| 00:00 | <zewt> | the "widths" is just a fake character-width mapping; the values are only chosen to be non-random |
| 00:04 | <TabAtkins> | Ah, okay. It's not very useful as a demonstration, then, since they're very off (and the first part of the alphabet is negative-width). |
| 00:04 | <zewt> | well, it's outputting in a fixed-width terminal, so it's not going to look like it really would anyway, heh |
| 00:05 | <zewt> | (also in reality you've got kerning, etc. to cope with) |
| 05:48 | <Von_Davidicus> | I have just learned that using HTML5 + SVG alleviates an inconsistency in browser implementation of JavaScript. |
| 05:49 | <Von_Davidicus> | Specifically, whether or not to include the SVG prefix. |
| 05:51 | <Von_Davidicus> | Some browsers (like Firefox and Internet Explorer) say "Thou Shalt." Others (like Opera, Chrome, and Safari) say "Thou Shan't." |
| 05:56 | <Von_Davidicus> | Since you don't need to use the prefix in HTML5, it becomes a non-issue. |
| 06:49 | <hsivonen> | yeah, I should review dglazkov's parser change proposal. sorry about the delay. |
| 06:52 | <hsivonen> | so when a parser gets aborted, the document's readyState in Gecko goes from "loading" to "complete" without going through "interactive" |
| 06:52 | <hsivonen> | is there any spec-based justification for that? |
| 07:01 | <annevk> | hmm |
| 07:01 | <annevk> | html5.org gives 500 errors |
| 07:03 | <hsivonen> | hg blame points to sicking for the readyState oddity, but he's on vacation. :-( |
| 07:03 | <Hixie> | hsivonen: looks like the spec doesn't change readyState when the parser is aborted |
| 07:03 | <Hixie> | i doubt this is an especially well tested area of the spec though |
| 07:04 | <hsivonen> | Hixie: not even to "complete"? |
| 07:04 | <Hixie> | nothing mentions changing it when a document or parser is aborted |
| 07:04 | <Hixie> | you also don't fire 'load', etc |
| 07:04 | <hsivonen> | Hixie: I tought data was discarded from the parser but the parser still went through "the end" |
| 07:05 | <hsivonen> | *thought even |
| 07:05 | <Hixie> | oh hm maybe, let me check again |
| 07:05 | <Hixie> | no, "abort a parser" doesn't really do anything except a hard stop |
| 07:05 | <Hixie> | again, though, this may be severely undertested |
| 07:06 | <hsivonen> | this spec thing was supposed to spare me of testing what other browsers do |
| 07:08 | <Hixie> | well you're welcome to do what the spec says :-) |
| 07:08 | <Hixie> | if everyone does that, then we'll all be interoperable :_) |
| 07:08 | <annevk> | hmm |
| 07:08 | <annevk> | so both Unicode and the utf-8 RFC don't say anything about five and six byte sequences anymore |
| 07:08 | <nesta_> | good morning! :) |
| 07:08 | <annevk> | morning |
| 07:09 | <annevk> | guess that makes the case of having a utf-8 decoder in the spec more compelling |
| 07:09 | <hsivonen> | Hixie: I doubt I'd get r+ for leaving readyState to "loading" when the throbber has stopped |
| 07:09 | <annevk> | Hixie: btw, any concerns about integrating with http://dvcs.w3.org/hg/encoding/raw-file/tip/Overview.html in due course? |
| 07:09 | <hsivonen> | Hixie: without proof of other browsers doing so |
| 07:09 | <annevk> | Hixie: quite a lot of the encoding related stuff would be simplified per my analysis |
| 07:10 | <Hixie> | hsivonen: convince another browser to follow the spec first? :-) |
| 07:10 | <Hixie> | annevk: sgtm |
| 07:10 | <hsivonen> | Hixie: I'd still have to test the other browser |
| 07:11 | <Hixie> | hsivonen: you could hope they wrote the tests as part of checking in the code |
| 07:11 | <Hixie> | annevk: "The Encoding" lol |
| 07:12 | <Hixie> | annevk: wasn't there some punctuation stripping on encoding names too? |
| 07:15 | <annevk> | no, just whitespace |
| 07:15 | <annevk> | it used to be punctuation and we changed that because it was incompatible |
| 07:16 | <Hixie> | ah k |
| 07:16 | <Hixie> | this is pretty awesome |
| 07:16 | <Hixie> | basically means i can drop all mention of the IANA charset registry, right? |
| 07:17 | <annevk> | yes |
| 07:17 | <Hixie> | fantastic! |
| 07:17 | <annevk> | and the special utf-8 text |
| 07:17 | Hixie | claps his hands |
| 07:22 | <MikeSmith> | annevk: http://qa-dev.w3.org:8888/?doc=data%3Atext%2Fhtml%3Bcharset%3Dutf-8%2C%3C%21doctype%2520html%3E%250D%250A%3Ctitle%3E%3C%252Ftitle%3E%250D%250A%3Cinput%2520type%253Dimage%2520value%253Dfoo%2520alt%253Dbar%3E&showsource=yes |
| 07:22 | <MikeSmith> | now with highlighting |
| 07:22 | <MikeSmith> | zcorpan: ↑ |
| 07:23 | <zcorpan> | :thumbsup: |
| 07:23 | <zcorpan> | "formnovalidate" gets split on two lines for me |
| 07:24 | <MikeSmith> | yeah, I noticed that problem in opera |
| 07:24 | <MikeSmith> | I don't know how to fix that |
| 07:24 | <annevk> | nice |
| 07:25 | <hsivonen> | Objection and revert request at the HTML WG over dropping IANA registry mention coming up in 1, 2, 3, ... |
| 07:26 | <zcorpan> | MikeSmith: .inputattrname { word-wrap:normal } |
| 07:27 | <MikeSmith> | zcorpan: OK, will make that change |
| 07:37 | <jgraham> | hsivonen: BTW if you were going to write some tests for this, I would be especially interested in them :) |
| 07:39 | <zcorpan> | heycam|away: why does Array[] conversion from JS value to IDL value not accept a user object like {'0':'foo', 'length':1}? (it uses the same rule as sequence<T> which throws TypeError for user objects, if i'm reading it correctly) |
| 07:46 | <MikeSmith> | zcorpan: can you please re-check and confirm if that wrapping problem is fixed? |
| 07:46 | <MikeSmith> | btw, I couldn't reproduce that in Opera on OSX |
| 07:46 | <MikeSmith> | only on Opera Mobile |
| 07:46 | <MikeSmith> | maybe it's a Linux bug |
| 07:47 | <MikeSmith> | hsivonen: when you have a minute to look at http://qa-dev.w3.org:8888/?doc=data%3Atext%2Fhtml%3Bcharset%3Dutf-8%2C%3C%21doctype%2520html%3E%250D%250A%3Ctitle%3E%3C%252Ftitle%3E%250D%250A%3Cinput%2520type%253Dimage%2520value%253Dfoo%2520alt%253Dbar%3E&showsource=yes please let me know what you think |
| 07:49 | <hsivonen> | MikeSmith: cool |
| 07:50 | <MikeSmith> | will send you the patch |
| 07:50 | <hsivonen> | MikeSmith: since <input> is special-cased, did you actively decide to list all the attributes for <input> instead of showing a pruned list for each <input type=foo>? |
| 07:51 | <hsivonen> | (I'm not suggesting that either is the right way. Just wondering what was considered.) |
| 07:51 | <MikeSmith> | I did consider listing only the ones for each type, but it seems useful to have them all shown in context, so that the user can see where they are allowed and not allowed |
| 07:51 | <zcorpan> | MikeSmith: i have OS X |
| 07:52 | <hsivonen> | actually, showing them all is probably a good way to pre-empt "but <input> does have attribute foo. the validator is wrong" |
| 07:52 | <hsivonen> | MikeSmith: ok. let's get this landed |
| 07:52 | <MikeSmith> | hsivonen: bingo |
| 07:52 | <MikeSmith> | OK |
| 07:52 | <MikeSmith> | zcorpan: strange. I couldn't reproduce it in either 11.6 nor Opera.Next |
| 07:53 | <zcorpan> | MikeSmith: did you try making the window narrower? |
| 07:53 | <MikeSmith> | ah |
| 07:53 | <MikeSmith> | yeah, that's right |
| 07:53 | <MikeSmith> | I was just looking at in at normal width |
| 07:54 | <MikeSmith> | zcorpan: OK, yeah, so it does seem to be fixed now. Right? |
| 07:55 | <zcorpan> | MikeSmith: class="highlight" the semanticists and accessibility-ists will haunt you :-) |
| 07:56 | <MikeSmith> | heh |
| 07:56 | <MikeSmith> | I guess I should change that |
| 07:56 | <MikeSmith> | suggestions? |
| 07:56 | <zcorpan> | source highlight uses <b> |
| 07:56 | <MikeSmith> | hmm, "highlight" is not presentational |
| 07:57 | <MikeSmith> | anyway, I can tweak that stuff later I guess |
| 07:58 | <annevk> | hsivonen: lets see what happens to https://www.w3.org/Bugs/Public/show_bug.cgi?id=16768 |
| 08:07 | <MikeSmith> | hsivonen: patch sent |
| 08:13 | <hsivonen> | MikeSmith: Java could use some syntactic sugar for literal initializers for collections... |
| 08:13 | <hsivonen> | JavaScript FTW |
| 08:13 | <MikeSmith> | heh |
| 08:13 | <MikeSmith> | yeah |
| 08:25 | <hsivonen> | MikeSmith: thanks. reviewed. |
| 08:26 | <MikeSmith> | thanks |
| 08:49 | <hsivonen> | pulseaudio got memes. still waiting for Opera and IE: http://pulseaudiomemes.tumblr.com/ |
| 09:50 | <hsivonen> | is there a JS-exposed way to abort a document without immediately causing navigation away from the document? |
| 09:52 | <annevk> | there was window.stop() at some point I thought |
| 09:53 | <annevk> | hmm |
| 09:53 | <annevk> | "The stop() method on Window objects should, if there is an existing attempt to navigate the browsing context and that attempt is not currently running the unload a document algorithm, cancel that navigation and any associated instances of the fetch algorithm. Otherwise, it must do nothing." |
| 09:53 | <annevk> | I forgot whether it's called navigating if a document is loading |
| 09:54 | <jgraham> | Yeah I was thinking the same |
| 09:54 | <jgraham> | I'm not sure if "cancel the navigation" is the same as "abort the document" |
| 09:55 | <jgraham> | Seems like it isn't per spec |
| 09:56 | <jgraham> | Or at least, I don't really know how "cancel the navigation" is defined |
| 09:57 | <jgraham> | But the navigation algorithm cancels previous navigations and aborts as a seperate step |
| 09:59 | <hsivonen> | readyState in Firefox is flaky, episode 42: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1478 |
| 09:59 | <hsivonen> | I'm increasingly unhappy about readyState support landing without thorough assertions |
| 09:59 | <hsivonen> | bonus points for test cases |
| 10:01 | <hsivonen> | the above case I fixed for Firefox 14 |
| 10:02 | <jgraham> | hsivonen: Can we expect to you make a W3C-sharable testsuite gor this (hint ;) |
| 10:02 | <jgraham> | *for |
| 10:04 | <hsivonen> | jgraham: not a safe expectation, unfortanately |
| 10:04 | <hsivonen> | I do plan to add fatal-in-debug-builds assertions to make sure people don't get to regress this stuff easily |
| 10:09 | jgraham | is pretty sure that the Mozilla Manifesto requires writing sharable tests ;) |
| 10:09 | <hsivonen> | Opera Next fires onload after window.stop() |
| 10:10 | <hsivonen> | Firefox and Chrome don't |
| 10:10 | <hsivonen> | Chrome transitions to readyState == "complete" synchronously upon window.stop(), which worries me |
| 10:10 | <hsivonen> | I'd expect a transition to "interactive" and an async transition to "complete" later |
| 10:11 | <jgraham> | The Opera thing sounds like a bug |
| 10:11 | <hsivonen> | jgraham: shareable test case: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1479 |
| 10:11 | <hsivonen> | I believe the Mozilla Manifesto doesn't require filing Opera bugs after pointing them out on IRC :-) |
| 10:11 | <annevk> | why should load not fire? |
| 10:12 | <jgraham> | hsivonen: Point :) |
| 10:12 | <jgraham> | annevk: Why would it? |
| 10:12 | <hsivonen> | IE9 doesn't support window.stop() |
| 10:13 | <annevk> | jgraham: nothing delays the load event anymore |
| 10:15 | <annevk> | kind of depends on how all the algorithms hook into each other of course |
| 10:15 | <annevk> | there's quite a lot of them |
| 10:15 | <jgraham> | annevk: Where is the load event fired? |
| 10:15 | <annevk> | when you stop parsing |
| 10:15 | <hsivonen> | I expected the load event not to fire when aborting |
| 10:15 | <annevk> | well an async process is started at that point |
| 10:15 | <jgraham> | Hmm, maybe that is still called if you abort the navigation then |
| 10:15 | <jgraham> | s/abort/cancel/ |
| 10:16 | <hsivonen> | window.stop() in Gecko and WebKit seems to abort |
| 10:16 | <jgraham> | But "cancel the navigation" isn't obviously well defined |
| 10:16 | <hsivonen> | at least superficially |
| 10:16 | <hsivonen> | I'm going to be unhappy if Facebook depends on the details of this stuff |
| 10:19 | <hsivonen> | boo. IE10 doesn't support window.stop() |
| 10:19 | <smaug____> | what? |
| 10:20 | <hsivonen> | smaug____: Object doesn't support property or method 'stop' on line 3 |
| 10:20 | <jgraham> | So does anyone disagree that "cancel the navigation" is undefined? If so, where is it defined? Otherwise I will file a bug |
| 10:21 | <hsivonen> | which browsers fire event for readyState transitions? |
| 10:21 | <hsivonen> | *events |
| 10:21 | <hsivonen> | do any? |
| 10:22 | <hsivonen> | Chrome does |
| 10:23 | <hsivonen> | even worse, it looks like it indeed transitions straight from "loading" to "complete" |
| 10:23 | <hsivonen> | sigh. is skipping "interactive" in some cases an IEism that Gecko and WebKit emulate? |
| 10:25 | <gsnedders> | hsivonen: We have interactive |
| 10:26 | <gsnedders> | (Gmaps relies on it, or did) |
| 10:27 | <hsivonen> | gsnedders: "interactive" in what situation? |
| 10:27 | <hsivonen> | gsnedders: WebKit and Gecko skip "interactive" in *some* situations |
| 10:27 | <gsnedders> | hsivonen: Dunno. Not looked at the code :) |
| 10:27 | <hsivonen> | gsnedders: Gecko skips it when a parser is aborted or when an XSLT transform fails to compile |
| 10:28 | <hsivonen> | the former seemed to be an intentional change by sicking |
| 10:28 | <hsivonen> | the latter seems like an oversight |
| 10:28 | <hsivonen> | Chrome skips "interactive" when calling window.stop() from a script-created document |
| 10:30 | <hsivonen> | I'd like documents not to skip "interactive" unless Web compat requires "interactive" to be skipped sometimes |
| 10:30 | <hsivonen> | because skipping "interactive" makes it unclear if there's a browser bug or an intentional skip |
| 10:32 | <hsivonen> | readyState is one of those things that seem simple on surface but are a huge mess when you start investigating what's going on |
| 11:03 | <annevk> | if you do set([...]) and you get TypeError: 'dict' object is not callable in Python what is going on? |
| 11:09 | <zcorpan> | annevk: i see 16 cp1250 encoding declarations in web200904 |
| 11:09 | <zcorpan> | 10 of which are in the http header |
| 11:09 | <annevk> | yeah, labels are somewhat conservative at the moment |
| 11:10 | <annevk> | http://wiki.whatwg.org/wiki/Encoding#Labels |
| 11:10 | <annevk> | the reason is probably because in one browser nothing happened for that label |
| 11:11 | <annevk> | right |
| 11:11 | <annevk> | Internet Explorer |
| 11:11 | <annevk> | Had all these as labels for "obviouslyfake" in my single-byte encoding research: csiso88598i, iso8859-11, iso8859-10, iso8859-14, csisolatin6, l6, obviouslyfake3, iso-8859-8-e, obviouslyfake2, iso8859-3, cp1255, iso8859-7, iso8859-5, cp1254, iso8859-9, iso8859-8, mac, cp1253, x-mac-roman, csmacintosh, iso-8859-16, iso8859-6, iso8859-13, iso-8859-6-e, iso-8859-6-i, iso8859-4, iso-ir-157, iso8859-15, cp1257, iso-8859-14, cp1251, cp1250, cp1258, |
| 11:11 | <annevk> | latin6, iso-8859-10 |
| 11:13 | <hasather> | annevk: did you overwrite `set` with a dict? |
| 11:13 | <annevk> | less conservative is probably okay, but you cannot be liberal, recognizing euc_jp as euc-jp is wrong |
| 11:13 | <zcorpan> | $ grep -aPih "text/html\s*;\s*charset\s*=\s*[\"']?(866|ansi_x3.4-1986|asmo-708|cn-gb|cp1250|cp1251|cp1252|cp1254|cp1257|cp367|cp50220|cp51932|cp819|cp932|cp936|csascii|cscp50220|cscp51932|cseuckr|cseucpkdfmtjapanese|csibm866|csinvariant|csiso646basic1983|csiso88596e|csiso88596i|iso-8859-6-i|csiso88598e|csiso88598i|cskoi8r|csunicode|csunicode11|csunicode11utf7|utf-7|csunicodeascii|csunicodejapanese|csunicodelatin1|csviscii|viscii| |
| 11:13 | <zcorpan> | cn|euc-tw|euc-tw|extended_unix_code_packed_format_for_japanese|ibm367|ibm819|invariant|iso-10646|iso-10646-j-1|iso-10646-ucs-2|iso-10646-ucs-basic|iso-10646-unicode-latin1|iso-2022-cn|iso-2022-cn|iso-2022-jp-1|iso-2022-jp-1|iso-8859-6-e|iso-8859-6-i|iso-8859-6-i|iso-8859-8-e|iso-celtic|iso-ir-100|iso-ir-199|iso-ir-226|iso-ir-6|iso646-us|iso8859-11|iso8859-12|iso-8859-12|iso8859-13|iso8859-15|iso8859-16|iso8859-3|iso8859-4|iso8859 |
| 11:13 | <zcorpan> | 859-7|iso8859-8|iso8859-9|iso88591|iso885910|iso885911|iso885912|iso-8859-12|iso885913|iso885914|iso885915|iso885916|iso88592|iso88593|iso88594|iso88595|iso88596|iso88597|iso88598|iso88599|iso_646.basic:1983|iso_646.irv:1991|iso_8859-10:1992|iso_8859-14|iso_8859-14:1998|iso_8859-16|iso_8859-16:2001|iso_8859-1:1987|iso_8859-2:1987|iso_8859-3:1988|iso_8859-4:1988|iso_8859-5:1988|iso_8859-6-e|iso_8859-6-i|iso-8859-6-i|iso_8859-6:198 |
| 11:13 | <zcorpan> | |iso_8859-8-e|iso_8859-8-i|iso_8859-8:1988|iso_8859-9|iso_8859-9:1989|ks_c_5601-1987|l10|l8|latin-9|latin10|latin8|microsoft-cp1250|microsoft-cp1251|microsoft-cp1252|microsoft-cp1253|microsoft-cp1254|microsoft-cp1255|microsoft-cp1256|microsoft-cp1257|microsoft-cp1258|ms932|ms936|ref|sjis|tis-620-2533|unicode-1-1|unicode-1-1-utf-7|utf-7|us|utf-7|utf-7|viscii|viscii|windows-936|x-cp1252|x-cp1253|x-cp1254|x-cp1255|x-cp1256|x-cp1257| |
| 11:13 | <zcorpan> | |x-mac-ce|x-mac-greek|x-mac-greek|x-mac-turkish|x-mac-turkish|x-user-defined|x-user-defined)" web200904 > labels.txt |
| 11:15 | <annevk> | hasather: thanks, thought it was scoped to the for loop |
| 11:15 | <annevk> | thanks zcorpan |
| 11:32 | <annevk> | zcorpan: what does -aPih mean here? |
| 11:34 | <zcorpan> | annevk: see grep --help |
| 11:34 | <zcorpan> | http://simon.html5.org/dump/encoding-labels/ |
| 11:36 | <zcorpan> | cp1250: 9 ... hmm, the original grep found 16 |
| 11:37 | <zcorpan> | cp1254: 10646 |
| 11:38 | <zcorpan> | maybe the python counter is bogus |
| 11:41 | <zcorpan> | changed |
| 11:43 | smaug____ | wants better map services. Although Bing map is mostly great, satellite pictures should get updated daily |
| 11:43 | <smaug____> | ...so that I could check whether certain lake still has ice |
| 11:44 | <zcorpan> | hmm. the grep is also bogus... |
| 11:46 | <MikeSmith> | hsivonen: btw, as far as open validator stuff, a while back you said you were planning to land the VerifierCommandLine code from the pfalcon bitbucket fork. You still planning on doing that? |
| 11:50 | <annevk> | cp1254: 10646 is kind of alarming |
| 11:51 | <zcorpan> | yeah that was bogus, got 10 in the second run :-) |
| 11:52 | <zcorpan> | annevk: ok now the data should be more accurate |
| 11:53 | <zcorpan> | "us" had a high number previously, but is now 0, because the previous grep didn't check the next char so picked up "us" in e.g. "us-ascii" |
| 11:54 | <annevk> | sweet thanks |
| 11:54 | <zcorpan> | np |
| 11:55 | <annevk> | ideally we do a similar run over the top 1 million pages with help from foolip |
| 11:56 | <foolip> | annevk, zcorpan, to collect encoding label statistics? |
| 11:56 | <annevk> | yeah |
| 11:56 | <annevk> | to see if there's any labels that might be worth adding |
| 11:56 | <annevk> | though we'd have to be careful with adding false-positives |
| 11:57 | <foolip> | yeah, it wouldn't be hard to modify the script to get ~5 pages per domain, but I guess extracting the labels is the hard bit, currently I just use a regexp |
| 11:57 | <zcorpan> | we could grep for *all* labels |
| 11:57 | <foolip> | all labels in which list? |
| 11:57 | <zcorpan> | all labels that are specified in the wild |
| 11:58 | <zcorpan> | no list |
| 11:58 | <foolip> | right, so the problem becomes writing the encoding sniffing per spec, which I haven't done |
| 11:58 | Philip` | imagines it might be worth running the data through an HTML parser so you can look for <meta http-equiv=content-type ...> more reliably |
| 11:58 | <foolip> | right, that might work, even if it isn't exactly what the spec does |
| 11:59 | <annevk> | might also be slow |
| 12:00 | <Philip`> | Parsing in Java is fast |
| 12:00 | <foolip> | yeah, the spec sniffing of the first 1024 bytes should be fast and allow you to close the connection without getting the entire resource, if that's what you want |
| 12:00 | <Philip`> | (where "fast" is like tens of thousands of pages per minute, if I remember correctly) |
| 12:01 | <foolip> | well, I guess whoever writes the code will decide :) |
| 12:25 | <zcorpan> | annevk: now the data says 16 for cp1250 |
| 12:26 | <zcorpan> | annevk: i forgot to do a case-insensitive regexp in the python script |
| 12:26 | <annevk> | ah k |
| 12:26 | <annevk> | have done that too |
| 12:28 | <annevk> | some of the weird long ones are relatively common too |
| 12:28 | <zcorpan> | annevk: 'utf-7' is present 5 times in http://wiki.whatwg.org/wiki/Encoding#Labels |
| 12:31 | <annevk> | zcorpan: yeah I used some internal data |
| 12:31 | <annevk> | zcorpan: that encoding cannot be activated on the web |
| 12:32 | <annevk> | still used for email I think |
| 12:35 | <annevk> | zcorpan: is it hard to get a list of URLs along with http://simon.html5.org/dump/encoding-labels/label-count.txt ? |
| 12:38 | <annevk> | AryehGregor++ for https://bugzilla.mozilla.org/show_bug.cgi?id=702948 |
| 12:39 | <annevk> | AryehGregor: might be interesting for NodeIterator too |
| 12:41 | <zcorpan> | annevk: shouldn't be too hard |
| 12:54 | <gsnedders> | Does it make me a bad computer scientist to end up hating thinking about data structures, unsure what to use? |
| 12:58 | <annevk> | gsnedders: just means you need more practice :) |
| 12:58 | <annevk> | gsnedders: fwiw, I have that each time I write scripts for research |
| 12:59 | <Philip`> | gsnedders: A good computer scientist would just list the big-O complexity of every possible data structure and pick whichever one is smallest |
| 13:00 | <jgraham> | And then use it on sets of 2 items |
| 13:01 | <gsnedders> | Philip`: Ah, I'm down to everything being O(n), but with vastly different perf in Python (although fairly similar in C). |
| 13:02 | <jgraham> | hsivonen: You might like to look at http://w3c-test.org/html/tests/submission/Opera/documentwrite/ Not that Gecko fails (m)any (although some seem to fail if you run them in an iframe, at least on my system, although it could be a cache issue after I fixed some bugs in the tests) |
| 13:04 | <Philip`> | gsnedders: A good computer scientist would ignore constant factors - they're just a trivial implementation detail |
| 13:05 | <gsnedders> | Philip`: :) |
| 13:05 | <gsnedders> | Philip`: Problem I think is the big-O complexity of Python built-ins… |
| 13:05 | <jgraham> | So… you didn't actually make everything O(N)? |
| 13:06 | <jgraham> | Pretty sure not knowing what data structure/algorithm you are *actually* using makes you a bad computer scientist, yes :) |
| 13:07 | <gsnedders> | jgraham: I think not :( |
| 13:07 | <gsnedders> | Working out the complexity of algorithms where you don't know what the complexity of operations on the interpreter builtins are is hard. |
| 13:07 | <gsnedders> | Esp. because it's an impl-detail of the interpreter. |
| 13:07 | <jgraham> | That is kind of silly |
| 13:07 | <gsnedders> | Okay, so I can probably care most about CPython, but still. |
| 13:08 | <jgraham> | Clojure has the right idea |
| 13:08 | <gsnedders> | jgraham: ? |
| 13:08 | <Philip`> | gsnedders: By your logic, surely any algorithm could be described as O(1), just with the problem that it's a constant number of built-in operations that don't have constant cost themselves |
| 13:08 | <gsnedders> | Philip`: Yes. |
| 13:09 | <gsnedders> | (Provided any problem > O(1) has a solution as a built-in) |
| 13:09 | <gsnedders> | (or sequence thereof) |
| 13:09 | <gsnedders> | (which surely isn't certain?) |
| 13:09 | <jgraham> | gsnedders: The built-in data structures in clojure all have a stated complexity for each operation they support |
| 13:09 | <gsnedders> | jgraham: Ah. |
| 13:09 | <Philip`> | (You can just add a built-in solveAlgorithm582() function to your language) |
| 13:10 | <jgraham> | That is the only fact I know about clojure |
| 13:10 | <jgraham> | Well apart from the fact that it is a lisp that runs on the jvm |
| 13:10 | <Philip`> | C++ STL typically states complexity too, which is nice |
| 13:10 | <jgraham> | and that it looks like it is quite fun |
| 13:10 | <jgraham> | On the other hand C++ STL only just got a hash table type |
| 13:12 | <Philip`> | There's been generally compatible vendor-specific ones for years, and Boost's one that I think the standard got based on, so that's a mostly theoretical problem if you don't care about extreme portability |
| 13:13 | <gsnedders> | (And if you care about extreme portability, plenty of embedded devices don't have the STL anyway) |
| 13:13 | <jgraham> | OTOH as a C++ beginner, none of that is at all obvious |
| 13:14 | <jgraham> | Well itg is pretty obvious that someone would have implemented hash tables somewhere |
| 13:15 | Philip` | likes that boost::unordered_map lets you do lookups using a key object that's not the same as the unordered_map's actual key type (assuming compatible hash and equality functions) |
| 13:15 | <jgraham> | … why? |
| 13:16 | <Philip`> | It means you can e.g. have a std::string key type (which does automatic memory management), but do lookups using a char*, so you don't have to do an allocation to construct the key for every lookup |
| 13:16 | <jgraham> | Oh, that's quite neat |
| 13:16 | <jgraham> | Or really stupid, hard to tell which :) |
| 13:18 | <jgraham> | Presumably you don't just use a std::string* for both because you want the hash table to own the strings, or something? |
| 13:18 | <gsnedders> | I guess theoretically the complexity of operations could vary between CPU architectures, too, but that's probably caring too much. |
| 13:18 | <annevk> | foolip: btw, I can volunteer for writing that algorithm in Python |
| 13:18 | <annevk> | foolip: not Java |
| 13:20 | <Philip`> | jgraham: Yeah, you don't want to have to manually free pointers inside complex data structures, because that usually becomes an ugly mess and will leak everywhere |
| 13:21 | <Philip`> | (whereas using std::string means it'll automatically deal with all the copying and freeing, and would only be noticeably inefficient if you're modifying the structure heavily and causing a load of copies, and isn't a problem if you're primarily doing lookups) |
| 13:35 | <zcorpan> | annevk: http://simon.html5.org/dump/encoding-labels/labels-urls.txt |
| 13:59 | <zcorpan> | first interesting url, cp1251, http://www.b2blogger.com/pressroom/tag/%F7%E0%F1%F2%ED%FB%E9%20%EA%EB%E8%E5%ED%F2 |
| 14:05 | <annevk> | zewt: lower boundary does not work with interval I think |
| 14:05 | <annevk> | zewt: as lower boundary can be greater than 0x10FFFF |
| 14:05 | <annevk> | zewt: also, where is this interval syntax defined? |
| 14:07 | <zewt> | where is the phrase "greater than" defined? :) |
| 14:08 | <zewt> | [0x200000,0x10FFFF] works, and is empty |
| 14:09 | <zewt> | probably easy enough to define (havn't looked through the other encoders/decoders much so I don't know if there's enough cases for it to be worth it) |
| 14:16 | <matjas> | what does the TR in W3C spec URLs stand for? e.g. http://www.w3.org/TR/CSS2/ |
| 14:18 | <zcorpan> | Technical Report |
| 14:19 | <zcorpan> | or at least i think it used to, now TR/ says "All Standards and Drafts" |
| 14:20 | <annevk> | yup |
| 14:20 | <annevk> | zewt: there's quite a few of those, especially with bytes |
| 14:20 | <annevk> | zewt: just search for "to 0x" |
| 14:21 | <annevk> | zewt: still a bit skeptical since I don't like how that looks, but maybe |
| 14:21 | <annevk> | zewt: also, "greater than" is way different from [0x40,0x50] :) |
| 14:25 | <MikeSmith> | jgraham, annevk : please remind me how to fix this anolis problem |
| 14:25 | <MikeSmith> | anolis --dump-xrefs=data/xrefs/dom/crypto.json Overview.src.html /tmp/spec |
| 14:25 | <MikeSmith> | [Errno 2] No such file or directory: 'data/xrefs/dom/crypto.json' |
| 14:26 | <matjas> | zcorpan, annevk: ty |
| 14:26 | <MikeSmith> | I thought what --dump-xrefs=data/xrefs/dom/crypto.json does is to create the data/xrefs/dom/crypto.json file |
| 14:26 | <MikeSmith> | gsnedders: ↑ |
| 14:28 | <gsnedders> | MikeSmith: No idea, I have nothing to do with that functionality. |
| 14:28 | <MikeSmith> | ok |
| 14:31 | <jgraham> | MikeSmith: Ask Ms2ger |
| 14:32 | <annevk> | MikeSmith: make a crypto.json file consisting of {} |
| 14:32 | <annevk> | MikeSmith: it doesn't create the file, it just fills it up |
| 14:34 | <jgraham> | That seems quite silly |
| 14:38 | <zcorpan> | http://www.rediff.com/gujarati/2002/apr/19dalal.htm |
| 14:44 | <MikeSmith> | annevk: thanks |
| 14:44 | <MikeSmith> | jgraham: I think it merits some stronger word than silly |
| 14:46 | gsnedders | wonders whether requiring UTF-8 for text/* over HTTP/2.0 is viable, or just as fictional as text/* being ISO-8859-1. |
| 14:46 | <gsnedders> | Certainly moving away from any content-type sniffing for SPDY content would be desirable. |
| 14:47 | <gsnedders> | (Whether a Content-Type header is used or magic bytes instead is a separate issue) |
| 14:48 | <jgraham> | MikeSmith: The W3C testrunner tries to run tests in an iframe, right? |
| 14:49 | <dglazkov> | good morning, Whatwg! |
| 14:50 | <MikeSmith> | jgraham: it was using an object before |
| 14:50 | <jgraham> | MikeSmith: I thought people objected to that? |
| 14:50 | <MikeSmith> | I can't remember if it was updated to use an iframe or nt |
| 14:50 | <MikeSmith> | jgraham: yeah, it causes problems on mobile |
| 14:50 | <MikeSmith> | you can't scroll |
| 14:51 | <jgraham> | Anyway I object to iframe on the basis that testing window.top doesn't really work in an iframe (for example) |
| 14:52 | <jgraham> | Basically I think this whole harness might be a world of hurt |
| 14:52 | <jgraham> | Of course it's not strictly relevant to me since we won't ever use it |
| 14:52 | <jgraham> | But I will be annoyed if people get upset when tests break in it |
| 14:52 | <MikeSmith> | I checked and it's still using <object> |
| 14:52 | <MikeSmith> | jgraham: who's we? |
| 14:52 | <jgraham> | OK |
| 14:53 | <MikeSmith> | and why won't you use it? |
| 14:53 | <jgraham> | Opera. But I would expect "browser vendors" |
| 14:53 | <MikeSmith> | oh |
| 14:53 | <MikeSmith> | yeah |
| 14:53 | <jgraham> | Because we have a highly custom solution designed for our specific requirements |
| 14:53 | <MikeSmith> | yeah |
| 14:53 | <MikeSmith> | understood |
| 14:54 | <jgraham> | (I guess you know that, but one must always think of the people trawling the logs) |
| 14:54 | <MikeSmith> | heh |
| 14:54 | <MikeSmith> | I think people trawling the logs should find some better way to spend their time |
| 14:54 | <annevk> | MikeSmith: jgraham: I'm sure Ms2ger takes patches |
| 14:55 | <annevk> | it does prevent people from creating files they didn't want to, and given that it's only for the first time, it's not too bad |
| 14:55 | <Philip`> | MikeSmith: Indeed, reading the channel's logs is almost as bad as reading the channel live |
| 15:03 | <MikeSmith> | Philip`: indeed |
| 15:03 | <MikeSmith> | annevk: I wonder where the name big5 came from |
| 15:03 | <MikeSmith> | I mean, why name it that? |
| 15:04 | <jgraham> | Size matters? |
| 15:05 | <zcorpan> | MikeSmith: https://twitter.com/#!/brucel/status/189664480391741440 |
| 15:05 | <Ms2ger> | brucel++ |
| 15:06 | <annevk> | 'The name "Big5" is in recognition that the standard emerged from collaboration of five of Taiwan's largest IT firms: Acer (宏碁); MiTAC (神通); JiaJia (佳佳), ZERO ONE Technology (零壹 or 01tech); and, First International Computer (FIC) (大眾).' |
| 15:06 | <annevk> | apparently not quite enough collaboration |
| 15:06 | <annevk> | at least not when it came to extending it |
| 15:07 | <MikeSmith> | heh |
| 15:07 | <zcorpan> | they should have made the label 宏碁-神通-佳佳-零壹-大眾 |
| 15:07 | <zcorpan> | (encoded as big5) |
| 15:08 | <jgraham> | So basically it was an ego thing? |
| 15:08 | <jgraham> | Impressive |
| 15:08 | <zcorpan> | (sorry, i mean, encoded as 宏碁-神通-佳佳-零壹-大眾) |
| 15:09 | <MikeSmith> | zcorpan: maybe they should have named it secret7 |
| 15:09 | <MikeSmith> | brucel maybe likes that one better |
| 15:12 | <zcorpan> | good idea for next year's april 1 |
| 15:12 | <jgraham> | I think of brucel as a maloryTowers man |
| 15:22 | <annevk> | zewt: I like the emit encoder/decoder error idea btw |
| 15:22 | <annevk> | zewt: that's quite neat |
| 15:23 | <annevk> | zewt: I didn't really like the current approach either |
| 15:23 | <annevk> | euh don't |
| 16:03 | <annevk> | MikeSmith: ah, the other reason it's not automatically generated is that you need to add a "url" field to crypto.json |
| 16:03 | <annevk> | MikeSmith: otherwise other specs can't use it |
| 16:03 | <MikeSmith> | oh |
| 16:04 | <annevk> | the xspec xref is a concatenation of that URL and the term |
| 16:04 | <MikeSmith> | but it seems like that could just be specified as an additional param |
| 16:05 | <annevk> | in the makefile? |
| 16:05 | <MikeSmith> | to anolis |
| 16:05 | <MikeSmith> | additional switch |
| 16:05 | <annevk> | i guess that could somehow be extracted too, yeah |
| 16:08 | <Ms2ger> | MikeSmith, yeah, patch welcome :) |
| 16:08 | <MikeSmith> | can't be arsed |
| 16:09 | <MikeSmith> | I prefer to complain |
| 16:09 | <MikeSmith> | seriously, it's not heinous thing |
| 16:09 | <MikeSmith> | just hard to remember |
| 16:09 | <Ms2ger> | Yeah |
| 16:09 | <MikeSmith> | I think this is the 3rd time I've need to re-remember it |
| 16:09 | <Ms2ger> | 3rd time you added a spec? :) |
| 16:09 | <MikeSmith> | yeah |
| 16:10 | <MikeSmith> | I guess so |
| 16:10 | <MikeSmith> | in this case it was trying to help the domcrypt guys get their stuff working |
| 16:10 | <Ms2ger> | Didn't I add that? |
| 16:10 | <MikeSmith> | dunno |
| 16:10 | <MikeSmith> | but don't think so |
| 16:10 | <MikeSmith> | because it was failing at the point where it was looking for that json file |
| 16:11 | <MikeSmith> | unless you put it in a different subdir |
| 16:14 | <Ms2ger> | Great news, gsnedders stopped being dumb :) |
| 16:14 | <gsnedders> | Ms2ger: Esp. considering I have exams next week! |
| 16:15 | <arunranga> | annevk, should HTML also define what a microtask is, in addition to "perform a microtask checkpoint"? That allows other specs (like File) to make some conditions only valid within a microtask. |
| 16:22 | <annevk> | arunranga: if you need a new concept from HTML related to task I'd suggest filing a bug on HTML with the requirements |
| 16:23 | <annevk> | tasks* |
| 16:23 | <annevk> | arunranga: especially if you need some kind of hook |
| 16:23 | <annevk> | arunranga: it seems though for the resolve URL thing we might need a bit more |
| 16:24 | <arunranga> | annevk: ok. I guess I wanted you (and Ms2ger's) take from a DOM perspective. |
| 16:24 | <arunranga> | annevk: yes, I'm trying to figure out *what* we might need. |
| 16:24 | <Ms2ger> | Ah, it's arunranga! |
| 16:24 | <arunranga> | Hello Ms2ger :) Formerly otherarun. Also, TheArun. |
| 16:25 | <Ms2ger> | Microtasks? 300 smaug____ annevk aklein |
| 16:25 | <annevk> | maha 301 hixie |
| 16:25 | <annevk> | I should understand this stuff, but each time it dazzles me |
| 16:26 | <arunranga> | All this redirection is giving me a time out. |
| 16:26 | <arunranga> | annevk, if by "dazzles" you mean "totally confuses me to the point of tears" I'm in the same boat as you. |
| 16:26 | <Ms2ger> | I decided against 301 because I still have hope I'll figure it out at some point :) |
| 16:26 | <Ms2ger> | The reason for that, of course, is that I haven't looked at it |
| 16:27 | <smaug____> | what about microtasks ? |
| 16:27 | <smaug____> | ah, some spec things... |
| 16:27 | <annevk> | arunranga: haha, it's not quite that bad, but the resolving URLs concept is quite confusing |
| 16:27 | <Ms2ger> | s/resolving// |
| 16:28 | <arunranga> | Greetings smaug___ ! Currently, HTML specifies "perform a microtask checkpoint" without an explanation of what constitutes a microtask. I think a microtask is a convenient concept upon which to hang a few other checks. |
| 16:28 | <arunranga> | smaug___, Ms2ger, annevk: this includes the validity of a Blob URL. I'd like them to be scoped within a microtask. |
| 16:28 | <arunranga> | (if possible) |
| 16:28 | <Ms2ger> | Blob URLs? |
| 16:29 | Ms2ger | runs |
| 16:29 | <Ms2ger> | zewt cares about that, IIRC? |
| 16:29 | arunranga | catches Ms2ger and forces the issue :) |
| 16:29 | <Ms2ger> | r- |
| 16:30 | <Ms2ger> | MikeSmith, I'm going to see if I can add a better error message for the --dump-xrefs thing |
| 16:30 | <MikeSmith> | OK |
| 16:30 | <MikeSmith> | that would be helpful |
| 16:30 | <arunranga> | zewt and I chatted a while ago about resolution of Blob URLs. FileAPI currently uses the confusing "dereference…" which I think should be replaced. |
| 16:31 | <smaug____> | arunranga: yeah, HTML spec could perhaps specify better what is a microtask. (The outermost script execution of the innermost task) |
| 16:32 | <arunranga> | smaug___ I'll file a bug I think. |
| 16:33 | <arunranga> | We've got performing a microtask checkpoint defined in terms of MutationObserver. |
| 16:34 | <smaug____> | well, because that is the original reason for microtasks |
| 16:34 | <smaug____> | sicking was thinking to use microtasks IDB |
| 16:36 | <arunranga> | I think we can use them for File as well, maybe for validity of Blob URLs. |
| 16:38 | <Ms2ger> | jgraham, MikeSmith, fwiw, Mozilla runs mochitests in an iframe too, so I don't think there would be an objection from our side |
| 16:38 | <annevk> | dglazkov: xmp is not a valid element fwiw |
| 16:38 | <Ms2ger> | jgraham, MikeSmith, I think we use popups, and disable popup-blocking, if necessary |
| 16:39 | <annevk> | arunranga: validity of blob URLs doesn't need to happen in a task I think |
| 16:39 | <gsnedders> | Anyone have IE10 at hand to see if const is supported in ES? |
| 16:39 | <arunranga> | annevk: What I mean is, *scoped* to a microtask. |
| 16:40 | <arunranga> | annevk: this makes leaks less likely. |
| 16:40 | <Ms2ger> | gsnedders, I can if you give me a test |
| 16:40 | <annevk> | ooh, the expiring stuff |
| 16:40 | <annevk> | right |
| 16:40 | <arunranga> | annevk: that is, having unused Blob URLs or "long lived" ones. Also, it mitigates the need for something like oneTimeOnly. |
| 16:40 | <Ms2ger> | Preferably by uploading it to the live dom viewer |
| 16:40 | <annevk> | arunranga: yeah I remember reading that, I think zewt was all on top of that :) |
| 16:41 | <gsnedders> | Ms2ger: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1480 |
| 16:41 | <annevk> | but part of that sounded like it needed to be defined in either URL resolving or fetching |
| 16:41 | <arunranga> | annevk: won't exactly stop race issues, but can mitigate them. |
| 16:41 | <gsnedders> | Ms2ger: Basically, does it give an error or not? |
| 16:41 | <annevk> | and then XHR and everything dealing with URLs needed updating |
| 16:41 | <arunranga> | annevk, right. Are those the purview of HTML? |
| 16:42 | arunranga | concludes it is. |
| 16:43 | <Ms2ger> | gsnedders, Syntax error |
| 16:43 | <gsnedders> | Ms2ger: Interesting. |
| 16:43 | <dglazkov> | annevk: ok :) |
| 16:45 | jgraham | tries to test some simple things, finds a selection of unrelated issues |
| 16:45 | <jgraham> | In other news, I am starting to understand why hsivonen hates about:blank |
| 16:46 | <jgraham> | Ms2ger: We run some tests in popup windows and some tests (including testharness.js tests) as top level documents |
| 16:46 | <smaug____> | that happens in Mozilla too |
| 16:46 | <smaug____> | some tests open a new window |
| 16:47 | <smaug____> | or new tab etc |
| 16:47 | <smaug____> | but most are run in an iframe |
| 16:47 | <jgraham> | Right, your tests can have chrome permissions |
| 16:47 | <jgraham> | Whereas we drive the tests externally so we can load them in a top-level window |
| 16:47 | <jgraham> | aiui |
| 16:48 | <jgraham> | Well I mean I understand the Opera system. I think I understand what Mozilla do :) |
| 16:48 | <Ms2ger> | Plain mochitests are run in an iframe and don't get chrome permissions |
| 16:49 | <Ms2ger> | Modulo SpecialPowers, which is an extension that grants access to some stuff, and UniversalXPConnect, which we want to kill |
| 16:49 | <annevk> | arunranga: URLs and fetch are at this point |
| 16:49 | <Ms2ger> | There are also chrome mochitests, which are run in a top-level window, and do have chrome permissions |
| 16:49 | <Ms2ger> | And other stuff |
| 16:50 | <annevk> | dglazkov: I guess what I should say, is that we have not introduced any new features in the HTML syntax that depart wildly from what is possible in XML |
| 16:50 | <jgraham> | Do SpecialPowers allow you to open a top-level browsuing context if you need it? |
| 16:50 | <Ms2ger> | That's what window.open is for |
| 16:51 | <annevk> | dglazkov: so making such a proposal can be somewhat controversial and also begs the question of how you anticipate <template> to work in a namespaced context |
| 16:52 | <jgraham> | Fair point. I always forget that window.open gives a top-level browsing context |
| 16:55 | <dglazkov> | annevk: I haven't been keeping XML and namespaces in mind at all. It's a good point -- though I have no sense whether XML is just an edge case at this point. |
| 17:00 | <annevk> | dglazkov: I think it is, and I don't necessarily think we need to cater to it, but unless we remove XML from browsers it needs to be defined what <template> does in that context |
| 17:00 | <annevk> | dglazkov: and it's also important in the DOM |
| 17:00 | <annevk> | dglazkov: if appendChild() something to <template> you'd have the same thing |
| 17:01 | <dglazkov> | annevk: that last one doesn't seem much different from appending a child to textarea |
| 17:06 | <dglazkov> | annevk: can you help rafaelw_ understand the issue with XML? |
| 17:07 | <rafaelw_> | annevk, hsivonen, dglazkov: just joining... |
| 17:07 | <rafaelw_> | annevk: if you would be so kind ^^^ |
| 17:07 | <annevk> | the issue is that this feature would not work in XML |
| 17:07 | <annevk> | just like <noscript> I suppose |
| 17:07 | <annevk> | if that's by design, it probably should state that explicitly so people who get annoyed by that kind of thing get annoyed early on |
| 17:08 | <annevk> | I don't really care |
| 17:09 | <rafaelw_> | honestly, this working in XML hadn't even occurred to me. |
| 17:09 | <rafaelw_> | but can you help me understand why it doesn't work. |
| 17:15 | <annevk> | rafaelw_: because you'd have to change the XML parser if you want it to work in the same way |
| 17:15 | <annevk> | rafaelw_: I'm assuming you're not planning on doing that |
| 17:15 | <annevk> | oh shit, I should have dinner |
| 17:17 | <rafaelw_> | hmm... that kinda makes it sound like this isn't so much a problem with XML as a question about whether the <template> element should have the same behavior in an XML document. is that a fair way of stating it? |
| 17:18 | <annevk> | I don't think so, because the XML situation also arises in the DOM |
| 17:19 | <annevk> | <template><test/></template> is not much different from createElement("template"); createElement("test"); template.append(test); body.append(template) |
| 17:19 | <annevk> | whereas in HTML per your suggestion <template><test/></template> would give <template/> in the DOM |
| 17:20 | <annevk> | which is "magic" in the HTML parser that we cannot introduce in the XML parser |
| 17:26 | <rafaelw_> | (adam is explaining to me locally). so is that the we have the ability to change the HTML parser, but no such ability with the XML parser? |
| 17:26 | <annevk> | right, changing those kind of invariants of the XML parser is not going to fly |
| 17:36 | <TabAtkins> | gsnedders: I can't decipher what you are attempting to say in your post about "legacy const". This isn't helped by your opening sentence being ungrammatical and nonsensical. |
| 17:44 | <gsnedders> | TabAtkins: Yay for not reading emails before I send them! |
| 17:46 | <rniwa> | arv: http://www.whatwg.org/specs/web-apps/current-work/multipage/editing.html#the-inert-attribute |
| 17:46 | <Ms2ger> | rniwa, do you know if anybody else than webkit does something as complex as http://trac.webkit.org/browser/trunk/Source/WebCore/css/CSSStyleSelector.cpp?rev=113922#L4920 ? |
| 17:46 | <annevk> | TabAtkins is correct though that insert="" is not suitable for <template>; my bad |
| 17:47 | <annevk> | maybe insert="" should be named inactive="" |
| 17:47 | <gsnedders> | TabAtkins: In related news, why does Standard English not contain "outwith" in its lexis? Grr! |
| 17:47 | <TabAtkins> | What does "outwith" mean? |
| 17:48 | <Ms2ger> | gsnedders, I would ask an English language majo... Oh |
| 17:48 | <rniwa> | Ms2ger: as far as I know, Gecko does something similar |
| 17:48 | <Ms2ger> | rniwa, including the mode-dependence? |
| 17:48 | <rniwa> | Ms2ger: think so |
| 17:49 | <Ms2ger> | Because I didn't see that in the code |
| 17:49 | <annevk> | "is a Scottish preposition meaning 'outside, beyond', and along with wee (for 'small') is the first non-English English word that visitors to Scotland notice (and sometimes adopt): Do you live outwith the city? / We can discuss that outwith the meeting." |
| 17:49 | <annevk> | brilliant |
| 17:49 | <rniwa> | Ms2ger: I swear I even checked gecko's implementation at one time |
| 17:49 | <rniwa> | Ms2ger: okay, can I see your code? |
| 17:49 | <gsnedders> | TabAtkins: Opposite of within, basically |
| 17:49 | <rniwa> | Ms2ger: lost my URL :( |
| 17:49 | <Ms2ger> | rniwa, I may have changed it since :) |
| 17:49 | <rniwa> | Ms2ger: oh |
| 17:49 | <rniwa> | Ms2ger: where can I see the code in Gecko though? |
| 17:50 | <TabAtkins> | gsnedders: That's just silly. |
| 17:50 | <gsnedders> | TabAtkins: so "outwith the meeting" means "not within the meeting", basically |
| 17:50 | <TabAtkins> | "outside of" |
| 17:50 | <Ms2ger> | http://mxr.mozilla.org/mozilla-central/source/content/html/content/src/nsHTMLFontElement.cpp , right? |
| 17:50 | <gsnedders> | TabAtkins: Bah, ScEn has a single word for it! |
| 17:50 | <Ms2ger> | Oh, wait |
| 17:50 | <gsnedders> | (I've undoubtedly used this a number of times on W3C mailing lists, as I often forget it's a ScEn word) |
| 17:51 | <annevk> | gsnedders: does Scotland have a two-level code? |
| 17:51 | <annevk> | en-sc or some such |
| 17:51 | <gsnedders> | annevk: No. en-gb-scot. |
| 17:51 | <annevk> | ugh |
| 17:51 | <gsnedders> | Oh, no, that's untrue. |
| 17:51 | <gsnedders> | Only Google hit for that is a Tweet by me. Must be misremebering. |
| 17:52 | <Ms2ger> | rniwa, sorry, I think I lied |
| 17:52 | <Ms2ger> | Gah. |
| 17:52 | <rniwa> | Ms2ger: ? |
| 17:52 | <Ms2ger> | http://mxr.mozilla.org/mozilla-central/source/layout/style/nsStyleUtil.cpp#64 |
| 17:52 | <annevk> | gsnedders: sco per ISO |
| 17:52 | <gsnedders> | annevk: sco |
| 17:53 | <rniwa> | Ms2ger: yup... |
| 17:53 | <rniwa> | Ms2ger: i think dhyatt wrote both tables |
| 17:53 | <Ms2ger> | And got some off-by-one errors in yours? :) |
| 17:53 | <gsnedders> | annevk: Though that implies it's the Scots Language and not a dialect of English (which is a very interesting discussion…) |
| 17:53 | <rniwa> | Ms2ger: maybe LOL |
| 17:54 | <rniwa> | Ms2ger: anyway, i've got to head to Mozilla's SF office now |
| 17:54 | <annevk> | Ms2ger: broken links in your source code |
| 17:54 | <rniwa> | so ttyl ;) |
| 17:54 | <Ms2ger> | Though I see erik⊙nc has blame for that table |
| 17:54 | gsnedders | also finds that Old English and Middle English have their own language codes, interestingly |
| 17:54 | <Ms2ger> | rniwa, won't find me there, sorry ;) |
| 17:54 | <rniwa> | Ms2ger: i know ;) |
| 17:55 | <gsnedders> | annevk: sco is really the Scots Language as opposed to Scottish English, and I'm really meaning the latter here. |
| 17:55 | <Ms2ger> | annevk, patches welcome :) |
| 17:55 | <Ms2ger> | "Bug 18136 - Fixing the font size mess {font} {ll}" |
| 17:55 | <gsnedders> | "Whaur dae ye bide?" would be Scots but not Scottish English, for example |
| 17:55 | <annevk> | Ms2ger: been ages since I patched Gecko |
| 17:55 | <annevk> | Ms2ger: and I was rather bad at it back then |
| 17:55 | <TabAtkins> | gsnedders: Guessing "Where do you live?" |
| 17:56 | <Ms2ger> | annevk, surely Opera has trained you by now :) |
| 17:56 | <annevk> | in the Gecko ways? not so much |
| 17:56 | <gsnedders> | TabAtkins: Yes. |
| 17:56 | <hober> | scots gaelic is gd iirc |
| 17:56 | <Ms2ger> | Well, I guess the Gecko ways are about as weird as Opera's |
| 17:56 | <hober> | i don't remember the three letter code for it |
| 17:57 | <gsnedders> | hober: yeah |
| 17:57 | <gsnedders> | hober: gla |
| 17:57 | <rektide> | anyone have suggestions for how do to drag and drop SVG? my problem is that i'm dragging the item along the screen under the mouse, covering up all the event handlers i would like to be firing |
| 17:57 | <hober> | sco is the sister-to-english language spoken in scotland :) |
| 17:57 | <rektide> | and elementFromPoint identifies not <svg:use/> elements but what the <svg:use/> elements point to |
| 17:57 | <gsnedders> | hober: Ever less so, though. Scottish English is probably about as common as Scots in colloquial usage now. |
| 17:58 | <gsnedders> | (Gaelic is practically non-existant for most purposes, and has had very little influence over Scottish English or Scots) |
| 17:58 | <rektide> | i'm interested to know what techniques are out there for making the MouseEvents useful, not all target this one being-dragged item |
| 17:59 | <othermaciej> | wikipedia sea: "Scottish English refers to the varieties of English spoken in Scotland. It may or may not be considered distinct from the Scots language." |
| 17:59 | <othermaciej> | is that wrong? |
| 18:00 | <Ms2ger> | [citation needed] |
| 18:00 | <Ms2ger> | rniwa (~rniwa@nat/mozilla/x-rkjncmdkxlybjmcu) has joined #whatwg |
| 18:00 | <Ms2ger> | We got him! |
| 18:00 | <rniwa> | Ms2ger: LOL. |
| 18:00 | <rniwa> | Ms2ger: SF goole office & SF mozilla office are in the same building complex |
| 18:00 | <shepazu> | rektide: turn off pointer events on the svg element you're dragging |
| 18:00 | <gsnedders> | othermaciej: You can argue either way. The academic view tends to be that language in Scotland is a continuum from Standard English through Scottish English to Scots, but still not entirely agreed upon. |
| 18:01 | <Ms2ger> | rniwa, remind me never to set foot in that complex, then :) |
| 18:01 | <MikeSmith> | othermaciej: http://en.wikipedia.org/wiki/Scots_language "It is sometimes called Lowland Scots to distinguish it from Scottish Gaelic, the Celtic language variety spoken in most of the western Highlands and in the Hebrides." |
| 18:01 | <gsnedders> | othermaciej: Given that it's basically a continuum, there's some debate as to whether Scots is a separate language from English or not. |
| 18:01 | <smaug____> | Ms2ger: :) |
| 18:01 | <smaug____> | Ms2ger: SF isn't that bad place |
| 18:01 | <rektide> | shepazu: thank you so much. of course. |
| 18:01 | <shepazu> | np |
| 18:01 | <rektide> | NOT ANY MORE |
| 18:02 | <Ms2ger> | smaug____, I thought you would be in favour of me not joining Google... Unless you saw all my regressions :) |
| 18:02 | <rektide> | omg the hackery that's been going on for a day and a half |
| 18:02 | <jgraham> | Uh, wait how did I manage to entirely overlook something like <template><test/></template> == <template></template>? |
| 18:02 | <othermaciej> | MikeSmith: I do know that Scottish Gaelic is distinct (and only spoken by 1% of the population of Scottland or so) |
| 18:02 | <rektide> | ;) |
| 18:02 | <jgraham> | That's not going to work |
| 18:02 | <gsnedders> | othermaciej: Go back to 1707 (Acts of Union, where the Scottish and English parliments effectly ceased to exist to be replaced by a single parliment), Scots was a separate language to English. |
| 18:02 | <shepazu> | rektide: remember to turn them back on when you drop the element :) |
| 18:02 | <smaug____> | Ms2ger: you could join Mozilla ;) |
| 18:03 | <smaug____> | as an employee |
| 18:03 | <jgraham> | gsnedders: Is the Scottish plan to leave the union and then fund independence by suing Linux for IP violations? |
| 18:03 | <othermaciej> | gsnedders: are you scottish yourself? |
| 18:03 | smaug____ | can hope someone drags rniwa out from the dark side :p |
| 18:03 | <MikeSmith> | wow, Ulster Scots? wtf |
| 18:03 | <MikeSmith> | othermaciej: you met gsnedders man |
| 18:03 | <othermaciej> | MikeSmith: I'm borderline autistic, I don't remember stuff like that :-) |
| 18:04 | <MikeSmith> | heh |
| 18:04 | <jgraham> | gsnedders: Is like the opposite of someone who plays Scottish on TV |
| 18:04 | <Ms2ger> | Glasgow? |
| 18:04 | <MikeSmith> | gsnedders looks like a young Robert Plant |
| 18:04 | <gsnedders> | othermaciej: Yes. |
| 18:04 | <MikeSmith> | except sexier |
| 18:04 | <gsnedders> | MikeSmith: Also, I have never met othermaciej. |
| 18:04 | <jgraham> | He was born is Scotland but looks/sounds about as Scottish as I do |
| 18:04 | <gsnedders> | But what jgraham said is the truth. |
| 18:04 | <jgraham> | But with a slight resembelence to a heroin addict |
| 18:05 | <jgraham> | Which is quite glaswegian at least |
| 18:05 | <gsnedders> | (Also, jgraham has more Scottish ancestory than I do, IIRC) |
| 18:05 | <jgraham> | I never understood how one did sums on ancestors |
| 18:05 | <MikeSmith> | gsnedders has a unique sound, just as Robert Plant does |
| 18:05 | <othermaciej> | I am not very good at distinguishing UK/commonwealth accents |
| 18:05 | <jgraham> | s/did/does/ |
| 18:05 | <gsnedders> | My accent is approximately RP. |
| 18:05 | <jgraham> | I think I have more Scottish parents than you |
| 18:06 | <gsnedders> | jgraham: Well, yes, that's what I was going by basically. |
| 18:06 | <jgraham> | I'm not sure how one weights backward in time |
| 18:06 | <gsnedders> | (I have none.) |
| 18:06 | <MikeSmith> | gsnedders, othermaciej : I thought you dudes met in Nice but I guess slacker ot hermaciej didn't show that year |
| 18:06 | <jgraham> | Although I am pretty sure you do it differently in America |
| 18:06 | <gsnedders> | (I have a single Scottish grandparent.) |
| 18:06 | <jgraham> | Like you find one ancestor in a country you like and claim to be from there |
| 18:06 | <gsnedders> | MikeSmith: Yeah, that's what happened. |
| 18:09 | <MikeSmith> | my ancestors were mostly pig thieves who fled to America to avoid getting branded with whatever tf the English bastards did to honest pig thieves at that time |
| 18:10 | <MikeSmith> | plus French Canadians |
| 18:10 | <Ms2ger> | Just your ancestors? |
| 18:10 | <MikeSmith> | I've stolen a pig or two in my time |
| 18:10 | <gsnedders> | http://en.wikipedia.org/wiki/Middle_Scots#Sample_text is a good example of what Scots was like a few hundred years ago |
| 18:11 | <MikeSmith> | plus I've "had" a pig or two i mine time |
| 18:11 | <gsnedders> | (Compare with, for example, Shakespeare in first folio form — modern editions are fairly different from the original language) |
| 18:11 | <gsnedders> | Obviously related to English, but really quite radically different to it. |
| 18:13 | <MikeSmith> | modern glasgow scottish is one of the greats in terms of its utter incomprehensibility |
| 18:13 | <gsnedders> | Heheh. It's not that bad! |
| 18:13 | <gsnedders> | (I didn't find it that bad when I first moved to Glasgow.) |
| 18:14 | <MikeSmith> | glasgow scots are basically barbarians who have barely mastered speech |
| 18:14 | <MikeSmith> | pretty sure some thug is now going to kill me for saying that |
| 18:17 | <annevk> | rafaelw_: is the idea that the HTML parser will define the parsing algorithm for <template>? |
| 18:18 | <annevk> | rafaelw_: because if that's the case, and that would make sense to me, it might be worth filing a bug against HTML at this point too, with some explanation of the goal |
| 18:19 | <annevk> | rafaelw_: so the HTML WG is not surprised and we don't get useless revert requests that hober then has to waste time on |
| 18:19 | <annevk> | (I guess the alternative is to buy hober even more beer, but I'm not sure that's wise) |
| 18:22 | <rafaelw_> | annevk: is there alternative to that? |
| 18:22 | <hober> | annevk: :) |
| 18:23 | <rafaelw_> | (that was certainly my idea) |
| 18:27 | <annevk> | rafaelw_: k |
| 18:29 | <MikeSmith> | kickstarter for hober beer |
| 18:30 | <annevk> | jsbell: "// TODO: Typo in spec?" is fixed |
| 18:30 | <annevk> | jsbell: noticed it myself, then went to check your code, then found that :) |
| 18:31 | <MikeSmith> | amazing |
| 18:32 | <MikeSmith> | my OSX software update actually updated software without requiring me to restart |
| 18:32 | <rafaelw_> | annevk: https://www.w3.org/Bugs/Public/show_bug.cgi?id=16789 |
| 18:32 | <MikeSmith> | there is hope |
| 18:32 | <annevk> | rafaelw_: cool, fixed the component so the email will go to the HTML WG |
| 18:45 | <annevk> | hober: maybe we can meet April 30 for Notifications? |
| 18:46 | <hober> | annevk: sure. want to come to apple for lunch? |
| 18:46 | <annevk> | hober: apparently doing it at the Microsoft venue is not really okay... although I'm not sure anyone actually asked |
| 18:47 | <hober> | weird |
| 18:47 | <annevk> | cool so if that's feasible we should just find John Gregg I guess and ask whoever else wants to come to come? |
| 18:48 | <hober> | sgtm |
| 18:48 | <annevk> | sweet, I'll mail the list again |
| 18:53 | <Ms2ger> | So, zcorpan gets to spec http://mxr.mozilla.org/mozilla-central/source/layout/style/nsStyleUtil.cpp#64 / http://trac.webkit.org/browser/trunk/Source/WebCore/css/CSSStyleSelector.cpp?rev=113922#L4920 |
| 19:02 | <annevk> | hober: I'm starting to feel like those people that email "meeting" emails to mailing lists around all the time instead of technical content |
| 19:03 | <annevk> | it's only my second such email thus far, but the feeling is strong |
| 19:15 | <hober> | annevk: :( |
| 19:17 | <Ms2ger> | MikeSmith, should be done now |
| 19:20 | <arunranga> | annevk, https://www.w3.org/Bugs/Public/show_bug.cgi?id=16790 for micro tasks now exists #FYI |
| 19:21 | <annevk> | nice |
| 19:45 | <MikeSmith> | Ms2ger: thanks |
| 19:45 | <Ms2ger> | Np |
| 19:49 | <annevk> | "Ack: MikeSmith, decoder." |
| 19:51 | <Ms2ger> | Not your kind of decoder |
| 19:54 | <Ms2ger> | "But test suites will soon be generated based on spec links." |
| 19:54 | <Ms2ger> | We live in a brave new world |
| 19:54 | <TabAtkins> | Yesssss |
| 19:55 | Ms2ger | enters whatwg.org/html |
| 19:55 | Ms2ger | waits for the millions of tests to roll out |
| 19:58 | <MikeSmith> | hey does anybody here read Dr. Dobbs journal, ever? |
| 19:59 | <MikeSmith> | I ask because i wrote up something for that site as a favor |
| 20:00 | <MikeSmith> | but jackass editor at that site seems to have found that it didn't meet his narrow view of what he's interested in publishing |
| 20:02 | <MikeSmith> | and I'm like, hey I never read anything you publish anyway, nor does anybody else I know, so if you don't choose to publish it, then nobody's going to care anyway |
| 20:02 | <annevk> | what was it about? |
| 20:02 | <Philip`> | He's not even a real doctor |
| 20:03 | <MikeSmith> | "HTML5" |
| 20:03 | <annevk> | MikeSmith: publish on blog.whatwg.org or your own blog? |
| 20:03 | <annevk> | (don't read dr. dobbs journal btw) |
| 20:04 | <MikeSmith> | this fuckartard says he wants an article about "where the HTML5 formalization process is, what still remains to be done, how long that might take, and finally what can be done today that won't be affected by the remaining items to be done" |
| 20:04 | <MikeSmith> | which I have no interest in writing |
| 20:05 | <MikeSmith> | instead I attempted to write something for him that actual developers might be interested in reading |
| 20:05 | <MikeSmith> | annevk: yeah, will post it somewhere else for sure |
| 20:06 | <annevk> | let me know when you tweet about it or when it's up |
| 20:06 | <annevk> | can share from @whatwg |
| 20:06 | <MikeSmith> | plus he's asking, "might I ask why the original article was not written?" |
| 20:07 | <MikeSmith> | the answer to which is, he never offered to pay us dime one for writing it |
| 20:07 | <MikeSmith> | so I care fuck all what he thinks it should be |
| 20:08 | <MikeSmith> | this guy is apparently the executive editor for Dr. Dobbs |
| 20:08 | <MikeSmith> | which doesn't give me much confidence |
| 20:08 | <MikeSmith> | asshat |
| 20:09 | <MikeSmith> | annevk: yeah, will do |
| 20:09 | <Philip`> | An editor wanting some control over the articles that get written and published in his publication? That sounds crazy |
| 20:10 | <MikeSmith> | everybody please remind me to never again waste my time trying to accommodate tech journalist |
| 20:11 | <MikeSmith> | Philip`: apparently they can find lots of people who will write crap for them without being paid |
| 20:12 | <MikeSmith> | in contrast to me, where I wrote this as a favor to somebody else on the W3C team who asked me to do it |
| 20:12 | <MikeSmith> | not for this moron |
| 20:13 | <gsnedders> | u"a" in UCS-4 Python is 56 bytes. Ow. |
| 20:13 | <MikeSmith> | heh |
| 20:13 | <TabAtkins> | ...what? |
| 20:13 | <MikeSmith> | python ftw |
| 20:14 | <annevk> | gsnedders: you meant to write bytes, not bits? |
| 20:15 | <gsnedders> | annevk: Yes. |
| 20:15 | <annevk> | o_O |
| 20:15 | <gsnedders> | I mean, type flag + repr shouldn't be that big, surely!? |
| 20:16 | <gsnedders> | I guess there's no type flag but a pointer to its class |
| 20:16 | <gsnedders> | Which would be eight bytes, plus four for the actual repr. |
| 20:18 | <gsnedders> | Interestingly, "a" (i.e., a byte string) is 38 bytes, even though I'd expect it to only be be three bytes smaller. |
| 20:30 | <Philip`> | gsnedders: x86 or amd64? |
| 20:30 | <gsnedders> | Philip`: amd64, lp64 |
| 20:49 | <jsbell> | annevk: re: "TYPO in spec?" - heh, yeah, was going to dig in a bit before bugging you |
| 20:52 | <jgraham> | gsnedders: I think assuming 12 bytes would be wildly optimistic. It stores a bunch of things e.g. length, reference count |
| 20:53 | <gsnedders> | jgraham: Oh, yeah, right, Python still has ref counting… |
| 20:54 | <jgraham> | Well CPython |
| 20:54 | <jgraham> | But yes |
| 20:58 | <jgraham> | They also seem to store their hash value, an integer representing whether they are UCS_2, UCS_4 or various other things, an integer indicating whether they are compact, an integer indicating whether they are ascii and some sort of ready flag |
| 20:58 | <gsnedders> | jgraham: What version of Python are you looking at? |
| 20:59 | <jgraham> | http://hg.python.org/cpython/file/b5e6cbacd6ab/Include/unicodeobject.h |
| 20:59 | <jgraham> | Which I assume is latest trunk |
| 20:59 | <gsnedders> | Irrelevant. The Unicode repr is changing in 3.3. |
| 20:59 | <jgraham> | Seems like you could do a bunch of bit packing here to improve the memory usage |
| 20:59 | <jgraham> | gsnedders: You mean has changed? |
| 21:00 | <gsnedders> | (As it now stores stuff as ASCII when possible, and then UCS-2 if possible, and finally UCS-4 if it has to, instead of it being set by compiled.) |
| 21:00 | <gsnedders> | jgraham: Has changed is trunk, is changing in the next release… |
| 21:00 | <gsnedders> | s/by compiled/at compile time/ |
| 21:01 | <gsnedders> | http://docs.python.org/dev/whatsnew/3.3.html#pep-393-flexible-string-representation |
| 21:01 | <jgraham> | 2.7 seems to be http://hg.python.org/cpython/file/dd23333b579a/Include/unicodeobject.h#l413 |
| 21:01 | <jgraham> | gsnedders: Ah, that would explain a lot |
| 21:02 | <jgraham> | But it will make the memory usage for your TC worse I would think |
| 21:02 | <gsnedders> | jgraham: Makes me want to rewrite this in C more and more… |
| 21:03 | <Ms2ger> | gsnedders, but think of the kittens! |
| 21:03 | <jgraham> | I really don't see why they need a whole int to store a compact flag and another int to store an ascii flag and another int to store a ready flag |
| 21:03 | <gsnedders> | jgraham: Yeah, that just seems silly. |
| 21:03 | <gsnedders> | jgraham: Or maybe compat with the old API? |
| 21:03 | <Ms2ger> | Access speed? |
| 21:03 | <gsnedders> | Ms2ger: ASCII/UCS2/UCS4 should all be one flag, logically. |
| 21:03 | <jgraham> | gsnedders: Maybe? |
| 21:04 | Ms2ger | points at the topic |
| 21:04 | <jgraham> | The old struct looks pretty differnt at least |
| 21:04 | <gsnedders> | jgraham: ascii is only one bit |
| 21:04 | <gsnedders> | jgraham: I think you need to learn more C :) |
| 21:05 | <gsnedders> | jgraham: (actually, it depends on how the compiler decides to layout the struct in memory, but the code dictates ascii is one bit) |
| 21:05 | <Ms2ger> | "Only 7-bit ASCII data is excepted." |
| 21:05 | <Ms2ger> | Grr |
| 21:06 | <jgraham> | gsnedders: Oh, I see |
| 21:07 | <jgraham> | Well ignore my previous complaint then |
| 21:07 | <gsnedders> | jgraham: Ah, 1BYTE (within kind) and ascii are different: 1BYTE_KIND allows ISO-8859-1, ascii allows US-ASCII |
| 21:07 | <jgraham> | I was assuming that C was a logical language |
| 21:07 | <gsnedders> | Probably so it knows what it can pass to system functions which expect ASCII |
| 21:07 | <jgraham> | So that if it said "unsigned int" you actually got an unsigned int |
| 21:08 | <gsnedders> | (I'm trying to implement a trie in Python without using ALL THE MEMORY.) |
| 21:09 | <jgraham> | gsnedders: You know short strings are intered, right? |
| 21:09 | <gsnedders> | jgraham: Yeah. |
| 21:10 | <gsnedders> | Still, my initial, trivial impl used a lot. |
| 21:33 | <myndzi> | that document - holy cow. incredible. |
| 21:33 | <myndzi> | is there a list of compliant (or close) parsers somewhere? |
| 21:43 | <jgraham> | What document? What parsers? |
| 21:46 | <myndzi> | the html parsing document on whatwg.org? |
| 21:46 | <Hixie> | you mean the spec? :-) |
| 21:46 | <myndzi> | sure |
| 21:46 | <myndzi> | The Spec |
| 21:46 | <myndzi> | :) |
| 21:46 | <Hixie> | as, the HTML standard? |
| 21:47 | <Hixie> | there's no browser that does exactly what the HTML standard says to do |
| 21:47 | <Hixie> | but there's several browsers who are basically trying to coverge on it |
| 21:47 | <Hixie> | if that's what you mean |
| 21:47 | <myndzi> | not a browser, a library, and i was just curious if there's a list somewhere |
| 21:47 | <myndzi> | to help identify the ones that do it the best |
| 21:47 | <Hixie> | just for the HTML parser part? |
| 21:48 | <Hixie> | html5lib is one i know about, validator.nu has one too |
| 21:48 | <myndzi> | yeah, just the html parser part |
| 21:48 | <Hixie> | and webkit has a parsing component, dunno how easy it would be to reuse |
| 21:48 | <myndzi> | tokenizing and fixing all the messed up stuff |
| 21:48 | <myndzi> | etc. |
| 21:48 | <myndzi> | i'll check em out |
| 22:03 | <jgraham> | Right, Opera, Gecko, WebKit, (IE10?) all have (almost) compliant parsers although they are not really easy to reuse; the Gecko parser (also used in validator.nu) probably has the most potential. html5lib has a python implementation and a less good PHP implementation. It also has some Ruby code that we don't talk about much. |
| 22:03 | <jgraham> | I'm not sure about other implementations. It would be nice if someone fixed libxml2 |
| 22:03 | <myndzi> | there's one that's looking interesting in node |
| 22:04 | <myndzi> | looks pretty clean and thorough |
| 22:04 | <myndzi> | not sure if i can hijack it to my purposes or not yet ;) |
| 22:04 | <myndzi> | jsdom gave me trouble last time i had anything to do with it |
| 22:07 | <hober> | myndzi: aredridel's parser is really nice |
| 22:07 | <smaug____> | jgraham: there is the dom.js parser |
| 22:08 | <myndzi> | yeah, the more i look at it the more i'm like (!) |
| 22:08 | <myndzi> | this may jsut save me some work :> |
| 22:08 | <jgraham> | Oh yeah, there's at least two javascript implementations! |
| 22:09 | <smaug____> | https://github.com/andreasgal/dom.js/blob/master/src/impl/HTMLParser.js |
| 22:09 | <jgraham> | And hober was going to write one in elisp :p |
| 22:09 | <smaug____> | no idea how good that is |
| 22:09 | <hober> | jgraham: well, i wrote the tokenizer |
| 22:09 | <hober> | jgraham: but stalled out working on the treebuilder |
| 22:10 | <myndzi> | smaug____: lookin like the other one is nicer; this code is beautiful ;) |
| 22:11 | <zewt> | <arunranga> annevk: won't exactly stop race issues, but can mitigate them. |
| 22:12 | <zewt> | those can be eliminated, it'll just take a bit more work |
| 22:12 | <myndzi> | i was gonna say i wish i could clean all the excess stuff out since there's a bunch i don't need, but it looks like it's going "quite fast enough, thank you very much" |
| 22:12 | <arunranga> | Hi zewt, were you referring to the aftermath of https://www.w3.org/Bugs/Public/show_bug.cgi?id=16790 ? |
| 22:13 | <zewt> | dunno, firefox is being kicked in the head by the one-page spec at the moment :) |
| 22:14 | <arunranga> | Which one-page are you referring to? |
| 22:14 | <zewt> | http://www.whatwg.org/specs/web-apps/current-work/, the one that everyone fears clicking |
| 22:14 | <smaug____> | it isn't that bad |
| 22:14 | <gsnedders> | jgraham: Unless I'm being stupid here, unicode strings aren't interned in Python 2 |
| 22:14 | arunranga | heh |
| 22:14 | <Hixie> | i haven't had any trouble with the one-page spec for a while |
| 22:15 | <Hixie> | dunno what hardware y'all are using but mine isn't THAT new :-) |
| 22:15 | <arunranga> | I actually kicked the one page spec in the head. |
| 22:15 | <Hixie> | (mid 2009 macbook pro and mid 2010 imac, mainly) |
| 22:15 | <arunranga> | (by adding 4G more RAM) |
| 22:15 | <Hixie> | ah, yeah, low RAM might be a problem |
| 22:15 | <Hixie> | i always try to max out my ram |
| 22:15 | <gsnedders> | Hixie: RAM is the big issue with the spec |
| 22:16 | <jgraham> | gsnedders: Oh, maybe |
| 22:16 | <arunranga> | Life on 8G of RAM makes me happy. |
| 22:16 | <gsnedders> | jgraham: Though looking at this again, I'm doing everything with single-char byte strings |
| 22:16 | <jgraham> | gsnedders: 127 characters should be more than enough for anyone |
| 22:17 | <zewt> | 16gb, still takes a good 7-8 seconds to load (in a sub-500ms web, that's a lifetime :) |
| 22:17 | <smaug____> | loading the spec seems to take about 300MB |
| 22:17 | <gsnedders> | jgraham: In related news, we need to sort out all the strings in html5lib, because we rely upon implicit conversion often |
| 22:17 | <jgraham> | Yes |
| 22:17 | <gsnedders> | jgraham: And that conversion time probably costs us in places. |
| 22:17 | <jgraham> | No |
| 22:17 | <jgraham> | I would expect |
| 22:17 | <jgraham> | But we should actually do a python 3 port |
| 22:17 | <jgraham> | so yes, we should fix it |
| 22:18 | <jgraham> | Anyway I should be asleep |
| 22:18 | <gsnedders> | jgraham: entitiesStartingWith was last time I looked by far the slowest thing when parsing the spec |
| 22:18 | <gsnedders> | jgraham: And relies upon implicit conversion |
| 22:18 | <jgraham> | Well if we fix it and it makes us faster then I won't be unhappy |
| 22:19 | <jgraham> | If you fix it and it makes us faster I will be even less unhappy :) |
| 22:23 | <gsnedders> | jgraham: Yup, the implicit coercion costs us ~0.8s on my fairly high-powered machine |
| 22:24 | <gsnedders> | 17.1 -> 16.3s to parse the spec |
| 22:30 | <gsnedders> | Pushed. |
| 22:47 | gsnedders | concludes that a 10% improvement in parsing the spec is good enough for half an hour of hacking, and heads to bed :) |
| 23:08 | <zewt> | gsnedders: looks like it's 32 bytes in 32-bit builds ... a lot of extra overhead for excessively large fields in the 64-bit build |
| 23:08 | <zewt> | (python strings) |
| 23:09 | <gsnedders> | zewt: Not that surprising seeming a lot of it is pointers. |
| 23:09 | <zewt> | more are just integers, actually |
| 23:10 | gsnedders | is obviously misremembering from looking earlier XD |
| 23:10 | <zewt> | string length (could have a separate "long string" type so regular sub-4gb strings don't pay for a 64-bit field), the hash (which appears to be 64-bit on 64-bit systems; should only be 32), and what appears to be a bitfield for interning (could also be 32-bit) |
| 23:11 | <zewt> | i'll worry about it if i ever decide to involve myself in python core development :) |