URL should return an absolute URL |
| 04:30 | <annevk2> | Hixie, as a bonus for being so late (I would do it after 10 minutes I believe) I'll add that WebSocket.URL has the same issue |
| 04:30 | <annevk2> | nice copy & paste job |
| 04:31 | <annevk2> | Hixie, it's sort of tempting to suggest they should have a common base interface |
| 04:41 | <Hixie> | yeah i considered that |
| 04:42 | <Hixie> | WebSocket.URL doesn't have to return an absolute one |
| 04:42 | <Hixie> | because the constructor throws if the URL isn't absolute |
| 04:42 | <Hixie> | anyway i filed a bug |
| 04:42 | <Hixie> | thanks |
| 05:15 | <MikeSmith> | Hixie: are you planning to add accesskey to the Rendering section? |
| 05:21 | <annevk2> | ah, fair point about WebSocket.URL |
| 05:38 | <Hixie> | MikeSmith: i don't know what to do about accesskey |
| 05:40 | <Hixie> | fundamentally it doesn't work |
| 05:40 | <Hixie> | but we need _something_ |
| 05:40 | <MikeSmith> | yeah, seems like |
| 05:42 | <MikeSmith> | even documenting it to say "this doesn't work interoperably, but here's what the behavior is in some contexts" |
| 05:43 | <Hixie> | that's the minimum we'll do, yeah |
| 05:43 | <Hixie> | i was hoping for better though |
| 05:43 | <Hixie> | (that'll be in the obsolete apis section at some point) |
| 05:43 | <MikeSmith> | I see |
| 05:43 | <MikeSmith> | we definitely do still need some replacement for it that actually works |
| 05:43 | <Hixie> | yeah |
| 05:44 | <Hixie> | i have a folder with 52 e-mails about it |
| 05:44 | <Hixie> | eta april |
| 05:49 | <MikeSmith> | OK |
| 06:12 | <Hixie> | writing author and UA criteria for DOM APIs poses an interesting problem |
| 06:12 | <Hixie> | i want it to be possible to hide the UA requirements |
| 06:12 | <Hixie> | but that means that DOM requirements end up reading as: |
| 06:13 | <Hixie> | The /foo/ attribute returns the foo. On getting, the user agent must return the foo. |
| 06:13 | <Hixie> | which sounds stupid. |
| 06:13 | <Hixie> | I wonder how to address this problem. |
| 06:34 | <annevk2> | that's not an author criteria though |
| 06:35 | <annevk2> | that's just explaining how it works |
| 06:45 | <annevk2> | hmm, cool, HTML5 doesn't create CDATASection nodes for text/html |
| 06:47 | <Hixie> | well we need something that tells the author how it works, i mean, i can't just remove the UA criteria leaving just the IDL |
| 06:47 | <Hixie> | or can we? |
| 06:51 | <annevk2> | no, but an introduction section that explains all methods should suffice |
| 06:52 | <MikeSmith> | If the UA criteria are accurately expressed by the IDL, seems like the prose could actually be removed. |
| 06:52 | <MikeSmith> | nm, annevk2 just clarified |
| 06:53 | <annevk2> | it might be a lot of work, but both an author and UA section for each IDL would be ideal, I think |
| 06:53 | <annevk2> | with the same pedantic detail as writing versus parsing HTML |
| 07:01 | <hsivonen> | In the future, I want to run a benchmark to see if dropping foster-parented stuff on the floor if parent gone would speed up the parser by making tree operation objects smaller by 25% |
| 07:08 | <annevk2> | sounds like it would use less memory |
| 07:13 | <hsivonen> | annevk2: using less memory for the operation queue would be the obvious effect. the execution speed effect off a more tightly packed queue and one fewer smart pointers to zero on each event is less obvious |
| 07:16 | <Hixie> | annevk2: it's not really clear to me what the author side would say |
| 07:17 | <annevk2> | it would say what the effect would of invoking the method or getting/setting an attribute |
| 07:17 | <annevk2> | would be* |
| 07:20 | <Hixie> | hmm |
| 07:21 | <annevk2> | how else would authors know whether it throws an exception for instance? |
| 07:21 | <annevk2> | and when it would throw, etc. |
| 07:28 | <MikeSmith> | wording could be "Set the foo attribute to..." |
| 07:28 | <MikeSmith> | hmm, or does RFC 2119 discourage that? |
| 07:29 | <hsivonen> | Hixie: is it bad if authors see that a setter must set and a getter must get instead of seeing that a setter sets and a getter gets? |
| 07:32 | <Hixie> | annevk2: generally exceptions only happen when the input is invalid |
| 07:34 | <annevk2> | e.g. how does an author know how to use WebSocket()? |
| 07:35 | <Hixie> | hsivonen: not sure what you mean? |
| 07:35 | <Hixie> | annevk2: the intro section |
| 07:35 | <annevk2> | or that document.domain cannot be set to www.example.org after it has been set to example.org |
| 07:35 | <Hixie> | annevk2: yeah, dunno |
| 07:35 | <Hixie> | an intro paragraph i guess |
| 07:36 | <annevk2> | Hixie, or the canvas 2D interface |
| 07:36 | <Hixie> | that one's gonna be tough |
| 07:36 | <annevk2> | I guess that intro paragraph is the authoring section I'm talking about |
| 07:37 | <annevk2> | I see intro more as explaining what the feature is for |
| 07:38 | <hsivonen> | wow. http://code.google.com/p/doctype/wiki/ArticleHereComesTheSun |
| 07:50 | <roc> | hsivonen: fortunately that is going away |
| 07:57 | <MikeSmith> | http://d.hatena.ne.jp/gyuque/ |
| 08:02 | <hsivonen> | MikeSmith: very cool. Too bad anti-aliasing in the cloth demo in Firefox antialiases against the black background leaving black lines in the ceiling |
| 08:03 | <hsivonen> | the problem is not visible in the Chrome screenshot, but then Chrome cheats and doesn't do AA |
| 08:05 | <hsivonen> | I wonder if the issue could be remedied by multiplying the canvas backing buffer dimensions by an integer, telling canvas to turn AA off upon draw to the backing buffer and then doing a bicubic downsample for display |
| 08:06 | <MikeSmith> | hsivonen: gyuque is the same guy who put together the reflow demos for the last Mozialla 24 |
| 08:06 | <MikeSmith> | dunno if you saw those |
| 08:06 | hsivonen | has no idea how seamless AA is done properly in 3D accelerators |
| 08:06 | <hsivonen> | MikeSmith: do you mean the ones that visually annotated wikipedia reflow? |
| 08:07 | <MikeSmith> | yeah, that was one of them |
| 08:11 | <roc> | hsivonen: they use multi-sampling, i.e., roughly what you describe |
| 08:16 | <hsivonen> | roc: in that case, perhaps it would be useful to be able to request a high-res backing buffer without AA in the buffer itself for canvas |
| 08:16 | <roc> | it's not clear that's really what you want for a software renderer |
| 08:17 | <hsivonen> | too slow? |
| 08:17 | <roc> | maybe |
| 09:00 | jgraham | sees public-html email backlog since 19:00Z last night, cries |
| 09:02 | <hsivonen> | tantek: you asked earlier if I had questions about XMDP. What should microformat consuming software do with XMDP profiles? |
| 09:02 | <tantek> | short answer is - build a dictionary |
| 09:02 | <hsivonen> | tantek: do they in practice? |
| 09:03 | <tantek> | longer answer will take too long at this hour for me I'm afraid, I will have to take your question and get back to you (probably with a wiki page) |
| 09:03 | <hsivonen> | tantek: ok. thanks |
| 09:03 | <tantek> | in practice they haven't needed to no. |
| 09:03 | <tantek> | but parsers *could* parse XMDP profiles, create a dictionary of terms (each with unique URI from the XMDP) and then have fully URI qualified terms |
| 09:04 | <tantek> | like other semantic techniques that use URI qualified terms :) |
| 09:04 | <hsivonen> | tantek: only if people actually used XMDP profiles all the time they use microformats |
| 09:04 | <tantek> | this was a deliberate design decision on my part, so one could define and use microformats, and have them be based in URIs |
| 09:05 | <tantek> | hsivonen, the 15-years-ago-parallel to that statement is: only if people actually used DOCTYPES all the time they use HTML. |
| 09:06 | <hsivonen> | tantek: was basing them in URIs a feature for microformats themselves or is it a feature for compatibility with other semantic technology communities? |
| 09:06 | <jgraham> | It seems that that case is a bit different because the use of doctypes has been driven by a desire to enable backwards-incompatible features in UAs |
| 09:06 | <tantek> | I created XMDP intentionally so that microformats always would/could have compatibility with the semantic technologies being used at the time. |
| 09:07 | <hsivonen> | tantek: ok. Do you see such compatibility still as a necessary thing and XMDP as a good way to get the compatibility? |
| 09:08 | <tantek> | yes I see XMDP as a good way to get the compatibility |
| 09:08 | <tantek> | is the compatibility "necessary"? I think that is still up for debate, but while it is up for debate, I prefer taking the conservative path and at least continuing to *enable* the compatibility. |
| 09:09 | <hsivonen> | OK. (I'm inclined to disagree, since the compatibility relies on the author instead of being in the control of the consumer) |
| 09:09 | <hsivonen> | (that is, I'm inclined to disagree that XMDP is a good way to achieve the compatibility) |
| 09:10 | <tantek> | the problem is that such conclusions (in the absolute) rarely tend to be accurate or useful |
| 09:10 | <tantek> | the better way to think of it would be, what would be a *better* way to achieve the compatibility |
| 09:11 | <tantek> | with a solid (written and published on the web) alternative proposal or proposals |
| 09:11 | <hsivonen> | A better way would be to have a defined hFoo to RDF mapping for each hFoo. |
| 09:11 | <tantek> | the point is, with XMDP you don't need to redefine mappings for each microformat |
| 09:12 | <tantek> | with a URI for each term, you're already done |
| 09:12 | <tantek> | since all RDF needs is URIs for terms |
| 09:12 | <tantek> | e.g XFN rel="friend" is defined by http://gmpg.org/xfn/11#friend |
| 09:12 | <tantek> | etc. |
| 09:12 | <hsivonen> | ok, then my concrete proposal would be a single base URI for all microformats.org microformats |
| 09:13 | <hsivonen> | such that consumers can know the URI instead of having to rely on author participation to discover it |
| 09:13 | <tantek> | yes that is a very reasonable request, and in http://microformats.org/wiki/to-do I believe |
| 09:13 | <tantek> | just like W3C DTDs |
| 09:13 | <tantek> | the parallels are there |
| 09:13 | <tantek> | and with that I will bid you good night. i will take the task of writing a longer answer to your original question of "What should microformat consuming software do with XMDP profiles?" |
| 09:14 | <hsivonen> | because if I wanted to process hFoo in an RDF system, I'd want to be able to ingest all hFoo--not just those hFoo instances with RDF-friendly authors |
| 09:14 | <hsivonen> | tantek: thanks |
| 09:14 | <tantek> | thanks for your excellent questions. |
| 09:21 | jgraham | is scared by all the people syaing that proprietry heuristics for filtering out layout tables can be used as a justification for spec features known to break in the face of layout tables |
| 09:25 | <jgraham> | Particularly in the absence of any meaningful data about the mechanisms or accuracy of those heuristics |
| 09:27 | <Philip`> | It does suggest the possibility that maybe HTML5 should define some standard heuristics, and should gather meaningful data on its accuracy, and then could use that as a reason to pay less attention to abuse of attributes on layout tables |
| 09:28 | <hsivonen> | if there were standard heuristics, should validators perform the heuristics, too, and whine about non-empty summaries on what are deemed layout tables? |
| 09:28 | <jgraham> | I suspect the heuristics contain at least a component that depends on the computed style of the table, so it's hard to specify |
| 09:28 | <hsivonen> | ok. |
| 09:29 | <hsivonen> | I don't want HTML conformance to depend on computed style |
| 09:30 | <jgraham> | I wonder (and I hope Philip` or someone will tell me) if there is any meaningful way to use something like a neural network to perform automatic classification of tables into layout/non-layout with high accuracy, given an initial training set |
| 09:30 | <jgraham> | What I don't know is what you would use as inputs. |
| 09:31 | <jgraham> | (an interesting consequence of that is that it would be effectively impossible to specify the lgorithm other than by "use this neural network") |
| 09:31 | <Philip`> | By "neural network" do you mean "magic box"? :-) |
| 09:31 | <jgraham> | *algorithm |
| 09:32 | <jgraham> | Philip`: Well I mean "neural network" al;though that is rather close to "magic box", yes |
| 09:35 | <Philip`> | If you had a lot of training data and a large set of characteristics, I guess you could easily find which characteristics are correlated with data/layout tables, and then I guess any supervised learning algorithm would give you some kind of prediction function |
| 09:35 | <Philip`> | (I've no idea if neural nets would be better/worse than any other approach) |
| 09:36 | <Philip`> | (except that I think you really wouldn't want an algorithm in the spec to be a definition of a neural net) |
| 09:36 | <Philip`> | (just because it'd be clearly insane) |
| 09:36 | <jgraham> | Can the characteristics be boolean things? I had a vauge notion they had to be continuous variables? |
| 09:36 | <jgraham> | Maybe I should just FGI |
| 09:39 | <Philip`> | jgraham: Boolean variables are just continuous variables that only ever happen to be in one of two states :-) |
| 09:39 | <Philip`> | I can't think of any particularly fundamental difference |
| 09:40 | <jgraham> | You might not get convergence or something |
| 09:41 | <hsivonen> | jgraham: when the tryserver Minefield/HTML5 parsing builds crashed for you on Linux, did they work on the second run and only crashed on the first run? |
| 09:42 | <Philip`> | If I remember correctly (aided by Wikipedia), in simple neural nets you have a perceptron which applies a threshold function to a weighted some of its input values, to get a boolean output |
| 09:42 | <Philip`> | and then you connect a load of them together, and do some learning |
| 09:44 | <jgraham> | The wikipedia page on this does not reallyt take universal accessibillity to heart. It it hard to understand :) |
| 09:44 | <Philip`> | Wikipedia describes some condition for convergence which sounds it probably should sometimes be true even if all your inputs are boolean |
| 09:45 | <Philip`> | jgraham: Try http://simple.wikipedia.org/wiki/Neural_network |
| 09:45 | <jgraham> | Philip`: But that is just wrong |
| 09:46 | <Philip`> | jgraham: But it's accessible |
| 09:46 | <jgraham> | The truth? You cna't handle the truth |
| 09:53 | <jgraham> | So it seems like a viable project |
| 09:54 | Philip` | notes that he is not exactly an expert on AI |
| 09:54 | <jgraham> | I guess it would be interesting to do something like a principal component analysis on the result to see what the primary indicators of layout tables are |
| 09:54 | <Philip`> | (nor, indeed, on anything else) |
| 09:55 | <Philip`> | I did sit through two AI lecture courses, but decided pretty early on that I wasn't going to answer any of the exam questions about it :-) |
| 09:57 | <Philip`> | jgraham: Why do anything that complicated, rather than simply computing a table of probabilities like p(table is data table | given has property P) for all Ps you can think of and then looking for the extreme ones? |
| 09:57 | <Philip`> | s/given/table/ |
| 09:58 | <Philip`> | Hmm, maybe the point is to find the independent characteristics, rather than "table has > 4 columns", "table has > 5 columns", "table has > 6 columns", ... all being high-probability cases |
| 10:04 | <jgraham> | Philip`: Yes, I think that is the idea |
| 10:16 | <Philip`> | The hard part would be accurately classifying ten thousand tables to use as training data and for verification |
| 10:16 | <Philip`> | or maybe much more than that would be needed, given how high a proportion will be layout tables |
| 10:30 | <jgraham> | hsivonen: Sorry I missed your question earlier. No, they crashed any time the profile dialog was shown before the browser proper started |
| 10:30 | <hsivonen> | jgraham: ok. thanks |
| 10:30 | <hsivonen> | I'm still clueless when it comes to the reason of the crashes |
| 10:30 | <jgraham> | Philip`: Doesn't seem too hard. It is generally pretty easy to classify the tables and it would parallelise easilly |
| 10:39 | <Philip`> | jgraham: s/ten thousand/a million/ - now it's much harder :-) |
| 10:39 | <jgraham> | Philip`: OK fair enough. But you're just guessing anyay |
| 10:39 | <jgraham> | *anyway |
| 10:40 | <Philip`> | jgraham: Indeed |
| 10:40 | <Philip`> | jgraham: Probably the main concern is that there would have to be enough data tables in the sample, which might mean the sample has to be huge |
| 10:41 | <jgraham> | Is there a command that is the opposite of diff? |
| 10:41 | <slaaya> | is this about rwifi |
| 10:41 | <Philip`> | (It's no good having a heuristic which says "given a table T, it is a layout table" even if it's right 99% of the time, because it's important that it's right for 99% of data tables too) |
| 10:41 | <slaaya> | ? |
| 10:42 | <Philip`> | jgraham: Like something that shows which lines are shared between files? |
| 10:42 | <jgraham> | Philip`: Yes |
| 10:42 | <slaaya> | if i a connect a router to a tv atenna would it work |
| 10:42 | <slaaya> | to boost its signal? |
| 10:42 | <slaaya> | with dd-wrt |
| 10:43 | <jgraham> | slaaya: This is almost certianly the wrong channel. What you suggest seems unlikely to work on general physical grounds through |
| 10:43 | <slaaya> | why though? |
| 10:43 | <slaaya> | if i a solder it the cable that goes to the tv atennna |
| 10:43 | <slaaya> | why would it not work? |
| 10:44 | <jgraham> | Because the design of antennae is wavelength-specific |
| 10:44 | <slaaya> | humm ok |
| 10:44 | <jgraham> | and the power output of the router is fixed |
| 10:44 | <slaaya> | so i cant just use antenna |
| 10:44 | <slaaya> | any |
| 10:44 | <slaaya> | its a vhf atenna |
| 10:44 | <slaaya> | i mean just using the metal as a booster in it |
| 10:45 | <Philip`> | jgraham: diff -C99999 a b | grep '^ ' |
| 10:46 | <jgraham> | Philip`: Ah, clever, thanks |
| 10:52 | <Philip`> | jgraham: I consider it to be a disgusting hack, rather than clever |
| 11:02 | <jgraham> | Well, however you want to put it it seems to do something mildly useful |
| 11:02 | <Lachy> | Hixie, yt? |
| 11:55 | <Philip`> | Is it safe to use xml:base in XHTML? |
| 11:55 | <Philip`> | like, no browser is planning to remove support for it or anything? |
| 11:55 | <zcorpan_> | Philip`: does webkit support it? |
| 11:55 | <Philip`> | zcorpan_: Don't know |
| 11:56 | <zcorpan_> | i think it doesn't |
| 11:56 | <Philip`> | I suppose that's a bit of a pain |
| 12:43 | gsnedders | tries to remember if he actually met danbri at TPAC |
| 12:43 | gsnedders | met too many people to remember who half them are :P |
| 12:45 | <karlcow> | gsnedders: probably. He was at TPAC. Did you talk with him is different. ;) http://flickr.com/photos/bblfish/2290417713/ |
| 12:45 | <gsnedders> | karlcow: I am aware he was there :P |
| 14:28 | gsnedders | finds more and more mistakes with uncertainties in his physics project. yay. |
| 14:33 | <jgraham> | gsnedders: If it is any comfort, getting uncertainties calculations right is really really hard |
| 14:34 | <jgraham> | And then they are usually not htat helpful because they don't tell you about systematic error |
| 14:34 | <gsnedders> | I've already concluded my results are useless because of the uncertainties, so adding more uncertainties on to it makes little difference |
| 14:35 | <karlcow> | :D |
| 14:36 | <gsnedders> | (And my previous physics teacher wondered why I was applying to do _theoretical_ physics.) |
| 14:37 | <jgraham> | There is a plot somewhere of the best-measured value of the speed of light against time. There are large steps, much bigger than the calculated error bars, corresponding to the development of new *techniques* to measure the speed of light |
| 14:38 | <karlcow> | :) calibration inferno |
| 14:38 | <Philip`> | jgraham: How do you know it's not due to the speed of light changing? |
| 14:39 | karlcow | wonders if Philip` is talking about the travel of light between two elements or the speed of light as a constant. |
| 14:40 | <Philip`> | Maybe light is scared of new measurement technologies, and slows down / speeds up a bit |
| 14:40 | <Lachy> | Philip`, because astronomers, astrophysists, or whatever, can look back in time by looking at distant stars and galaxies and make measurements |
| 14:40 | <karlcow> | Lachy: there are more steps doing that |
| 14:40 | <Lachy> | karlcow, of course. |
| 14:40 | <karlcow> | that I was calling calibration inferno |
| 14:40 | <Lachy> | I don't know what that means |
| 14:42 | <karlcow> | first you try to have a black body spectrum in laboratory, then closest bright stars not too hot with a similar temperature or close from your black body spectrum (to avoid interstellar dust), then more distant stars, then close galaxies, etc etc. |
| 14:42 | <karlcow> | each time in the process you had a layer of indirection and errors. |
| 14:43 | <karlcow> | but that's part of the game |
| 14:43 | <karlcow> | which is quite fun |
| 14:43 | gsnedders | maintains his position that astrophysics is crazy |
| 14:43 | <karlcow> | ;) |
| 14:43 | <karlcow> | maybe it's why I like it |
| 14:43 | <gsnedders> | hehe |
| 14:43 | <karlcow> | and I like the Web too ;) |
| 14:43 | <jgraham> | It's no more crazy than ECMAScript |
| 14:43 | <karlcow> | both crazy and imperfect |
| 14:44 | <jgraham> | Actually that probably isn't true |
| 14:45 | <Philip`> | Space is all fake anyway - it's just a painted dome with some lights hanging off it |
| 14:45 | <Philip`> | I've seen The Truman Show, I know how these things work |
| 14:45 | <Lachy> | LOL |
| 14:45 | <karlcow> | Philip`: that! is reasonable :) |
| 14:47 | <Lachy> | there was a nice simple video I saw on YouTube once that explained in laymens terms how we know the speed of light has remained constant. |
| 14:49 | <gsnedders> | hmm… |
| 14:49 | gsnedders | is trying to search his tweets for " and failing |
| 14:55 | <zcorpan> | hsivonen: couldn't rel be case-sensitive in general but various keywords are case insensitive? |
| 14:55 | <hsivonen> | zcorpan: I guess triple generation could be case-sensitive |
| 14:57 | hsivonen | #ifdefs out XLink autolink and SVG onload in text/html |
| 14:57 | <annevk2> | zcorpan, yes, that would be possible |
| 14:57 | <jgraham> | Er, an attribute that is sometimes casesensitive seems like a whole heap of badness waiting to happen |
| 14:57 | <annevk2> | zcorpan, see e.g. xhr.open(method, ...) |
| 14:58 | <hsivonen> | jgraham: possibly, but a smaller heap than CURIEs |
| 14:58 | <annevk2> | it's not verry pretty, no |
| 14:58 | <jgraham> | hsivonen: Yes. But I can't see authors getting it right, ever |
| 14:59 | <hsivonen> | jgraham: can you see authors getting RDFa right? |
| 14:59 | <jgraham> | hsivonen: I think my fellings on that are well known :) |
| 14:59 | <jgraham> | *feelings |
| 14:59 | <annevk2> | nope |
| 14:59 | <jgraham> | annevk2: Was tha directed at hsivonen or me? |
| 15:00 | <annevk2> | I can't see authors getting RDFa right |
| 15:01 | <annevk2> | they already mess up <meta name=keywords> |
| 15:02 | <Philip`> | Authors don't need to know how to get RDFa right - they can just copy-and-paste from examples produced by experts (like with Creative Commons) |
| 15:02 | <annevk2> | deeply technical QA and developers at Opera don't get namespaces right |
| 15:02 | <annevk2> | in fact, we have a bug to this very day for several years and the only persons that have encountered it so far are in this channel |
| 15:02 | <karlcow> | a lot of possible jokes about opera just now ;) I will avoid |
| 15:03 | <annevk2> | hsivonen wasted a day of his life explaining the PFWG how namespaces work |
| 15:03 | <gsnedders> | Peh! Nobody uses Opera! |
| 15:03 | <annevk2> | it's just not gonna fly |
| 15:03 | <Philip`> | karlcow: In this channel, we prefer the jokes to be about namespaces |
| 15:04 | <Philip`> | and then we consider them to be facts rather than jokes :-) |
| 15:04 | <gsnedders> | annevk2: What if we put wings on it, and tell it to not fly to close to the sun or the sea, for fear of wrecking the wings? |
| 15:05 | <karlcow> | get the beat with names paces |
| 15:05 | gsnedders | wonders how many people are confused by the latter half of that |
| 15:05 | <annevk2> | gsnedders, they did, it's called XML |
| 15:05 | <gsnedders> | reversed(list(references.finditer(node.text))) strikes me as kinda stupid |
| 15:06 | <karlcow> | Philip`: it is why it is not very funny ;) people are far too serious in this "real world" channel. |
| 15:06 | <annevk2> | surely the world has heard of Icarus? |
| 15:06 | <jgraham> | gsnedders: Why? |
| 15:07 | <gsnedders> | jgraham: Having to cast it to list. I'd expect reversed to work on iterables, not sequences |
| 15:07 | gsnedders | can see why it only works with sequences though |
| 15:07 | <Philip`> | A modern-day Icarus would likely come undone by flying into power cables, or being intercepted by fighter jets for straying into reserved airspace |
| 15:07 | <jgraham> | gsnedders: def count():n= 1; yield n; n+=1 |
| 15:07 | <karlcow> | annevk2: Icarus, even if he falls, had a dream and it's why no matter the morality of the story, it is the dream which is pushing people :))) |
| 15:08 | <gsnedders> | The moral of Icarus: humans can't fly. |
| 15:08 | <karlcow> | serious people die and are forgotten, poets and dreamers survive. |
| 15:08 | <karlcow> | gsnedders: mwahaha |
| 15:09 | <jgraham> | karlcow: Not sure that is really true. Euler isn't noted for being a good party guest |
| 15:09 | <annevk2> | Einstein on the other hand... oh wait |
| 15:09 | <karlcow> | CQFD = Ce Qu'il Fallait Démontrer. jgraham is a serious person :) |
| 15:10 | <karlcow> | http://z.about.com/d/physics/1/0/C/0/-/-/Einstein_tongue.jpg |
| 15:10 | <gsnedders> | Démontrer? |
| 15:10 | hsivonen | listened to http://www.cybercodeur.net/conferences/parisweb2006/22092006/hq/ogg/08-karl_dubost-enrichissons_nos_relations.ogg |
| 15:10 | <annevk2> | gsnedders, demonstrated afaict |
| 15:11 | <karlcow> | hsivonen: ahaha. I'm talking about women and train and encounters |
| 15:11 | gsnedders | has a reputation of being mad and a girl among his friends, so guesses he falls out of the category of being serious |
| 15:11 | <hsivonen> | it seems that the same microformat & RDFa & profile arguments have gone round and round for years |
| 15:11 | <gsnedders> | annevk2: Ah, that makes sense |
| 15:11 | <karlcow> | I'm also talking about infobesity |
| 15:11 | <zcorpan> | hsivonen: "* hsivonen #ifdefs out XLink autolink and SVG onload in text/html" - for gecko, for security? |
| 15:12 | <hsivonen> | zcorpan: in the Gecko version of the HTML5 parser, mainly for simplicity |
| 15:12 | <jgraham> | karlcow: Certianly I doubt I will be remembered. |
| 15:12 | <annevk2> | XLink autolink? |
| 15:13 | <annevk2> | and SVG onload is that the load event for each fricking element that is inserted? I don't really like that |
| 15:13 | <hsivonen> | annevk2: it seems that XLink has a feature where setting a couple of attributes stop the load and load a new URL in the browsing context |
| 15:13 | <zcorpan> | o_O |
| 15:13 | <hsivonen> | annevk2: SVG onload is a synchronous event firing when </svg> is seen |
| 15:14 | <annevk2> | oh yeah, XLink is pretty advanced |
| 15:14 | <annevk2> | I would strongly encourage you to only support the bits actually needed to make references from SVG and not support any of the additional stuff |
| 15:15 | <annevk2> | e.g. not have any actual XLink support |
| 15:15 | <hsivonen> | annevk2: I care about parity between the HTML and XML tree builders, but personally, I'd be OK with zapping XLink autolinks from XML as well |
| 15:16 | <annevk2> | I don't think they should work in XML |
| 15:16 | annevk2 | was talking about DOM support |
| 15:17 | <hsivonen> | without SVG onload, </script> is the only 'pipeline hazard' that require the concrete tree insertions, the tree builder and the tokenizer to sync up |
| 15:18 | <annevk2> | SVG onload for </svg> would make sense if we allow <svg> as root, but I'm not sure if that's workable |
| 15:18 | <hsivonen> | but it's defined to fire for nested </svg>, too |
| 15:18 | <annevk2> | hmm ok, "SVG SUCKS" |
| 15:19 | <hsivonen> | annevk2: speaking of which, 'daddy' has a white background box in the snapshot of WebKit that NNW uses (obscuring some text) |
| 15:28 | <yael> | hello, I have a question about localStorage. Can I ask it on this channel? |
| 15:28 | <annevk2> | just ask |
| 15:28 | <jgraham> | yael: You can ask it on any channel, but this one has a particularly high chance of giving a useful answer |
| 15:29 | zcorpan | has found a way to inject arbitrary attributes that ie would pick up (not in ie8 mode) when roundtripped through html5lib, but hasn't found a way to make script run |
| 15:29 | <jgraham> | zcorpan: Do you mean through the sanitizer |
| 15:29 | <yael> | What is the impact of setting document.domain on localStorage? Should the browser stop using the localStorage for the current domain and start using the localStorage of the parent? |
| 15:29 | <zcorpan> | yeah... although i haven't actually tested with the sanitizer |
| 15:30 | <zcorpan> | <br title=``foo> |
| 15:30 | <annevk2> | yael, document.domain does not affect localStorage |
| 15:31 | <annevk2> | if it did the spec would say so |
| 15:31 | <annevk2> | document.domain only affects a limited number of APIs |
| 15:32 | <yael> | annevk2: thank you. :-) |
| 15:38 | <zcorpan> | jgraham: <head><meta charset=utf-8><br title=`><xmp>`><script>alert(1)</script></xmp> |
| 15:40 | zcorpan | wonders why '<head>' is serialized |
| 16:04 | <zcorpan> | hmm maybe the spec should require quotes for <, `, and unicode whitespace in attribute values |
| 17:50 | <virtuelv> | hendry: yt? |
| 18:15 | <karlcow> | gsnedders: http://twitter.com/kidehen/statuses/1257796237 |
| 18:37 | <gsnedders> | karlcow: awesome. |
| 20:28 | <webben> | Is it well-known already that the "whatwg Archives" link at http://lists.whatwg.org/listinfo.cgi/whatwg-whatwg.org takes you to a Forbidden page? |
| 20:28 | <Philip`> | webben: It's now http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org , apparently because Dreamhost is stupid |
| 20:40 | <virtuelv> | Hixie: should html5 cover encoding of form data in POST requests? |
| 20:41 | <webben> | Philip`: ta |
| 20:41 | <virtuelv> | or, rather, it does, and I missed it on first glance |
| 20:42 | <webben> | hmm the links on that page are all broken too :( |
| 20:42 | <webben> | hurray at least the search still works |
| 21:09 | <gsnedders> | What are the most common assumptions that people make about others? |
| 21:23 | <Philip`> | gsnedders: That they exist |
| 21:26 | <Dashiva> | CURIEs in @rel are not a problem. We have always been at war with Eurasia. |
| 23:00 | <heycam> | i suppose http://www.w3.org/1999/xhtml/vocab wasn't minted in 1999 |
| 23:01 | <gsnedders> | heycam: Why, of course not! That might be expected! |
| 23:01 | <heycam> | :) |
| 23:02 | <heycam> | and yes i can see why they might want to use that URI, but it does sort of break the expectation of the year being relevant to when it was minted |
| 23:02 | <heycam> | it's a useful expectation since it helps you narrow down the range of years to guess-and-check when you can't remember the correct uri :) |
| 23:04 | <Hixie> | the whole concept of putting dates in permanent urls is ridiculous |
| 23:04 | <Hixie> | almost as ridiculous as the concept of permanent urls as identifiers! |
| 23:04 | Hixie | ducks |
| 23:04 | gsnedders | cows |
| 23:05 | Hixie | duckers |
| 23:07 | <heycam> | i can reliably remember the svg, xhtml and xslt namespace uris, but that's it |
| 23:07 | <heycam> | the xml one i have particular difficulty with... |
| 23:07 | <Hixie> | i'm especially proud of the xbl one |
| 23:08 | <heycam> | what is it, /ns/xbl or something? |
| 23:08 | <Hixie> | yup. so awesome you guessed it without knowing it. |
| 23:08 | <heycam> | i knew it at one stage, but had forgotten |
| 23:08 | <Hixie> | hehe |
| 23:09 | <heycam> | i think you still need special dispensation to use http://www.w3.org/ns/blah namespace uris |
| 23:09 | <Hixie> | you need special dispensation for /yyyy/blah urls too |
| 23:09 | <Hixie> | /yyyy/mm/blah are the only ones you can mint without agreement |
| 23:09 | <Hixie> | at least in theory |
| 23:09 | <Hixie> | last i chekced |
| 23:09 | <heycam> | oh, so even less memorable :) |
| 23:10 | <heycam> | i don't have difficult remembering java package/class names, on the other hand |
| 23:10 | <Hixie> | indeed! |
| 23:10 | <heycam> | import java.1999.awt.Button |
| 23:13 | <heycam> | Hixie, so i see that you noted in some forum recently that you weren't tracking public-html mails and that was up to someone else |
| 23:13 | <heycam> | since i've sent a few things there that haven't been dealt with yet, should i instead have filed bugzilla entries or something? |
| 23:14 | <heycam> | (i think those particular ones are on your issues list, but i was just wondering about procedure) |
| 23:14 | <Hixie> | in theory the chairs are supposed (i am told) to tell me what to respond to |
| 23:14 | <Hixie> | in practice i respond to anything that is a comment on the spec that i don't think is inane |
| 23:14 | <Hixie> | which doesn't actually include that much |
| 23:14 | <Hixie> | but that's another story |
| 23:14 | <Hixie> | what was your feedback? i can see if it's in the folders |
| 23:14 | <heycam> | i did check, and i saw a few things of mine there, so i'm sure it's fine |
| 23:15 | <heycam> | just wanted to know what i *should* have been doing |
| 23:15 | <Hixie> | officially, filing a bug is the way to ensure that i don't miss your feedback |
| 23:16 | <Hixie> | or asking the chairs to track it |
| 23:16 | Hixie | waits for heycam to return |
| 23:16 | <Hixie> | officially, filing a bug is the way to ensure that i don't miss your feedback |
| 23:16 | <Hixie> | or asking the chairs to track it |
| 23:17 | <heycam> | k |
| 23:17 | <heycam> | (wireless is dodgy here) |
| 23:17 | <Hixie> | in practice if you send feedback i'll see it and track it |
| 23:17 | <Hixie> | i do read every e-mail on public-html, the only ones i don't add to my pile are the ones that are either rehashing old ground, making proposals that are prosterous on their face, trolling, ignorant, etc |
| 23:18 | <Hixie> | sometimes something falls through the cracks though |
| 23:18 | <Hixie> | so if you want to make sure, either ask me or file a bug |
| 23:18 | <heycam> | ack |
| 23:20 | <Hixie> | bbiab |
