| 05:27 | <zcorpan> | Ms2ger: jgraham: done |
| 05:33 | <zcorpan> | TabAtkins: i thought implementations didn't store shorthands anywhere |
| 05:34 | <zcorpan> | TabAtkins: i'm not sure storing the shorthands helps much given that the values can be changed later and then you need to decide how to serialize anyway |
| 05:35 | <zcorpan> | TabAtkins: but yeah, if syntax eats the duplicates, and maybe i can point to css-cascade for the expansion of shorthands (since it has a paragraph about that already), maybe that works |
| 06:12 | <SimonSapin> | TabAtkins, zcorpan: Re Syntax eats the duplicate, it’s actually not that simple when you have !important |
| 06:13 | <zcorpan> | SimonSapin: oh right |
| 06:13 | <zcorpan> | so cascade is still the right place to say which decl should go in |
| 06:14 | <SimonSapin> | maybe |
| 06:15 | <SimonSapin> | zcorpan: well, that would be easiest: within valid declarations for the same longhand property (after shorhand expansion) in the same style rule, with one with the greatest Cascade precedence is kept |
| 06:16 | <SimonSapin> | that is, !important then source order |
| 06:28 | <SimonSapin> | zcorpan: I just wrote http://lists.w3.org/Archives/Public/www-style/2013Aug/0137.html |
| 06:39 | <zcorpan> | SimonSapin: thanks |
| 06:40 | <SimonSapin> | zcorpan: I’ll update Syntax one I figure out if Tab’s new processor does cross-spec linking to sections |
| 06:41 | <zcorpan> | why do you need sections? |
| 07:33 | zcorpan | ponders whether he should go to tpac |
| 07:51 | <SimonSapin> | zcorpan: linking to http://dev.w3.org/csswg/css-cascade/#cascading |
| 07:52 | <SimonSapin> | or linking to 'cascade' might work |
| 07:52 | <zcorpan> | i was going to say... :-) |
| 08:16 | <zcorpan> | were there other use cases for defaultStyle other than toggling display:none? |
| 09:11 | <jgraham> | zcorpan: Yay! |
| 09:12 | zcorpan | computes context, guesses the review |
| 09:13 | <jgraham> | Indeed |
| 09:13 | <jgraham> | Although also, you should go to TPAC |
| 09:14 | <Ms2ger> | Why? |
| 09:15 | <jgraham> | Ms2ger: Because it will increase the number of useful people there, making it more worth my while going :) |
| 09:15 | <Ms2ger> | You could not go yourself ;) |
| 09:16 | <jgraham> | So, I plan to sqaush that review before committing it because really there is no need for dozens of little bugfixes |
| 09:16 | <jgraham> | Anyone object? |
| 09:17 | <jgraham> | I guess that it means that Aryeh will be listed as the sole author and get all the credit/blame |
| 09:17 | <Ms2ger> | Sure |
| 09:39 | <jgraham> | OK, pull request #1 is no more |
| 09:39 | <jgraham> | Unles I screwed something up |
| 09:39 | <jgraham> | Which is possible |
| 10:18 | <zcorpan> | hmm, i need to remember to test both quirks and non-quirks before drawing conclusions |
| 10:19 | <zcorpan> | (i specced clientWidth as depending on whether it's top-level browsing context or not, but actually it depends on quirks mode) |
| 10:47 | <SimonSapin> | annevk: data:text/html,<style>body:before{content:"\d834\dd1e |
| 11:00 | <zcorpan> | TabAtkins: i understand that your preprocessor is optimized for css specs. how well would it work for dom specs? |
| 11:00 | <gsnedders> | jgraham: I was considering Carakan until shipping. |
| 11:01 | <jgraham> | gsnedders: sof worked on it before shipping |
| 11:01 | <gsnedders> | jgraham: Then so did Lachy, so you're still wrong. |
| 11:01 | <jgraham> | I think sof was on the project for longer |
| 11:02 | <jgraham> | Or he fixed even more bugs/unit time than I remember |
| 11:02 | <gsnedders> | (Disconcerting: I still know the Carakan bug number off by heart.) |
| 11:03 | <gsnedders> | jgraham: If you believe BTS, 2009-10-19 was Lachy, 2010-02-01 was sof. |
| 11:03 | <zcorpan> | jgraham: btw if there's anything else in critic that needs my attention, just ping me, i haven't caught up with that at all |
| 11:03 | <jgraham> | gsnedders: Oh |
| 11:03 | <gsnedders> | jgraham: I remembered them both being late Jan. |
| 11:04 | <jgraham> | zcorpan: I can't think of anything, but sure, will do |
| 11:04 | <zcorpan> | if there isn't, maybe i can mark all as read :-) |
| 11:04 | <gsnedders> | jgraham: I think Lachy really started slightly later, and sof slightly earlier. But still definitely Lachy first. |
| 11:04 | <jgraham> | gsnedders: I guess this is just more evidence that sof is an awesome bug killing machine |
| 11:04 | <gsnedders> | jgraham: Yes. Except when he introduces code-deleting bugs. |
| 11:05 | <jgraham> | s/awesome/awe-inspiring/, perhaps |
| 11:05 | <Lachy> | yeah, I didn't really start actively working on carakan for a while after that. I'm not sure when, and I can't remember what I was doing. |
| 11:05 | <gsnedders> | Lachy: Reducing sites and JS libraries. This was… about all we all did. |
| 11:06 | <Lachy> | yeah, I remember what I did when I really worked on carakan, but I can't remember what I was doing in the last few months of 2009 before I actually started work on carakan. |
| 11:06 | <zcorpan> | iirc i was officially carakan qa for a few days or so, but didn't actually do anything (or rather i was doing something else) |
| 11:06 | <jgraham> | (kilsmo also worked on it a bit earlier, mostly doing QAish bits. So the average number of people on the project was probably around 6) |
| 11:06 | <gsnedders> | jgraham: I thought kilsmo stopped more or less when I started, thus I averaged it out |
| 11:07 | <gsnedders> | (Per BTS he was working on it /far/ longer than is true.) |
| 11:07 | <jgraham> | Yeah, that could be true |
| 11:07 | <jgraham> | So maybe 5.5 |
| 11:07 | <zcorpan> | i was supposed to write a testing framework, but kilsmo beat me to it (dunno if his version got used later or not) |
| 11:07 | <gsnedders> | zcorpan: Yes, it did. |
| 11:09 | <hsivonen> | I go outside network reach and the HTML parsing algorithm changes. |
| 11:10 | <gsnedders> | https://github.com/nolanw/HTMLReader — Obj-C + Cocoa HTML parser |
| 11:12 | <hsivonen> | hmm. "in foreign" rules in the spec don't have an entry for the end-of-file token... |
| 11:13 | <jgraham> | hsivonen: Well probably if you were in network range you would have objected to the change ;) |
| 11:15 | <hsivonen> | annevk: how do I ask http://html5.org/tools/web-apps-tracker to show a larger number of recent changes? |
| 11:15 | <zcorpan> | hsivonen: ?limit=1000 |
| 11:15 | <annevk> | hsivonen: ?limit=1000 or if you feel like having fun, -1 |
| 11:15 | <hsivonen> | zcorpan, annevk: thanks |
| 11:17 | <smaug____> | argh, I had forgotten to log out from gmail |
| 11:18 | <smaug____> | I should just delete the account |
| 11:20 | <hsivonen> | Am I failing to see something obvious? Where is the behavior of EOF in foreign content specced? |
| 11:20 | <Ms2ger> | Writing testing frameworks seems like something Opera likes a lot |
| 11:21 | Ms2ger | wonders why |
| 11:21 | gsnedders | remembers jgraham going home for the weekend and being like, "I'll write a test framework this weekend", and thus testharness.js was born. |
| 11:23 | <hsivonen> | there must be someone here who has already looked at EOF in foreign content |
| 11:24 | <Ms2ger> | Not sure what makes you think that |
| 11:28 | <zcorpan> | Ms2ger: maybe because we write tests a lot |
| 11:28 | <Ms2ger> | I guess people who write tests are more inclined to write test harnesses |
| 11:28 | <Ms2ger> | ^full time |
| 11:29 | <hsivonen> | hah. the parsing algorithm now has inserting "in the appropriate place" as a defined concept |
| 11:29 | <hsivonen> | "do the appropriate thing" makes a lot of sense for a spec |
| 11:29 | <jgraham> | Well carakan needed a test harness because it had to run tests in the shell rather than in the browser |
| 11:30 | <gsnedders> | estest-futhark, jsunit, lots of custom one-offs, and testharness.js, I think basically accounts for all the test suites, ignoring opjsunit. |
| 11:30 | <jgraham> | Opera needed testharness.js because the thing we had before that dated to Hixie and had several deficiencies |
| 11:31 | <jgraham> | Like it discouraged writing tests |
| 11:32 | <Ms2ger> | jsframework.js? |
| 11:32 | <zcorpan> | that was jgraham's first attempt |
| 11:33 | <jgraham> | Oh, that was an earlier attempt to fix things |
| 11:33 | <Ms2ger> | I'm glad you came up with a better assert_throws for th.js :) |
| 11:37 | <jgraham> | I wish I had come up with a better way of composing assertions |
| 11:37 | <jgraham> | Like assert_false = invert(assert_true) |
| 11:39 | <gsnedders> | Did one of us not write something that did that? |
| 11:41 | <zcorpan> | jgraham: what would invert do? my guess that it just flips the condition seems wrong for a strict "false" check |
| 11:41 | <zcorpan> | e.g. 0 should fail both assert_true and assert_false |
| 11:42 | <gsnedders> | Equally an exception should fail both. |
| 11:45 | <jgraham> | Yeah, I guess it's not trivial |
| 11:45 | <gsnedders> | I remember hitting this problem before… |
| 11:45 | <jgraham> | But having to manually write an inverse for each assert is annoying |
| 11:45 | <gsnedders> | Maybe it was opjsunit I tried to do this with? |
| 11:46 | <jgraham> | gsnedders: Well it is easy to do not quite right |
| 11:46 | <gsnedders> | jgraham: Right. Which is why I suspected we tried. |
| 11:46 | <jgraham> | Maybe it isn't possible to do it right |
| 11:47 | <gsnedders> | I think you just need the actual assertion composed of smaller things. |
| 11:57 | <jgraham> | Have a server than can serve text files at least ;) |
| 12:01 | <Igor^> | we browsers use utf8 decode characters and they display character references literally? |
| 12:07 | <zcorpan> | Igor^: what? |
| 12:07 | <Igor^> | unicode code point is saved on the hard disk using utf8 utf16 or ucs2 encodings for example? |
| 12:07 | <Igor^> | I am not sure if I use character reference in html file how it is saved on hard disk and how it is decoded with web browser then? |
| 12:37 | <hsivonen> | so we aren't supporting frames in templates anymore? |
| 12:38 | <hsivonen> | but now we are supporting <title>? |
| 12:38 | <hsivonen> | is there any logic to this? |
| 12:39 | <hsivonen> | <title> is conforming but <frame> isn't? |
| 12:42 | <gsnedders> | What has Hixie been smoking? |
| 12:47 | <Igor^> | can you answer me one question? |
| 12:47 | <Igor^> | If I use character reference in html file to represent a character and web server sends the file on browser request, how the browser will decode the character reference? |
| 12:47 | <Igor^> | My Nginx web server is configured to not send character encoding in the header I have set character encoding in the meta tag on page level to utf8. |
| 12:49 | <Igor^> | and what is this? http://html6spec.com/ :D |
| 13:03 | <zcorpan> | Igorrrrr: character references in HTML are parsed the same regardless of hte set encoding |
| 13:04 | <Igorrrrr> | zcorpan, please explain this to me noone can and I can't find such ino on the web :S |
| 13:04 | <Igorrrrr> | zcorpan, how you mean they are parsed same |
| 13:05 | <Igorrrrr> | ok so when I write character reference - < in html file and save that file with html editor using utf8 encding how the character reference is saved on the hard disk? |
| 13:05 | <zcorpan> | there usually isn't any saving on the hard disk when loading a page in a browser (unless the user saves it to disk) |
| 13:07 | <zcorpan> | or do you mean just the step that you saved a file to disk from your editor? |
| 13:07 | <zcorpan> | in either case, < is still < |
| 13:08 | <zcorpan> | also, this channel probably isn't appropriate for this kind of question |
| 13:09 | <zcorpan> | or at least people here might not respond helpfully for this kind of question :-) |
| 13:17 | <Igorrrrr> | zcorpan, no I mean when I save the file on the server hard disk |
| 13:18 | <hsivonen> | yay spec bugs: https://bugzilla.mozilla.org/show_bug.cgi?id=884795 |
| 13:18 | <zcorpan> | Igorrrrr: why not inspect the file with a hex editor and see for yourself |
| 13:19 | <Igorrrrr> | zcorpan, becasue I don;t have hex editor and I am not sure how to do it never done it |
| 13:21 | <zcorpan> | Igorrrrr: < as utf-8 is the following bytes: 26 6C 74 3B |
| 13:25 | <Igorrrrr> | so zcorpan < will be saved using utf8? |
| 13:25 | <zcorpan> | Igorrrrr: yes. if you save as utf-8... |
| 13:25 | <Igorrrrr> | ok thanks zcorpan |
| 13:25 | <Igorrrrr> | the web browser then when see character reference will represent literally the character instead of its meaning? |
| 13:27 | <zcorpan> | the browser represents it as < |
| 13:28 | <Igorrrrr> | zcorpan, literally not what it means |
| 13:28 | <Igorrrrr> | or as a string escaped |
| 13:32 | <Igorrrrr> | ok zcorpan care to answer 1 more question? |
| 13:32 | <Igorrrrr> | in < how is "#x003C" part called without the ambiguous ampersand |
| 13:33 | <Lachy> | Igorrrrr, what do you mean by "ambiguous ampersand"? |
| 13:33 | <Igorrrrr> | we are finding definition on the #web ^^ |
| 13:33 | <Igorrrrr> | just "#x003C" part |
| 13:33 | <Igorrrrr> | without "&;" |
| 13:33 | <Igorrrrr> | how it is called? |
| 13:33 | <hsivonen> | Igorrrrr: I think there isn't a specific term for that |
| 13:33 | <Igorrrrr> | html hexadecimal entity or? what is it's definition? |
| 13:33 | <Lachy> | that part doesn't have a name on its own. < is a numeric character reference |
| 13:34 | <Igorrrrr> | so "lt" how is called? |
| 13:34 | <Igorrrrr> | ok then I guess they don;t have specific name ok |
| 13:34 | <hsivonen> | hexadecimal numeric character reference, rather |
| 13:35 | <Igorrrrr> | and "&;" is ambiguous ampersand? |
| 13:35 | <zcorpan> | no |
| 13:35 | <Lachy> | Igorrrrr, see the spec http://www.whatwg.org/specs/web-apps/current-work/multipage/syntax.html#character-references |
| 13:35 | <Igorrrrr> | no? -_- I though yes :( |
| 13:35 | <Igorrrrr> | I am Lachy but as I understood &; is ambiguous ampersand :S |
| 13:35 | <hsivonen> | Igorrrrr: & is an ampersand. whether it's ambiguous depends on context |
| 13:36 | <Igorrrrr> | see An ambiguous ampersand is a U+0026 AMPERSAND character (&) that is followed by one or more alphanumeric ASCII characters, followed by a U+003B SEMICOLON character (;), where these characters do not match any of the names given in the named character references section. |
| 13:36 | <Igorrrrr> | this is said in the spec so I guess I am right zcorpan ? |
| 13:36 | <zcorpan> | Igorrrrr: ...one or more alphanumeric ASCII characters... |
| 13:36 | <zcorpan> | "&;" doesn't have that |
| 13:37 | <zcorpan> | &lol; would have an ambiguous ampersand |
| 13:37 | <Lachy> | Igorrrrr, in case it's unclear, alphanumeric means A-Z a-z or 0-9. < is followed by a #, not an alphanumeric character. |
| 13:39 | <Igorrrrr> | yeah sorry my fault |
| 13:39 | <Igorrrrr> | yeye my fault ok thank you all thanks |
| 13:39 | <Igorrrrr> | and sorry for offtopic discussion but I wanted to hear from the makers ^^ |
| 13:39 | <Igorrrrr> | keep up the good work |
| 13:41 | <zcorpan> | np |
| 13:52 | <zcorpan> | "Copyright (c) 2013 Nolan Waite. All rights reserved." https://github.com/nolanw/HTMLReader |
| 13:56 | <jgraham> | zcorpan: It's iOS, he probably wants to chage you $0.99 to use it |
| 13:57 | <jgraham> | Although it also says public-domain |
| 13:58 | <gsnedders> | Where does it say that? |
| 13:58 | <gsnedders> | Oh, in podspec. |
| 13:58 | <gsnedders> | Someone should file an issue on that. I vote zcorpan or jgraham, as they noticed the contradiction. |
| 13:58 | <zcorpan> | i'll file |
| 14:00 | <zcorpan> | there |
| 14:01 | <gsnedders> | hsivonen: Also, now you're back from no-internet-land, you might be interested in all the various changes to html5lib-tests |
| 14:01 | <hsivonen> | gsnedders: ok |
| 14:01 | <gsnedders> | hsivonen: (Mainly in case you think I've accepted any pull request I shouldn't have) |
| 14:02 | hsivonen | is now working through bugzilla needinfos |
| 14:07 | <hsivonen> | annevk: so charset menu non-using sessions account for somewhere between 99.98% and 99.99% of Firefox sessions |
| 14:11 | <annevk> | hsivonen: wow |
| 14:11 | <annevk> | so .015 * .5B |
| 14:12 | <annevk> | that's still a lot of sessions, although the stats are a bit off I'm sure |
| 14:15 | <annevk> | jgraham: so there's input, base, canonical, and then up to nine components |
| 14:15 | <annevk> | jgraham: with maybe more components in the future, depending on how we do this |
| 14:16 | <jgraham> | annevk: Can you make the components optional somehow? |
| 14:16 | <jgraham> | Or perhaps they are deterministic once you have canonical? |
| 14:18 | <annevk> | Maybe if I write down the actual components and not what's exposed to JavaScript and then do the normalization later on... |
| 14:18 | <annevk> | The JavaScript components don't expose all the details. E.g. /html? and /html both have .search as "" |
| 14:19 | <annevk> | Maybe it could be something like input{space}base{space}[u:{username}][p:{password] etc. |
| 14:20 | <annevk> | And if you want a space, use \s or some such? |
| 14:20 | <annevk> | And \r, \n, \t should be there I guess and maybe some generic escape |
| 14:33 | <annevk> | I guess I could even optimize by making the base the same as the line before if you just use two spaces |
| 14:51 | <hsivonen> | annevk: https://bugzilla.mozilla.org/show_bug.cgi?id=865916#c55 |
| 14:53 | <annevk> | That's pretty interesting. I guess it means it's less and less needed to have this menu... |
| 14:56 | <hsivonen> | too bad I didn't include code to test for Ruby's Postulate |
| 15:17 | zcorpan | notices tpac TAG meetings are Member Confidential http://www.w3.org/2013/11/TPAC/ |
| 15:18 | <hsivonen> | where's the reform?!?! |
| 15:18 | <zcorpan> | annevk: i'm very much dissappoint!! |
| 15:26 | <annevk> | I'll ask. I didn't even know. |
| 15:26 | <annevk> | You're all welcome though. The chairs can fix that error. |
| 16:14 | <annevk> | So I got this format now: |
| 16:14 | <annevk> | http://example\t.\norg http://example.org/foo/bar s:http h:example.org p:/ |
| 16:14 | <annevk> | http://user:pass@foo:21/bar;par?b#c s:http u:user pass:pass h:foo port:21 p:/bar;par q:?b f:#c |
| 16:14 | <annevk> | http:foo.com s:http h:example.org p:/foo/foo.com |
| 16:15 | <annevk> | I guess I should just go with that for now and ask for wider input later... |
| 16:16 | <SimonSapin> | annevk: what if the path contains " q:"? |
| 16:16 | <annevk> | SimonSapin: what about it? |
| 16:17 | <annevk> | SimonSapin: ooh, that would become \sq: |
| 16:17 | <SimonSapin> | isn’t that syntax ambiguous? |
| 16:17 | <SimonSapin> | oh |
| 16:17 | <annevk> | I guess I could make it tab-separated as that's more conventional. I'm not a big fan of tabs |
| 16:19 | <jgraham> | No, tab seperated is bas |
| 16:19 | <jgraham> | *bad |
| 16:19 | <jgraham> | I would rather just escape spaces in some way |
| 16:19 | <jgraham> | So the \s thing seems fine |
| 16:19 | <annevk> | Why is it bad? |
| 16:20 | <jgraham> | Because literal tab characters are hard to enter in many editors |
| 16:20 | <jgraham> | And it's easy to accidentially add spaces rather than tabs |
| 16:21 | <annevk> | Yeah makes sense, that's why I preferred spaces too |
| 16:28 | <annevk> | And I guess I'll go with "#{anything}" lines are ignored and " #{anything}" at the end of line are ignored too for comments |
| 17:40 | <TabAtkins> | zcorpan, SimonSapin: Yeah, on further review, I think the squashing of duplicates goes in Cascade, not Syntax. I'll rejigger things. |
| 17:41 | <TabAtkins> | zcorpan: It should work fine, I guess. It just uses a bunch of Bert's preprocessor shortcuts, which Anolis also uses. You'd just need to write your own boilerplate files (the stuff in the /include dir). (Feel free to PR me with them if you do so.) |
| 17:47 | <TabAtkins> | zcorpan: Note that Shepherd already parses non-CSS specs to look for anchors: http://test.csswg.org/shepherd/api/spec |
| 17:48 | <TabAtkins> | zcorpan: You can just ask plinss if you want more to show up in that list. |
| 17:49 | <zcorpan> | TabAtkins: ok. i'll look into moving over cssom at some point |
| 17:49 | <TabAtkins> | zcorpan: Cool. I can play with it first if you'd like. |
| 17:49 | <TabAtkins> | Also: name suggestions for my processor? |
| 17:57 | <zcorpan> | some online name generator suggests Dead Fist |
| 17:58 | <TabAtkins> | zcorpan: SOLD |
| 18:06 | <zcorpan> | quick search for 'dead fist' gives http://en.wikipedia.org/wiki/Neil_Burke and http://www.fanfiction.net/s/7800685/1/Dead-Fist |
| 18:07 | <TabAtkins> | Both of these seem appropriate to link to my processor. |
| 18:09 | <zcorpan> | annevk: what is this url syntax thing for? |
| 18:10 | <annevk> | zcorpan: portable test format |
| 18:10 | <zcorpan> | ah |
| 18:10 | <annevk> | need something that supports lone surrogates and is fairly easy to write and supports comments |
| 18:10 | <annevk> | and since writing parsers is easy... |
| 18:13 | <zcorpan> | so you support character escapes for lone surrogates? |
| 18:13 | <annevk> | haven't really figured out exactly whether I want code units or code point escapes |
| 18:14 | <annevk> | I suspect \u{SIX HEX DIGITS} so e.g. \u00FFFD will be the format |
| 18:14 | <annevk> | well, unless code units, in which case \uFFFD |
| 18:14 | <annevk> | code units might be fine |
| 18:15 | <zcorpan> | i think code units seems saner if you want to represent surrogates |
| 18:15 | <annevk> | Unicode does both, but then maybe Unicode is insane |
| 18:18 | <zcorpan> | i guess more of an historical quirk |
| 18:21 | <TabAtkins> | Yeah, surrogates are a quirk. UTF-16 is the weird insane encoding. |
| 18:23 | <TabAtkins> | annevk: What's the best way to invoke Encoding to ensure that CSS doesn't receive unpaired surrogates? |
| 18:23 | <TabAtkins> | annevk: None of the encodings allow them - you can only get them via setting manually-created strings in JS. |
| 18:23 | <annevk> | TabAtkins: you'll get surrogates through JS |
| 18:23 | <annevk> | TabAtkins: that's right |
| 18:24 | <TabAtkins> | Can we just say to parse JS strings as UTF-16, so those surrogates turn into fffd? |
| 18:25 | <annevk> | TabAtkins: You can have all the API interactions make use of http://dev.w3.org/2006/webapi/WebIDL/#dfn-obtain-unicode currently [EnsureUTF16] in IDL, however, it's not entirely clear what that buys us |
| 18:25 | <annevk> | TabAtkins: except for worse perf |
| 18:26 | <annevk> | I had this discussion with SimonSapin earlier |
| 18:26 | <TabAtkins> | Ah, kk. ^_^ |
| 18:27 | <annevk> | We've been talking a lot about strings lately. Need to figure out the story for Servo too... My feeling at the moment is that we're rather stuck with the 16-bit integers |
| 18:27 | <SimonSapin> | JS strings are insane |
| 18:27 | <annevk> | Hmm, could be worse, like Python :p |
| 18:28 | <SimonSapin> | come on :p |
| 18:28 | <SimonSapin> | there are no accidental lone surrogate in python |
| 18:28 | <TabAtkins> | It's true, Python's strings are worse than JS. |
| 18:28 | <TabAtkins> | No, there's accidental encoding errors FUCKING EVERYWHERE. |
| 18:29 | <Domenic_> | ES6 supports \u{123456} |
| 18:29 | <SimonSapin> | TabAtkins: embrace the Unicode Sandwich: http://nedbatchelder.com/text/unipain.html#h_Pain_relief |
| 18:29 | <SimonSapin> | TabAtkins: JS doesn’t even have bytes, so sure you don’t get errors for byte/text conversions |
| 18:30 | <Domenic_> | Unicode in ES6 http://www.slideshare.net/domenicdenicola/es6-is-nigh/40 + http://www.slideshare.net/domenicdenicola/es6-is-nigh/41 |
| 18:31 | <SimonSapin> | TabAtkins: also you could do yourself a favor and switch to Python 3 |
| 18:33 | <TabAtkins> | Yeah, prolly should. |
| 18:33 | <SimonSapin> | TabAtkins: I doubt implementers will agree to do character decoding on JS strings |
| 18:34 | <zcorpan> | TabAtkins: just let the lone surrogates through, i'd say |
| 18:34 | <SimonSapin> | close you eyes and pretend you haven’t seen them |
| 18:34 | <zcorpan> | TabAtkins: like document.write and the DOM API |
| 18:35 | <SimonSapin> | also, "Unicode scalar values" is just ridiculous |
| 18:35 | <zcorpan> | put it on the band names wiki page |
| 18:35 | <annevk> | Domenic_: yeah, as a syntax feature that makes sense |
| 18:36 | <annevk> | Domenic_: haven't seen that iterator in the draft yet though |
| 18:37 | <Domenic_> | annevk: good point, i wonder if it was just forgotten or if consensus has changed since i put those slides together |
| 18:37 | <annevk> | I vaguely recall now it'll go in a module of sorts |
| 18:38 | <annevk> | Found out the other day ES6 has already been taking close to 4 years and the draft still has many holes... |
| 18:40 | <Ms2ger> | Ah, versioned standards |
| 18:43 | <zcorpan> | SimonSapin: fwiw getDefaultComputedStyle exists in gecko. whether it's expensive or not i dunno |
| 18:58 | <SimonSapin> | zcorpan: what does it do? |
| 18:59 | <zcorpan> | SimonSapin: basically the same as .defaultStyle in CSSOM |
| 19:00 | <zcorpan> | SimonSapin: returns cascaded style without author styles |
| 19:06 | <zcorpan> | or computed style |
| 19:07 | <zcorpan> | i guess i should remove defaultStyle |
| 19:08 | <zcorpan> | and being even more anal about use cases in the future |
| 20:02 | <zcorpan> | hmm, seems http://www.w3.org/2009/07/webidl-check doesn't check for errors in extended attributes |
| 20:02 | <zcorpan> | maybe that's intentional |
| 20:03 | <zcorpan> | but not good for spec-writing |
| 20:03 | <Ms2ger> | Extended attributes are for distributed extensibility |
| 20:04 | <Hixie> | hsivonen: can you elaborate on your question about <template> and <title>? I don't recall doing anything special for <title> when merging <template> in |
| 20:04 | <Hixie> | hsivonen: support for <frame> was dropped because as specified before it was quite broken, and actually making it work seemed like a lot of effort for something that has been deprecated for like 15 years |
| 20:05 | <zcorpan> | Ms2ger: then webidl shouldn't use them for its own features |
| 20:05 | <Ms2ger> | That's probably fair |
| 20:09 | <jsbell> | We could prefix (etc) the ones used by browser implementations to control code generation, if we can agree on a syntax. For blink we just have a doc that marks them as as non-standard or links to the standard clause. |
| 20:10 | <Ms2ger> | Well, the idea is that specs can extend it too |
| 20:10 | <Ms2ger> | I believe webgl even does that |
| 20:11 | <zcorpan> | nice, i'll replace all my spec prose with a custom extended attribute |
| 20:12 | <zcorpan> | the meaning of which i define in an appendix |
| 20:12 | <Hixie> | oooh, good idea |
| 20:13 | <Hixie> | we can replace all our specs with one spec that has one IDL block |
| 20:13 | <Hixie> | [Browser] |
| 20:13 | <Hixie> | [Browser] means "follow these specs..." |
| 20:13 | <Ms2ger> | "...and don't follow these" |
| 20:18 | <nimbu> | zcorpan: https://www.w3.org/Bugs/Public/show_bug.cgi?id=17518 most browsers dont obey that |
| 20:18 | <nimbu> | so i guess i would have to file browser bugsssss :|||||| |
| 20:18 | <nimbu> | and i dont even remember or have the code. |
| 20:20 | <Hixie> | how the heck is this happening: https://bugzilla.mozilla.org/show_bug.cgi?id=884795 |
| 20:20 | <Hixie> | happens in both chrome and firefox |
| 20:21 | <Hixie> | but i can't work out why |
| 20:21 | <Hixie> | there's no furthest block, so the AAA should just go 1-9 and stop, no? |
| 20:21 | <zcorpan> | Hixie: my testing suggests that blink and gecko do follow that. http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2445 |
| 20:22 | <zcorpan> | er |
| 20:22 | <zcorpan> | s/Hixie/numbu/ |
| 20:24 | <Hixie> | http://software.hixie.ch/utilities/js/live-dom-viewer/?saved=2446 |
| 20:24 | <Hixie> | removing the attribute fixes it |
| 20:24 | <Hixie> | wtf |
| 20:25 | <Hixie> | oh.... http://www.whatwg.org/specs/web-apps/current-work/#push-onto-the-list-of-active-formatting-elements |
| 20:25 | <Hixie> | the Noah's Ark clause |
| 20:26 | <Hixie> | but wait, no, that should still not matter |
| 20:32 | <Hixie> | ohhh, i get it |
| 20:33 | <zcorpan> | Hixie: "If formatting element is not the current node, this is a parse error. (But do not abort these steps.)" doesn't explain the behavior but still seems relevant |
| 20:34 | <zcorpan> | step 7 or AAA |
| 20:34 | <Hixie> | what's going on is this: |
| 20:34 | <Hixie> | after <code a> <code> <code><code><code> |
| 20:34 | <Hixie> | there's four <code>s on the formatting list |
| 20:34 | <Hixie> | and five on the stack |
| 20:34 | <Hixie> | and the ones on the list are given precedence for some reason |
| 20:35 | <Hixie> | so then the last three get closed and you see the next </code> |
| 20:35 | <Hixie> | and it closes the first one (<code a>) rather than the second one, because the first is on the list and the second isn't |
| 20:36 | <zcorpan> | so it should look at the stack of open elements and not leave the list of formatting elements alone if it's not in that list? |
| 20:37 | <zcorpan> | and step 7 should probably also just look at the stack of open elements |
| 20:39 | <Hixie> | I think the solution is to make the AAA check the stack first and only do its stuff if the current node isn't a matching formatting element, yeah. |
| 20:41 | Hixie | files a spec bug |
| 20:41 | <Hixie> | https://www.w3.org/Bugs/Public/show_bug.cgi?id=22926 |
| 23:29 | <Igor^> | can I send mail from html with a element? |
| 23:32 | <Hixie> | you can use <form> to prompt the user's e-mail client to be prefilled with an e-mail |
| 23:32 | <Hixie> | but you can't send mail directly (it would be used for spam) |
| 23:36 | <Igor^> | ok thanks Hixie |