| 00:28 | <Hixie> | anyone know of an online tool to normalise strings? |
| 00:28 | <TabAtkins> | like, nfc normalize? |
| 00:28 | <Hixie> | yeah |
| 00:29 | <TabAtkins> | No, but charlint is available for perl |
| 00:29 | <Hixie> | anything available for JS? |
| 00:29 | <TabAtkins> | I'm not finding anything. |
| 00:30 | <Hixie> | henri's validator just says "Text run is not in Unicode Normalization Form C" and doesn't tell me what it should be, which seems less than optimally helpful |
| 00:31 | <ttepasse> | Isn't an Unicode-Normalisation part of IDNA? There a JS-Tools for that. |
| 00:40 | <boogz> | has the "lh" tag been removed from HTML5? |
| 00:40 | <boogz> | anybody know? |
| 00:41 | <Hixie> | lh was never in html5 |
| 00:41 | <boogz> | aw crap |
| 00:41 | <Philip`> | Seems it was only ever in HTML 3.0 |
| 00:42 | <Philip`> | "User agents can use this in place of the full list when a mechanism is provided to fold and unfold nested lists." - sounds a bit like <details> |
| 00:45 | <ttepasse> | Hixie, I remembered I had a small js-script in which there is a String.prototype.toNFC() ... but there was a problem with higher plane chars. |
| 00:46 | <ttepasse> | If you're interested, you'll have to reprogramm the script which tranforms the Unicode DB into a JS-Object: |
| 00:46 | <ttepasse> | http://gist.github.com/42793 |
| 00:46 | <Hixie> | cool, thanks |
| 00:47 | <ttepasse> | If I remember correctly the problem is, that JS only allows four hexadecimal chars for Unicode escape codes. |
| 00:51 | <micheil> | I really need to remember to setup that whatwg svn mirror of mine on a cron job. |
| 02:04 | <TabAtkins> | Hixie: In the sentence above the Zalgo example, signed or unsigned integer? |
| 02:05 | <cheeser> | you guys are killing me with the websockets updates. ;) |
| 02:09 | <cheeser> | so all websocket frames are now length delimited? |
| 02:09 | <cheeser> | and i reading that correctly? |
| 02:09 | <cheeser> | s/and/am/ |
| 02:10 | <Hixie> | TabAtkins: fixed |
| 02:10 | <Hixie> | cheeser: yeah |
| 02:10 | <cheeser> | well, that's nicer actually. |
| 02:11 | <cheeser> | that 7-bit encoded length was ... weird. |
| 02:11 | <TabAtkins> | Do bugfixes get you in the acknowledgements? |
| 02:11 | <Hixie> | TabAtkins: you've been in the acknowledgements for a couple of days now |
| 02:11 | <TabAtkins> | Woo! |
| 02:11 | <Hixie> | :-P |
| 02:11 | <cheeser> | so there's no specific frame format for a close frame, yes? just series of 0 bytes in the frame yes? |
| 02:12 | <cheeser> | 0xFF { 8 x 0x00 } 0x00 ? |
| 02:12 | <TabAtkins> | That *is* the specific frame format. |
| 02:12 | <TabAtkins> | No, 9 x 0x00, I think. |
| 02:12 | <Hixie> | cheeser: 0x00 { 8 x 0x00 } |
| 02:12 | <Hixie> | cheeser: it's just an empty frame with frame type 0x00 |
| 02:12 | <cheeser> | right. so basically there's only one frame type now. and the closing frame is defined simply by that content, yes? |
| 02:12 | <Hixie> | cheeser: (note that text frames are now frame type 0xff) |
| 02:12 | <TabAtkins> | No, there are two frame types. |
| 02:12 | <TabAtkins> | 0xFF for test, and 0x00 for "closing frame". |
| 02:12 | <Hixie> | text |
| 02:13 | <TabAtkins> | s/test/text/ |
| 02:13 | <Hixie> | right |
| 02:13 | <cheeser> | ok. let me get caught up on the new doc. the diffs are harder to read than just rereading the doc again. |
| 02:13 | <Hixie> | yeah, the diffs are a pain sometimes, especially with major changes like this |
| 02:13 | <Hixie> | and on a related note, bummer, i actually went _up_ in total pending e-mails today despite emptying my websockets folder! |
| 02:14 | <Hixie> | booo |
| 02:14 | <cheeser> | yeah. but overall it looks like it might be simplifying some of that for me. |
| 02:14 | <cheeser> | i'll probably spend tomorrow updating :) |
| 02:14 | <Hixie> | :-) |
| 02:15 | <Hixie> | i hope we don't have to change it again before v1 |
| 02:15 | <Hixie> | i'm getting a little nervous about annoying all the early adopters |
| 02:15 | <cheeser> | me, too. ;) |
| 02:15 | <cheeser> | i have to present on this in a month. |
| 02:16 | <cheeser> | hrm. i'll probalby outpace the browser support. |
| 03:17 | <boblet> | anyone know of an outliner tool that also shows what content is associated with each title? want to check I understand 4.4.11.4 creating an outline algorithm… |
| 06:00 | <annevk5> | so no variable length after all |
| 06:00 | <annevk5> | meh |
| 06:05 | <annevk5> | Hixie, I might have missed it but it does not seem to say what happens if there is more data than 64-bit can express |
| 06:06 | <annevk5> | Hixie, at least for sending |
| 06:23 | <boblet> | does http://validator.w3.org/ also check HTML5+ARIA? or is that just the beta version + validator.nu? |
| 06:45 | <boblet> | for the record validator.w3.org doesn’t flag ARIA values as errors, so I guess so |
| 07:18 | <Hixie> | annevk5: let me know when a browser supports 16 exabyte strings and i'll fix it |
| 07:19 | <annevk5> | 64-bit is 16 exabyte? |
| 07:20 | <Hixie> | what did you think it was? |
| 07:21 | <annevk5> | I didn't think much about it to be honest |
| 07:23 | <annevk5> | if most websocket messages are going to be short this is quite the waste |
| 07:23 | <Hixie> | not really |
| 07:23 | <annevk5> | and if most is going to be video/audio different framing will likely be used... |
| 07:24 | <Hixie> | IPv4 + TCP is 54 bytes for the headers |
| 07:24 | <Hixie> | and additional 9 is nothing |
| 07:24 | <Hixie> | an |
| 07:24 | <Hixie> | and IPv6 + TCP is 60 bytes |
| 07:25 | <Hixie> | so for a 32 byte message, the 7 bytes extra is 7 / (32 + 9 + 54) = 7% overhead |
| 07:25 | <Hixie> | (with variable width it'd be ~1%) |
| 07:26 | <Hixie> | in practice it makes very little difference unless you think there'll be lots of frames per packet |
| 07:26 | <Hixie> | in which case it's easy to work around, make bigger messages :-) |
| 07:26 | <Hixie> | but if you can convince the wg, i'll back you up and be happy to change it |
| 07:26 | <Hixie> | i just made it fixed length because i couldn't get anyone to agree |
| 07:27 | <annevk5> | i don't feel strongly |
| 07:27 | <annevk5> | just have a slight preference for no limits |
| 07:27 | <annevk5> | at least no theoretical limits |
| 07:28 | <annevk5> | but 16 exabyte... well :) |
| 07:29 | <annevk5> | "As of May 2009, the size of the World's total Digital content has been roughly estimated to be 500 billion gigabytes, or 500 exabytes." |
| 07:29 | <annevk5> | though also |
| 07:29 | <annevk5> | "According to an IDC paper sponsored by EMC Corporation, 161 exabytes of data were created in 2006, "3 million times the amount of information contained in all the books ever written," with the number expected to hit 988 exabytes in 2010." |
| 07:32 | <annevk5> | seems like it's time to go to the office; no charger :/ |
| 07:37 | <abarth> | 16 exabytes ought to be enough for anybody -- ian hickson, 2010 |
| 07:37 | <Hixie> | i didn't say it'd be enough for anybody |
| 07:37 | <Hixie> | quite the contrary |
| 07:38 | <abarth> | :) |
| 07:38 | <Hixie> | however, i do expect that by the time it's a problem (2040?) that we'll have added fragmentation (about 6 months from now?) |
| 07:57 | <Hixie> | ok so seriously, wtf |
| 07:58 | <Hixie> | the wg hums for a fixed width length, i argue for a variable width length and everyone tells me i should just shut up and do what the experts hum |
| 07:58 | <Hixie> | so i change it, and now everyone's like "Well really I would prefer variable-width." |
| 08:18 | <annevk5> | Hixie, fwiw, they did not declare consensus on fixed-length fields |
| 08:20 | <Hixie> | http://www.ietf.org/mail-archive/web/hybi/current/msg02964.html says there was "consensus in the room" |
| 08:20 | <Hixie> | for what that's worth |
| 08:21 | <annevk5> | sorry |
| 08:21 | <Hixie> | but anyway, nobody was arguing for variable length but me, and nobody else seemed to care, and some arguments were made in favour of fixed length that seems somewhat compelling so long as the length was big enough |
| 08:21 | <Hixie> | so... |
| 08:21 | <annevk5> | i hummed against iirc |
| 08:21 | <annevk5> | maybe that's in the minutes |
| 08:22 | <annevk5> | i wonder how much it's worth trying to correct what has been minuted here |
| 08:22 | <Hixie> | easiest solution to avoid getting misminuted: don't attend the meetings |
| 08:23 | <annevk5> | meeting some of the people is useful though |
| 08:23 | <Hixie> | i used to think so too, but i'm not convinced of that anymore either |
| 08:24 | <annevk5> | the technical discussions, mwah, but then the mailing list fulfills that role mostly |
| 08:24 | <annevk5> | I'm still young and naive; I guess I'll find out in a couple of years ;p |
| 08:26 | <Hixie> | imho the only thing that face to face meetings could be good for is increasing the sense of being a team |
| 08:26 | <Hixie> | but if that's the goal, current meetings are an abject failure |
| 08:26 | <Hixie> | we should instead have team-building days |
| 08:26 | <Hixie> | where we play board games or whatever |
| 08:27 | <annevk5> | I've been thinking the same thing to be honest |
| 08:28 | <annevk5> | Something like that would be great |
| 08:28 | <annevk5> | (I was thinking more of physical games, but anything goes really :)) |
| 08:48 | <annevk5> | hybi list blargh |
| 09:27 | <annevk5> | so genius that TextWrangler notices when stuff happens on the file system |
| 09:27 | <annevk5> | on the fly |
| 09:27 | <annevk5> | gedit should learn that |
| 10:12 | <annevk5> | lol indeed |
| 10:12 | <annevk5> | everyone now talks about some kind of variable length with a maximum |
| 10:12 | <annevk5> | dave seems right that we're not nearing consensus anytime soon... |
| 10:15 | <Philip`> | Sounds more like you're continually nearing consensus |
| 10:15 | <Philip`> | until you get too close and then it jumps away |
| 11:13 | <Philip`> | http://crypto.stanford.edu/~dabo/pubs/abstracts/privatebrowsing.html - "We found that private browsing was more popular at adult web sites than at gift shopping sites and news sites [...] This observation suggests that some browser vendors may be mischaracterizing the primary use of the feature when they describe it as a tool for buying surprise gifts" - gosh! Who'd have thought it? |
| 11:52 | <annevk5> | god |
| 11:52 | <annevk5> | content-type is a mess |
| 14:19 | <MikeSmith> | I seem to be having some problems with Gmail in Webkit nightlies |
| 14:20 | <Peter`> | could be the html5 parser, did you try it earlier? |
| 14:21 | <MikeSmith> | yeah |
| 14:21 | <MikeSmith> | this is without the HTML5 treebuilder |
| 14:21 | <MikeSmith> | but it was working fine until maybe 3 days ago or so |
| 14:22 | <MikeSmith> | I guess I'll try again with the latest nightly |
| 14:22 | <jgraham> | Philip`: That's hilarious. I had to read the paper to check that you weren't making that up |
| 14:23 | <jgraham> | Although it is not clear that their methodology really supports their results |
| 14:23 | <jgraham> | s/results/conclusions/ |
| 14:24 | MikeSmith | updates to r64816 |
| 14:24 | <jgraham> | And they lose lots of points for not saying what the error bars on their graph represent |
| 14:31 | <MikeSmith> | Peter`: whoah! MathML enabled now too? |
| 14:31 | <MikeSmith> | excellent |
| 14:31 | <Peter`> | It doesn't work inline yet, but by default it gets compiled in |
| 14:32 | <Peter`> | in nightlies it works as separate files :) |
| 14:34 | <Peter`> | http://trac.webkit.org/export/64845/trunk/LayoutTests/mathml/presentation/fractions-vertical-alignment.xhtml |
| 14:40 | <MikeSmith> | Peter`: http://hsivonen.iki.fi/test/moz/html5-hacks-demo.html appears to mostly work |
| 14:41 | <MikeSmith> | though the entities seem to not work |
| 14:41 | <MikeSmith> | and excessive spacing around the radical sign |
| 14:42 | <Peter`> | I see both the MathML as the SVG, that's cool :) |
| 14:42 | <Peter`> | It's probably doctype related, thank you |
| 14:42 | <annevk5> | entities haven't landed yet |
| 14:42 | <MikeSmith> | ah |
| 14:43 | <MikeSmith> | what the hell kind of nonsense is ⁢ anyway? |
| 14:44 | <Philip`> | Sounds like semantics |
| 14:44 | <Philip`> | or maybe it's a space |
| 14:44 | <Peter`> | MikeSmith: http://www.fileformat.info/info/unicode/char/2062/index.htm |
| 14:44 | <MikeSmith> | thanks |
| 14:45 | <MikeSmith> | still not clear to me me what the purpose of it is, other than spacing |
| 14:45 | <cheeser> | isn't that what is for anyway? |
| 14:45 | <MikeSmith> | yeah, that's what I was about to say |
| 14:45 | <MikeSmith> | I guess it's a semantic no-breaking space |
| 14:46 | <MikeSmith> | or a semantic non-breaking non-space |
| 14:46 | <Peter`> | Citing the spec: The InvisibleTimes MathML character entity is used here to indicate to a renderer that there are special spacing rules between the 4 and the x, and that the 4 and the x should not be broken onto separate lines. |
| 14:46 | <Peter`> | Use of the entity is deprecated now |
| 14:47 | <MikeSmith> | Peter`: what should be used instead? |
| 14:47 | <Peter`> | <mchar name="InvisibleTimes" /> |
| 14:47 | <Peter`> | see the paragraph at http://www.w3.org/TR/2000/WD-MathML2-20000328/chapter2.html#N1720 |
| 14:50 | <Peter`> | Inline MathML should increase its usage a lot |
| 14:50 | <MikeSmith> | I didn't know about mchar |
| 14:51 | <jgraham> | srsly? |
| 14:51 | <jgraham> | <char name="InvisibleTimes"/> |
| 14:51 | <MikeSmith> | I know very little about MathML |
| 14:52 | <jgraham> | s/char/mchar/ |
| 14:52 | <jgraham> | I thought typing ⁢ was bad enough |
| 14:52 | <MikeSmith> | yeah well |
| 14:52 | <MikeSmith> | named entities that require DTDs were a misguided idea from the beginning |
| 14:52 | <jgraham> | MikeSmith: (the srsly was not directed at your ignorance; I shared it) |
| 14:53 | <MikeSmith> | ah, OK |
| 14:53 | <Peter`> | So did I by the way, I read it on that page |
| 14:53 | <MikeSmith> | anyway, editors just need a better way to display non-printing characters |
| 14:53 | <cheeser> | interesting |
| 14:53 | <annevk5> | most do |
| 14:54 | <annevk5> | but perhaps not Unicode invisible characters as they are rather uncommon |
| 14:54 | <jgraham> | Maybe MathML needs to be less finickity about making you put in all the invisible characters |
| 14:54 | <MikeSmith> | non-printing chars are the only good use case I can think of for named character references or named-character-reference equivalents like that mchar thing |
| 14:55 | <jgraham> | <mn>4</mn><mi>x</mi> is long enough |
| 14:55 | <jgraham> | without having to write |
| 14:55 | <jgraham> | <mn>4</mn><mo><mchar name="InviibleTimes"/></mo><mi>x</mi> |
| 14:55 | <Workshiva> | <mchar name="InvisibleTimes"/> |
| 14:56 | <jgraham> | In LaTeX you write |
| 14:56 | <jgraham> | 4x |
| 14:56 | <Workshiva> | <span hidden>x</span> |
| 14:58 | <Philip`> | LaTeX seems to confuse people in cases where they shouldn't have an invisible times between symbols, though |
| 14:59 | <Philip`> | They'll write $stuff = 1$ instead of $\mathit{stuff} = 1$ and it'll render like "stu f f = 1" which is extremely irritating for readers who are needlessly pedantic |
| 15:00 | <jgraham> | true |
| 15:02 | <gsnedders> | jgraham: You're on holiday. Go away. |
| 15:02 | <Philip`> | though I suppose that's less of an issue in many mathematical fields since they'll write $ζ = 1$ instead and it's your own fault if you can't remember how to pronounce ζ or mix it up with ξ |
| 15:03 | <Peter`> | MikeSmith: Not sure if you're interested in geo/mobile stuff, but DeviceMotionEvent ("previously" Accelerometer) just landed in WebKit :) http://trac.webkit.org/changeset/64845 |
| 15:04 | <MikeSmith> | coo |
| 15:04 | <MikeSmith> | cool |
| 15:04 | <MikeSmith> | oh |
| 15:04 | <MikeSmith> | dino |
| 15:04 | <MikeSmith> | I guess I know what he's been working on now |
| 15:07 | <MikeSmith> | the event is actually still called DeviceOrientation in http://dev.w3.org/geo/api/spec-source-orientation.html |
| 15:07 | <MikeSmith> | but I guess that name is a known problem |
| 15:07 | <MikeSmith> | I think Safari already has an event with that name |
| 15:07 | <annevk5> | what is the actual event name? |
| 15:08 | <annevk5> | these sound like event objects |
| 15:08 | <MikeSmith> | DeviceOrientationEvent |
| 15:08 | <annevk5> | that's the event object |
| 15:08 | <annevk5> | well, hopefully |
| 15:08 | <MikeSmith> | http://dev.w3.org/geo/api/spec-source-orientation.html#device_orientation_event |
| 15:09 | <MikeSmith> | http://dev.w3.org/geo/api/spec-source-orientation.html#accelerometer_event |
| 15:10 | <MikeSmith> | ah |
| 15:11 | <MikeSmith> | actually, I just see from Peter` tweet that DeviceMotionEvent seems to be same as AccelerometerEvent |
| 15:14 | <jgraham> | gsnedders: I have no where to go |
| 15:14 | <jgraham> | and Katie is getting her hair cut |
| 15:16 | <Peter`> | MikeSmith: apparently the spec is yet to be updated |
| 15:19 | <Peter`> | Weekend! :) Talk to you later |
| 15:27 | <gsnedders> | jgraham: The park! |
| 15:34 | <farious> | !help |
| 15:35 | <farious> | Hello |
| 15:36 | <erlehmann> | anyone looking at the [embed] in here in IE will notice that IE cannot into HTML5 http://gsoc2010.dieweltistgarnichtso.net/?p=63 |
| 15:36 | <erlehmann> | it outputs tags like <:figure> and </:figure> |
| 15:36 | <erlehmann> | any idea how a could fix that? |
| 15:39 | <erlehmann> | needs more remy sharp |
| 15:40 | <Philip`> | erlehmann: Do you really want to be putting things like <html><head/> inside the body of your page? |
| 15:41 | <erlehmann> | Philip`, NO. thank you. |
| 15:44 | <Workshiva> | erlehmann: Isn't that the same problem the shiv is supposed to fix? |
| 15:44 | <erlehmann> | Workshiva, that's a different problem. |
| 15:45 | <erlehmann> | Workshiva, apparently the shiv creates an unnamed namespace for elements. as you might see using IE7, VIDEO gets serialized correctly. |
| 15:45 | <erlehmann> | but if i don't use remy sharps IE shiv, I cannot style the elements |
| 16:10 | <erlehmann> | Philip`, thank you. I had assumed some things about HTML5lib::parseFragment() without checking them. |
| 16:12 | <cheeser> | can anyone explain to me the "why" of the security header fields in the websocket handshake? |
| 16:12 | <cheeser> | it almost makes sense but i feel like i'm missing something. |
| 17:08 | <cheeser> | Hixie: can you explain to me, if you have the time, what need the sec keys fill in the websocket handshake? the spec doc almost makes sense to me but I feel like i'm missing something. |
| 17:19 | <jgraham> | cheeser: (I guess Hixie is asleep right now) |
| 17:21 | <cheeser> | could be ;) |
| 17:21 | <cheeser> | it's not urgent anyway |
| 19:09 | <Hixie> | cheeser: mostly they help convince the browser that the server is really a websocket server and not (e.g.) an HTTP server being tricked |
| 19:28 | <cheeser> | ok. but it's necessarily meant to prevent man in the middle attacks, yes? |
| 19:28 | <cheeser> | just to make sure the other side really means to handle the websocket request, right? |
| 19:33 | <Hixie> | cheeser: the only prevention of man-in-the-middle attacks is TLS |
| 19:39 | <cheeser> | right. |
| 19:39 | AryehGregor | wants TLS+SRP :( |
| 19:39 | <cheeser> | that's why it wasn't making much sense to me. :) |
| 19:40 | <cheeser> | thanks, Hixie |
| 20:51 | <AryehGregor> | http://www.whatwg.org/specs/web-apps/current-work/multipage/the-button-element.html#dom-button-value |
| 20:52 | <AryehGregor> | "value" isn't enumerated, how can it be limited to only known values? |
| 20:54 | <Hixie> | oops will fix |
| 20:55 | <AryehGregor> | Is there any particular reason why other enumerated attributes aren't limited to known values, like preload on video/audio? annevk suggested that all enumerated attributes should be limited to only known values. |
| 20:56 | <AryehGregor> | Seems like it would be more consistent, and he didn't think it would break much. |
| 20:56 | <Hixie> | i'm happy to change them on a case-by-case basis (especially the new ones will require very little to convince me) |
| 20:56 | <Hixie> | send mail or file bugs |
| 20:56 | <AryehGregor> | I'll send mail. |
| 20:58 | <Hixie> | but name the ones you want changed, don't ask for a default change :-) |
| 20:58 | <Hixie> | (if i were to change the default i'd also change all the existing enumerated attributes so it'd have no effect on the spec :-) ) |
| 21:00 | <AryehGregor> | You could just say that an IDL attribute that reflects an enumerated DOM attribute always has the limit-to-known-values behavior, so you can drop the boilerplate "limited to only known values". |
| 21:05 | <Hixie> | AryehGregor: yeah but then i'd have to have a "not limited to known values", and currently there are more like that than that are limited |
| 21:05 | <AryehGregor> | Not if you just change all enumerated attributes to be limited to known values. annevk seemed to think that was a good idea. |
| 21:06 | <Hixie> | yeah what i'm saying is that i am not willing to do that without considering each one on a case by case basis, because it's likely that there are compat issues there. |
| 21:07 | <AryehGregor> | Presumably unless the implementers are willing to go with it. |
| 21:07 | <Hixie> | food time |
| 21:34 | <AryehGregor> | Hixie, should I report the "value limited to only known values" thing as a bug? |
| 21:36 | <TabAtkins> | Argh, is there a listing of all the spec views somewhere? I want to read the websrt subset. |
| 21:38 | <Philip`> | TabAtkins: http://www.whatwg.org/specs/web-apps/current-work/websrt.html |
| 21:39 | <TabAtkins> | danke, Philip`. |
| 22:35 | <Hixie> | AryehGregor: nah, i fixed it in the source already, it'll be fixed soon |
| 22:35 | <AryehGregor> | Hixie, k. |
| 22:35 | <AryehGregor> | Ṯ̤̩̙̖̣ͬ̈̅͊h̏ͭ͌ͦͨ͒̃i̱̹̙͔̟̞̙̔̑ͦ̊s̬̣͙ͭ̾̆ ͉͕͉̲͊͛ͧi̳̩̳͋ͥ͒͋͗ś̜͐̆ͧͥ̈̋ ͚̻̖̦̽͌̂̓̎̍w̥ͤeͣ̑̈ͮï̖̪̺̈̄ͧͦr̺ͣd̪̣̜̲̪̞̈́̉ͫ̔͐̚.̦̳͖̰̘ͧͫͥ |
| 22:35 | <AryehGregor> | (looks really crazy in my fixed-width font) |
| 22:37 | <TabAtkins> | Yeah, looks retarded in irssi, since it doesn't do composition. |
| 22:37 | <AryehGregor> | That's just broken. |
| 22:37 | <TabAtkins> | Indeed. |
| 22:40 | <AryehGregor> | I just realized that's the technique that this uses. http://stackoverflow.com/questions/1732348/regex-match-open-tags-except-xhtml-self-contained-tags/1732454#1732454 |
| 22:41 | <TabAtkins> | Yup. |
| 22:42 | <AryehGregor> | Oh, that probably originated the meme. |
| 22:43 | <AryehGregor> | Maybe? |
| 22:43 | <AryehGregor> | No, apparently not. |
| 22:43 | <AryehGregor> | Dates to 2004, from SomethingAwful. |
| 22:44 | <TabAtkins> | Everything ever originates from SA or one of the chan. |
| 22:45 | AryehGregor | tried Googling some Zalgo text and got a 414, URL too large |
| 22:46 | <Hixie> | so someone is arguing that rel=prefetch should be allowed on <a href=""> |
| 22:46 | <Hixie> | this seems reasonable, since you don't always know what needs prefetching until you get down the page |
| 22:47 | <Peter`> | I remember at least two threads about it on whatwg |
| 22:47 | <Hixie> | e.g. google search might want to say the first hit should be prefetched, but doesn't know the first hit at the time of writing the <head> </head> bit |
| 22:47 | <Peter`> | that was one of the rationales given indeed |
| 22:47 | <gsnedders> | Yet that opens up getting really large amounts of resources |
| 22:47 | <Hixie> | so here's the question: |
| 22:47 | <Hixie> | clearly rel=prefetch on <a href="">...</a> doesn't actually change what kind of link relationship it is |
| 22:47 | <Hixie> | so do we define that some rel="" types don't define relationships? |
| 22:48 | <Hixie> | if so, do we require that <link rel=prefetch> cases also list another keyword? |
| 22:48 | <Peter`> | alternative would be @prefetch, like there's @ping too, although it would be inconsistent with <link> |
| 22:48 | <TabAtkins> | You mean "How do we abuse the rel mechanism to do things other than express relationship semantics"? |
| 22:48 | <Hixie> | well it has to be rel=prefetch because that's what people do |
| 22:48 | <Hixie> | TabAtkins: yes |
| 22:49 | <TabAtkins> | Is it illegal to do <a rel="" href=foo>bar</a>? (That is, include @rel with an empty list.) |
| 22:49 | <AryehGregor> | Why would we require that <link rel=prefetch> also list another keyword? |
| 22:49 | <jgraham> | Hixie: We realise that our notions are at odds with the real word and so quietly kill them in a corner? |
| 22:49 | Hixie | checks |
| 22:49 | <Hixie> | rel="" (empty) is valid |
| 22:50 | <TabAtkins> | Then, no, rel=prefetch should be fine by itself. |
| 22:50 | <Hixie> | jgraham: i'm trying to kill them, i'm just trying to work out what to put in its place |
| 22:50 | <Hixie> | TabAtkins: i mean on <link> |
| 22:50 | <Hixie> | i guess <link rel=""> (empty) is also valid |
| 22:50 | <boblet> | I also like @prefetch, but @prefetch and rel="prefetch" would be bad |
| 22:50 | <TabAtkins> | I don't see anything particularly wrong with a <link> that doesn't express a valid relationship. |
| 22:50 | <Hixie> | though maybe it shouldn't be |
| 22:51 | <TabAtkins> | I just consider <link> to be the url-carrying version of <meta>. |
| 22:51 | <AryehGregor> | It seems highly pointless. |
| 22:51 | <Hixie> | just so we're clear, i'm not suggesting we do anything but rel=prefetch, i'm just trying to work out what we should redefine rel="" to be so that that makes sense |
| 22:51 | <jgraham> | Hixie: I think what I'm trying to say is that in-practice rel is used to trigger behaviour in UAs |
| 22:51 | <Hixie> | AryehGregor: well, it means the same as <link rel="bogus" href="...">, assuming bogus isn't recognised |
| 22:51 | <AryehGregor> | Just say it defines various properties of URLs, mostly relations. |
| 22:52 | <gsnedders> | jgraham: But saying it badly, like normal? :P |
| 22:52 | <Hixie> | jgraham: not always |
| 22:52 | <AryehGregor> | Which means nothing, though? |
| 22:52 | <jgraham> | Hixie: Counter examples? |
| 22:52 | <Hixie> | jgraham: rel=index |
| 22:52 | <Hixie> | jgraham: it just defines a hyperlink |
| 22:53 | <Hixie> | jgraham: the spec currently separates "external resource links" from "hyperlinks" |
| 22:53 | <TabAtkins> | AryehGregor: "rel provides additional information about the link it is associated with. Often, this expresses a particular relationship between the page and the linked resource, but it also sometimes expresses that the UA should treat the link specially." |
| 22:53 | <jgraham> | Hixie: Presumably triggers behaviour in some UAs e.g. search engines, browsers with specific support for that |
| 22:53 | <AryehGregor> | "presumably" |
| 22:53 | <AryehGregor> | Meaning, "I don't see why it shouldn't, so I'll assume it does." |
| 22:53 | <jgraham> | Well if it doesn't why are people using it? |
| 22:54 | <AryehGregor> | Why are people doing <script language="javascript">? |
| 22:54 | <jgraham> | They must be getting something out of it |
| 22:54 | <boblet> | that’s the way they learned to do it |
| 22:54 | <jgraham> | Maybe just a misguided sense of their own cleverness |
| 22:54 | <AryehGregor> | No, they're just cargo-culting or assuming someone uses it somehow. |
| 22:54 | <AryehGregor> | When maybe no one does. |
| 22:54 | <boblet> | yep |
| 22:54 | <jgraham> | Well that doesn't sound like behaviour we should encourage |
| 22:55 | <AryehGregor> | Obviously not. |
| 22:55 | <boblet> | and the spec doesn’t |
| 22:55 | <Hixie> | jgraham: well insofar as any description of semantics causes behaviour, sure |
| 22:55 | <boblet> | well, in that case (and <style>) anyhow |
| 22:55 | <Hixie> | jgraham: i thought you meant specifically defined behaviour (like rel=stylesheet) as opposed to UA-chosen behaviour based on semantics (like rel=index) |
| 22:55 | <jgraham> | Great so define rel as having a funcional effect and then maybe people who have no idea what the functional effect will be will think twice |
| 22:56 | <jgraham> | Hixie: Ah |
| 22:57 | <TabAtkins> | The unifier for @rel is that it expresses metadata about the linked resource. |
| 22:58 | <TabAtkins> | (And maybe the host resource.) |
| 22:58 | jgraham | wonders what sort of crappy webmail client tells you that your session is "invalid or expired" half way through you writing a message |
| 22:58 | <jgraham> | TabAtkins: How is "prefetch" metadata about anything? |
| 22:59 | <jgraham> | It is purely a hint to enable special UA processing |
| 22:59 | <Hixie> | yeah |
| 22:59 | <Hixie> | that's my problem |
| 23:00 | <TabAtkins> | jgraham: It's saying "this resource needs to be prefetched". That's metadata. |
| 23:00 | <jgraham> | You could in some sense say it is metadata about the link itself |
| 23:00 | <gsnedders> | jgraham: Horde? :P |
| 23:00 | <jgraham> | gsnedders: roundcube |
| 23:00 | <jgraham> | but not metadata about the resource |
| 23:01 | <Hixie> | TabAtkins: it's not metadata about the link or the resource, it's a pragma for the user agent regarding (if anything) the _current_ resource |
| 23:02 | <jgraham> | Anyway if you want a single unifying principle rather than just saying that it can do various different things, rel provides additional information to the UA to enable special processing of the link |
| 23:03 | <boblet> | metadata about the author’s desire for how the UA should treat the link |
| 23:04 | <TabAtkins> | Hm, yeah, saying it's metadata about the link itself would work. |
| 23:04 | <TabAtkins> | And be applicable to ordinary relations as well. |
| 23:05 | <Hixie> | it's not really even that imho |
| 23:05 | <TabAtkins> | Taking it from "this is an unnamed and meaningless link between two resources" to "this link shows the author of the current page". |
| 23:05 | <Hixie> | what would you replace "shows the author" with for rel=prefetch? |
| 23:07 | <TabAtkins> | "this is an unnamed and meaningless link" => "this link denotes a resource that should be grabbed now"? |
| 23:08 | TabAtkins | retires to his windows machine, where he will set things up so that he can easily refresh all four browsers at once. |
| 23:08 | <TabAtkins> | Fucking. Margin. Collapsing. |
| 23:08 | <cardona507> | windows machine... shudder |
| 23:09 | <AryehGregor> | Why only four browsers? |
| 23:09 | <Hixie> | TabAtkins: that doesn't sound like metadata, it sounds like a pragma :-) |
| 23:09 | <TabAtkins> | becasue chrome is an adequate webkit representative in things like this. |
| 23:09 | <AryehGregor> | True. |
| 23:09 | <TabAtkins> | Hixie: potato, potragma. |
| 23:10 | <cardona507> | only 4 browser? what about Sleipnir? :) |
| 23:10 | <cardona507> | *browsers |
| 23:11 | <boblet> | fwiw most authors would not worry about prefetch being described as a relationship imo |
| 23:11 | <boblet> | not that that helps much |
| 23:11 | <Hixie> | oh this is a completely theoretical issue |
| 23:11 | <Hixie> | i just want to make sure the spec is internally consistent |
| 23:13 | <Hixie> | ok i have "external resource links" (rel=stylesheet), "hyperlink links" (rel=index), what should I call the third kind of link that is rel=prefetch? |
| 23:14 | <TabAtkins> | pragma links |
| 23:15 | <Hixie> | trying to avoid the word pragma, since we used it with <meta> |
| 23:16 | <TabAtkins> | Why avoid it? It describes it accurately. It's just a pragma with a url argument, rather than a string. |
| 23:16 | <Hixie> | using the same term for two things in the same spec causes all kinds of hassle |
| 23:16 | <boblet> | links to suggest (desired) UA behaviour |
| 23:16 | <Hixie> | like searches finding both terms, people getting them confused, search-and-replace messing it up, etc |
| 23:16 | <Hixie> | how about "directive links", since pragma is another word for directive |
| 23:17 | <TabAtkins> | ... |
| 23:17 | <TabAtkins> | I was literally typing that just now. |
| 23:17 | <Hixie> | heh |
| 23:17 | <boblet> | much better than my attempt ;( |
| 23:17 | <AryehGregor> | Okay, this is somewhat weird. |
| 23:17 | <Hixie> | :-) |
| 23:17 | boblet | won’t give up his day … occupation |
| 23:17 | <AryehGregor> | Hmm. |
| 23:18 | <gsnedders> | TabAtkins: I believe that's called doing a Philip` here, seeming he always types the intelligent comment more quickly than someone else |
| 23:19 | <Workshiva> | gsnedders: It would be more like pulling a me, since I'm always not the first person to type the intelligent comment |
| 23:21 | <AryehGregor> | What is this expected to alert? data:text/html,<!doctype html><script>var el = document.createElement("form"); el.setAttribute("action", ""); alert(el.action);</script> |
| 23:22 | <AryehGregor> | Because I've tested in four browsers and they all get it wrong. |
| 23:22 | <AryehGregor> | (as far as I'm reading the spec) |
| 23:22 | <gsnedders> | Per spec? I think "" |
| 23:23 | <gsnedders> | Yeah, that seems right |
| 23:23 | <AryehGregor> | Shouldn't it be the resulting absolute URL? |
| 23:23 | <AryehGregor> | http://www.whatwg.org/specs/web-apps/current-work/multipage/urls.html#reflecting-content-attributes-in-idl-attributes |
| 23:23 | <gsnedders> | Oh, wait, yeah |
| 23:23 | <gsnedders> | Duh |
| 23:23 | <AryehGregor> | "If a reflecting IDL attribute is a DOMString attribute whose content attribute is defined to contain a URL, then on getting, the IDL attribute must resolve the value of the content attribute relative to the element and return the resulting absolute URL if that was successful, or the empty string otherwise" |
| 23:23 | <gsnedders> | I'm misremembering what @action is |
| 23:23 | <gsnedders> | (Like, confusing it with @method) |
| 23:24 | <AryehGregor> | Chrome dev, Safari 5, Firefox 4, and Opera 10.60 all alert "". |
| 23:24 | <AryehGregor> | Confusingly, the same bug exists for some random set of URL attributes in all of them. |
| 23:24 | <AryehGregor> | In Chrome dev it's audio/src, base/href, blockquote/cite, embed/src, form/action, iframe/src, ins/cite, source/src, video/src, video/poster. |
| 23:24 | <AryehGregor> | But a/href works as expected. And img/src, etc. |
| 23:24 | <AryehGregor> | Opera has a different set where it's wrong. |
| 23:25 | <AryehGregor> | Safari 5 has a still different set. |
| 23:25 | <AryehGregor> | Firefox 4 only has that one, all others behave correctly. |
| 23:25 | <AryehGregor> | So I'm very confused. |
| 23:25 | <AryehGregor> | (The test I'm using for this is: http://aryeh.name/tmp/html5tests.html) |
| 23:25 | <AryehGregor> | (Not totally sure if all the failures are legit, I'm working through them and filing bugs when I feel like it) |
| 23:27 | <gsnedders> | AryehGregor: Feel free to report a bug on Opera and just attach that |
| 23:27 | <gsnedders> | AryehGregor: (Well, you can link to it, but link-rot makes it preferable to have it attached…) |
| 23:27 | <TabAtkins> | Dammit, there just doesn't appear to be any way to get an empty element to show its position in Opera. |
| 23:28 | <TabAtkins> | It refuses to show an outline, won't show a box-shadow with a positive spread, and implements the old version of border-image which can't paint outside the box geometry. |
| 23:28 | <TabAtkins> | And the inspector wont' give me lines for its position either. |
| 23:29 | <AryehGregor> | gsnedders, I filed some bugs against Opera before and never got any response, so now I don't bother. You can use it if you want. I hope to polish it up and post it to the whatwg and/or submit it as a test to the W3C before the end of the summer, though. |
| 23:29 | <gsnedders> | AryehGregor: Please do report it, though |
| 23:30 | <AryehGregor> | You can report it for me if you want. I'm tired of reporting bugs and then never seeing anything come from them. |
| 23:30 | <AryehGregor> | Even IE doesn't do that. |
| 23:34 | <MikeSmith> | hmm, http://diveintohtml5.com/ is showing a Chinese-localized version when I try it |
| 23:35 | <Hixie> | me too |
| 23:36 | <MikeSmith> | maybe the .com site has nothing to do with Mark |
| 23:36 | <Hixie> | seems likely |
| 23:37 | <TabAtkins> | Yay, I have focus-follows-mouse behavior in w7 now! |
| 23:37 | <AryehGregor> | Yay! |
| 23:37 | <TabAtkins> | woo registry hacking! |
| 23:38 | AryehGregor | would care if he used Windows 7 other than for testing IE9 |
| 23:38 | <gsnedders> | TabAtkins: ewww |
| 23:39 | <TabAtkins> | opera didn't immediately install on my linux box, and i don't have the patience to debug it when it works on windows where i have to do my cross-browser testing anyway, so shrug. |
| 23:39 | <TabAtkins> | gsnedders: You're wrong. |
| 23:39 | <TabAtkins> | Note: *not* auto-raise. |
| 23:39 | <AryehGregor> | Focus-follows-mouse is unquestionably awesome. |
| 23:39 | <AryehGregor> | Anyone who thinks otherwise is crazy. |
| 23:39 | <AryehGregor> | But anyway, I have to go now. |
| 23:39 | <AryehGregor> | (by the way, I moved my test to a permanent home at http://aryeh.name/tests/reflection.html, you can treat that as a permalink) |
| 23:42 | <gsnedders> | hsivonen: http://trac.webkit.org/browser/trunk/WebCore/benchmarks/parser/html-parser.html appears to be benchmark for HTML5 parsing. Dunno if you saw that when you were asking before… |
| 23:56 | <Hixie> | should <link rel=prefetch media=audio> not preferch if the media is screen? |
| 23:57 | <Hixie> | and if so, should the same apply to <a rel=prefetch media=audio>? |
| 23:57 | <Hixie> | if so, this is gonna require far more changes |
| 23:58 | <TabAtkins> | <link>, sure, <a>, no. |
| 23:58 | <TabAtkins> | (I don't think there's any precedent for @media doing special things on <a>.) |
| 23:59 | <TabAtkins> | Rather than do <a rel=prefetch>, why not just <link rel=prefetch> in content? Parsers'll have to handle that anyway for Microdata. |
| 23:59 | <TabAtkins> | You'll have to duplicate links, but shrug. |