| 00:05 | <zewt> | wow, nobody in webgl understands how web specs work *at all* |
| 00:05 | <zewt> | "I think that we should mandate that as soon as a feature becomes available without prefix, support for the prefix should be dropped." |
| 00:05 | <zewt> | as if specs can force browsers to do things |
| 06:13 | <boblet> | can anyone give me an example showing the difference between s and del elements, where content that would be appropriate for one is not for the other? I’d like to check I understand the difference… |
| 06:26 | <zcorpan> | doesn't the spec have examples? |
| 06:49 | <boblet> | zcorpan: the spec uses an old price being replaced by a new price for s, and completed todo items, closed bugs, replaced words, and a removed table column for del. However I’m not sure why del wouldn’t be appropriate for the old price, or s for e.g. replaced words |
| 06:51 | <boblet> | for example, if an online newspaper misquoted someone, they’d want to show both the incorrect text (for context) and the new text. semantically it seems like s and ins, but I’d probably use del and ins to use @datetime on del |
| 06:53 | <zcorpan> | s+ins seems bogus |
| 06:56 | <zcorpan> | if you fixed a misquote, it seems like a deletion and addition, so del+ins (if you want to keep the misquoted text, otherwise just remove the misquoted text altogether) |
| 06:57 | <zcorpan> | (and don't use ins in that case) |
| 06:57 | <boblet> | zcorpan: s seems possible for a misquote under “no longer accurate”, but I’m also iffy about it. The only examples of s are obsolete prices (with new sale price) and sold out events. I’m wondering what other content would be appropriate for s |
| 06:59 | <zcorpan> | a misquote wasn't accurate to begin with :-) |
| 06:59 | <boblet> | :) any other examples you can think of? |
| 07:01 | <zcorpan> | *shrug* |
| 07:02 | <boblet> | zcorpan: thanks |
| 07:08 | <annevk> | false descriptions |
| 07:08 | <annevk> | he's a <s>piece of work</s> nice guy |
| 07:08 | <annevk> | or some such |
| 07:12 | <zcorpan> | name changes, maybe |
| 07:12 | <zcorpan> | <h1>Introducing <s>Crap</s> <s>Dirt</s> Dash</h1> |
| 07:13 | <annevk> | hsivonen: did you just write "when they" as "one day"? |
| 07:13 | <annevk> | hsivonen: because otherwise I'm not sure what you wrote |
| 07:16 | <hsivonen> | annevk: yes. speech recognition for the lose |
| 07:17 | <annevk> | oh, if it was all dictated that's quite good then I guess |
| 07:18 | <zcorpan> | annevk: did you get anywhere with big5? |
| 07:19 | <annevk> | I haven't done much during the weekend, I have some kind of headache I can't seem to get rid of |
| 07:20 | <zcorpan> | :-/ |
| 07:21 | <zcorpan> | hmm |
| 07:21 | <zcorpan> | Simons-MacBook-Pro:Dotnetdotcom zcorpan$ grep -aFic "big5" web200904 |
| 07:21 | <zcorpan> | 8882 |
| 07:21 | <annevk> | it occurs in content too, but that's a lot higher |
| 07:23 | <annevk> | actually it's not |
| 07:23 | <zcorpan> | do you want the subset with pages that contain "big5"? |
| 07:23 | <annevk> | if I search for big5 through the data you gave me I get 8511 |
| 07:24 | <zcorpan> | oh |
| 07:26 | <annevk> | hmm, but if I do the same grep I get 5780 |
| 07:26 | <zcorpan> | that's weird. does the i flag not work with F? |
| 07:28 | <annevk> | I think my Python script might be wrong somehow |
| 07:28 | <zcorpan> | grep -aPc "[bB][iI][gG]5" web200904 |
| 07:28 | <zcorpan> | 8892 |
| 07:29 | <zcorpan> | grep -aFizHZ "big5" web200904 > big5-all.txt resulted in a 91MB file |
| 07:30 | <annevk> | yeah my Python script must be wrong |
| 07:31 | <annevk> | though how exactly... |
| 07:31 | <annevk> | bytes = open("big5.txt", "rb").read() |
| 07:31 | <annevk> | find = b"big5" |
| 07:31 | <annevk> | found = b"" |
| 07:31 | <annevk> | c = 0 |
| 07:31 | <annevk> | for b in bytes: |
| 07:31 | <annevk> | if b.lower() in find: |
| 07:31 | <annevk> | found += b.lower() |
| 07:31 | <annevk> | else: |
| 07:31 | <annevk> | if found == find: |
| 07:31 | <annevk> | c +=1 |
| 07:31 | <annevk> | found = b"" |
| 07:31 | <annevk> | print c |
| 07:31 | zcorpan | is uploading it zipped (18MB) over 3G |
| 07:32 | <annevk> | why are you on 3G? |
| 07:32 | <zcorpan> | on the train |
| 07:38 | <annevk> | with the script above I get more results than you |
| 07:38 | <annevk> | 11k |
| 07:39 | <annevk> | zcorpan: is the grep thing just substring matches or are there other conditions? |
| 07:40 | <zcorpan> | your script seems to miss a count in e.g. "big55" |
| 07:40 | <zcorpan> | should be just substring |
| 07:41 | <annevk> | then I would get even higher numbers |
| 07:41 | zcorpan | switching trains |
| 07:41 | <annevk> | running it on the 29MiB file gives 11k results |
| 07:41 | <annevk> | oh well |
| 07:46 | <hsivonen> | http://lists.w3.org/Archives/Public/www-tag/2012Apr/0010.html |
| 07:49 | <annevk> | my MacBook crashed hard |
| 07:49 | <annevk> | lots of pretty colored lines on the screen |
| 07:49 | <zcorpan> | annevk: textedit finds 11669 "big5"s in big5.txt |
| 07:49 | <zcorpan> | annevk: and 15564 in big5-all.txt |
| 07:50 | <zcorpan> | annevk: maybe the grep count counts "lines" |
| 07:50 | <annevk> | that's the same amount I found |
| 07:50 | <annevk> | exact |
| 07:51 | <annevk> | great, so my scripts are not entirely buggy |
| 07:51 | <zcorpan> | it's buggy for the input i gave :-) |
| 07:52 | <annevk> | hence "not entirely" |
| 07:55 | <annevk> | the problem is that I haven't really found a way to analyze the data properly |
| 07:55 | <annevk> | there's about a 1000 files in what you gave me |
| 07:56 | <annevk> | and over 600000 potential code points in the PUA / HKSCS range |
| 07:57 | <annevk> | so far I thought of writing a tokenizer to strip out the HTTP related bits and store the HTML bits each in one file |
| 07:57 | <annevk> | then maybe have some statistics on a per file basis and study a couple of them |
| 08:01 | <MikeSmith> | hsivonen: "There really are spoons." |
| 08:02 | <hsivonen> | MikeSmith: did I sent more misdictated email? |
| 08:03 | <MikeSmith> | hsivonen: no, see the reply |
| 08:03 | <MikeSmith> | from Pat Hayes |
| 08:03 | <MikeSmith> | to the www-tag message you cited |
| 08:03 | <hsivonen> | I see |
| 08:05 | <MikeSmith> | bz implementing ruby support? |
| 08:05 | <MikeSmith> | https://bugzilla.mozilla.org/show_bug.cgi?id=256274 |
| 08:05 | <annevk> | www-tag is top of my list of lists I am considering to unsubscribe from |
| 08:05 | <MikeSmith> | oh, maybe he's just merging existing patches to the tip |
| 08:07 | <hsivonen> | MikeSmith: looks like it. it has been on his list of things to do |
| 08:07 | <MikeSmith> | ok |
| 08:07 | <MikeSmith> | good to see |
| 08:07 | <annevk> | original patch seems to be from Hajime Shiozawa |
| 08:08 | <zcorpan> | annevk: simon.html5.org/dump/big5-all.txt.zip |
| 08:09 | <hsivonen> | annevk: IIRC, the original needed some additional attention from bz |
| 08:15 | <annevk> | thanks zcorpan, even more data I don't know what to do with :p |
| 08:19 | <MikeSmith> | is the list of encoding names in the Encodings spec the complete list that the spec will contain? |
| 08:20 | <MikeSmith> | I realize some of the ones you have there you haven't specced out yet, but I just mean is that the complete outline at least |
| 08:20 | <annevk> | MikeSmith: unless someone finds another encoding we need to add |
| 08:20 | <MikeSmith> | OK |
| 08:21 | <annevk> | MikeSmith: and it has been suggested to remove the remaining IBM encodings |
| 08:21 | <MikeSmith> | oh |
| 08:21 | <annevk> | as well as iso-2022-cn |
| 08:21 | <MikeSmith> | why? |
| 08:21 | <annevk> | stronger case for the latter |
| 08:21 | <annevk> | because not all browsers implement them |
| 08:22 | <MikeSmith> | OK |
| 08:22 | <MikeSmith> | so the criteria for what's included is that it's limited to the set of encodings that all browsers support, right? |
| 08:22 | <annevk> | zcorpan: seems you did not upload a complete file |
| 08:22 | <annevk> | zcorpan: it downloads 14MiB and does not recognize it as a zip file |
| 08:23 | <annevk> | MikeSmith: or something that's needed for compatibility |
| 08:23 | <MikeSmith> | OK |
| 08:24 | <annevk> | MikeSmith: and potentially something that's really useful, but "better than utf-8" has not been found yet to my knowledge :) |
| 08:26 | <MikeSmith> | annevk: so that's basically what I already told Richard but I think at some point he's going to ask you himself, and maybe ask about specific encodings |
| 08:27 | <annevk> | MikeSmith: cool |
| 08:27 | <MikeSmith> | annevk: would publishing this as a deliverable of the i18n WG be an option? |
| 08:27 | <MikeSmith> | that would put it in the same case as the charmod spec |
| 08:28 | <MikeSmith> | does HTML5 normatively reference charmod? or do any other specs? |
| 08:28 | <MikeSmith> | I mean it would put it in the same status as charmod as far as W3C publication |
| 08:31 | <annevk> | I thought i18n didn't do RECs? |
| 08:31 | <annevk> | but charmod is a REC |
| 08:31 | <MikeSmith> | yeah |
| 08:31 | <annevk> | so that might be okay |
| 08:31 | <MikeSmith> | I'm not so clear on status of charmod |
| 08:31 | <MikeSmith> | is it actually a REC? |
| 08:32 | <annevk> | http://www.w3.org/TR/charmod/ says so |
| 08:32 | <annevk> | a frequently violated REC |
| 08:32 | <MikeSmith> | yeah, I see it is |
| 08:33 | <annevk> | oh sweet |
| 08:33 | <annevk> | the original Prince of Persia code is going online |
| 08:34 | <annevk> | I hope someone makes that playable in a browser |
| 08:35 | <zcorpan> | annevk: try again |
| 08:35 | <annevk> | trying again |
| 08:39 | <annevk> | I think it worked |
| 08:44 | <annevk> | it did and it seems I hit the bug zcorpan found in my script |
| 08:44 | annevk | finds 15537 hits for big5 |
| 09:30 | <annevk> | in the larger dataset not every file uses correct HTTP line endings |
| 09:31 | <Ms2ger> | Breaking news: people violate HTTP |
| 09:32 | <annevk> | the problem is I need to change my simplistic tokenizer |
| 09:32 | <jgraham> | HTTP working group being treated for shock. |
| 09:32 | <annevk> | I guess I should eat 0D when followed by 0A and otherwise use 0A |
| 09:32 | <annevk> | or not worry about the larger dataset for now |
| 09:33 | annevk | does that |
| 09:46 | <annevk> | only a third of the files defines HTTP level charset |
| 09:46 | <annevk> | of which a tenth is not big5/big5-hkscs |
| 09:47 | <annevk> | and of those 34 a couple are bogus, some utf-8, iso-8859-1, ms950, and x-ms950-hkscs |
| 09:47 | <annevk> | bogus is actually either the empty string or b"null" |
| 09:48 | <annevk> | (i.e. those bytes, no quotes) |
| 09:48 | <annevk> | oh, and one euc_kr |
| 09:48 | <annevk> | which is also bogus, as it should be euc-kr to be recognized |
| 09:49 | <annevk> | the way I search for charset is somewhat bogus too btw, but the cheat is justified for the dataset :) |
| 10:02 | <annevk> | zcorpan: hmm |
| 10:03 | <annevk> | zcorpan: are we sure they give the raw data? |
| 10:05 | <annevk> | zcorpan: if I open a couple of test pages, decoding them as utf-8 gives better results :/ |
| 10:06 | <annevk> | what's the dataset source again? |
| 10:07 | <annevk> | ah http://dotnetdotcom.org/ |
| 10:09 | <annevk> | to be clear, in processing I only opened files with the "b" flag set |
| 10:13 | <annevk> | Philip`: know anything about that? |
| 10:15 | <zcorpan> | annevk: also for big5.txt ? |
| 10:22 | <annevk> | zcorpan: you mean big5-all.txt? |
| 10:22 | <annevk> | zcorpan: this was on big5-.txt |
| 10:24 | <annevk> | big5.txt |
| 10:26 | <annevk> | I looked at the first 50 files |
| 10:27 | <annevk> | lots have big5 in HTTP set |
| 10:27 | <annevk> | but no big5 in the actual data |
| 10:33 | <zcorpan> | annevk: i meant big5.txt. big5-all.txt was zipped so might have been tampered with by zipping or unzipping, was my thought |
| 10:35 | <annevk> | it seems sort of plausible they have done normalization given the zero byte delimited files |
| 10:35 | <annevk> | but it's clearly not great for this |
| 10:38 | <annevk> | I think I'll email the dotnetdotcom guys just to be sure |
| 10:38 | <zcorpan> | textwrangler can't open big5.txt, but has no problems opening utf-8 files with nulls |
| 10:41 | <annevk> | well, there's no encoding conversion going on locally |
| 10:41 | <annevk> | bytes = open("big5.txt", "rb").read() |
| 10:41 | <annevk> | and |
| 10:41 | <annevk> | newbytes = open("test-" + str(c) + "." + charset + ".html", "wb") |
| 10:42 | <annevk> | the rest is just iterating over, testing on, and writing bytes |
| 10:52 | <annevk> | emailed dotdot |
| 11:13 | <annevk> | zcorpan: so yeah e.g. the euc_kr file which has big5 in <meta>, has EF BF BD as byte sequence which is UTF-8 for FFFD and is nothing in either other encoding |
| 11:14 | <zcorpan> | annevk: ok :( |
| 11:14 | <annevk> | and it has the same sequences in big5.txt as it has in my split out files |
| 11:14 | <annevk> | when I use a hex editor |
| 11:14 | <zcorpan> | too bad |
| 11:15 | <annevk> | I could prolly write a custom utf-8 decoder to find out why big5.txt cannot be opened in TextWrangler, but I'm not sure that's worth it |
| 11:44 | <annevk> | zcorpan: you could maybe quickly verify to be a 100% sure by checking some utf-16 data in the set |
| 11:44 | <annevk> | zcorpan: as additional sanity check |
| 11:46 | <Philip`> | The dotbot data can't contain any UTF-16 pages since it can't represent 0x00 bytes |
| 11:46 | <annevk> | doh |
| 11:47 | Philip` | never tried looking to see what they actually do with 0x00 bytes (maybe reject that page, or drop those bytes, or truncate, or whatever) |
| 11:47 | <annevk> | windows-1252 data with octets over > 0x7F works too |
| 11:49 | <zcorpan> | grep -aPc "^Content-Type\s*:\s*text/html\s*;\s*charset\s*=\s*[\"']?utf-16" web200904 |
| 11:49 | <zcorpan> | 0 |
| 11:54 | <zcorpan> | annevk: you could see what you find in http://webcrawl.s3.amazonaws.com/web.short.gz |
| 11:55 | <zcorpan> | annevk: maybe grep screws things up |
| 11:57 | <annevk> | no |
| 11:57 | <annevk> | search for windows-1252 |
| 11:57 | <annevk> | first octet sequence I find in that document is C3 96 |
| 11:57 | <annevk> | it's a German document, and in UTF-8 that is Ö |
| 11:58 | <annevk> | and that is followed by sterreich so I think it is indeed normalized :( |
| 11:58 | Philip` | notes that grep can screw things up if you don't run it with LANG=C |
| 11:58 | <annevk> | I'm just looking through the file zcorpan pointed out in a hex editor |
| 11:58 | <annevk> | nothing grep can screw up here |
| 11:59 | <annevk> | so sad panda face |
| 11:59 | <annevk> | big sad panda face |
| 12:00 | <zcorpan> | so i guess you need to go shopping for a different data set, or do a crawl yourself |
| 12:03 | <jgraham> | Probably wouldn't be that hard to do a custom crawl that just got the kind of data you want (i.e. rejected any pages that aren't the encoding you care about without storing them) |
| 12:05 | <Philip`> | Depends if by "crawl" you mean actually parsing pages and following links and trying to get a not disasterously biased dataset, or just downloading random pages from lists on dmoz.org or wherever |
| 12:05 | <jgraham> | Well the extent that you need to actually parse pages to do it is rather limited |
| 12:06 | <jgraham> | You just need to find things that look like URLs |
| 12:06 | <Ms2ger> | Doesn't the dotbot data give you the URLs as well? |
| 12:06 | <zcorpan> | Ms2ger: yes, but many urls are probably dead by now |
| 12:06 | <zcorpan> | http://s3.amazonaws.com/alexa-static/top-1m.csv.zip might be a useful starting point |
| 12:06 | <annevk> | what is that? |
| 12:07 | <zcorpan> | alexa's top 1 million sites |
| 12:07 | <Ms2ger> | Sounds like Alexa's top 1 million pages |
| 12:07 | <annevk> | is there a way to get all the URLs from the dotbot pages who have big5 somewhere? |
| 12:07 | <zcorpan> | it'll only give you front pages |
| 12:08 | <annevk> | 'cause then I could just download the dotbot pages again |
| 12:08 | <Philip`> | jgraham: Most are relative URLs, so you need some way to resolve them properly, and it's quite possible the URLs include non-ASCII characters so you need to decode the pages first |
| 12:08 | <annevk> | and those that don't 404 and are still big5 would be useful |
| 12:09 | <zcorpan> | Philip`: can you think of a good way to do that? |
| 12:09 | <Philip`> | and you probably want to avoid following links like <script>..."<a href='"+url+"'>"... etc, so you can't just dumbly tokenise everything |
| 12:09 | <jgraham> | Philip`: Sure, decoding is necessary |
| 12:10 | <jgraham> | Philip`: But only picking absolute urls is fine because it biases you towards covering many different sites, not lots of data from the same site |
| 12:10 | <Philip`> | I guess the harder part is trying to avoid downloading a zillion pages from database-driven sites and no pages from anywhere else |
| 12:10 | <jgraham> | (which one also does by storing some sort of map of domain -> # pages and capping the number of pages per domain) |
| 12:11 | <Philip`> | zcorpan: If you're finding the pages with grep -z, maybe you could add "-B2" so it prints the preceding two 'lines' (which includes the URL)? |
| 12:12 | <zcorpan> | ooh, clever |
| 12:12 | <annevk> | given the URL encoding stuff having everything in utf-8 makes perfect sense |
| 12:12 | <annevk> | although not necessarily for query parts, but maybe they do something special there too |
| 12:27 | <zcorpan> | annevk: http://simon.html5.org/dump/big5-all-with-urls.txt.zip - i'll leave it to you to parse out the urls :-) |
| 12:38 | <annevk> | ooh, that includes all the data? |
| 12:38 | <annevk> | hmm |
| 12:39 | <annevk> | guess that's not much of a problem |
| 13:16 | <davidb> | oat |
| 13:19 | <Ms2ger> | Morning |
| 13:27 | <mattwest> | What is everybody's thoughts about using the <article> element to wrap the content of standard web page. For example something like an about page. |
| 13:40 | <annevk> | alright so I make a string from start of file or 0A until I hit 00, append that to list of URLs and carry on... |
| 13:41 | <annevk> | zcorpan: can you actually make this one for the smaller big5.txt? |
| 13:41 | <annevk> | zcorpan: I think that contained more accurate data, this includes stuff such as big5.gif |
| 13:55 | <zewt> | Web Application Store Community Group Launched |
| 13:55 | <zewt> | |
| 13:55 | <zewt> | april fool's was yesterday, guys |
| 13:56 | <zcorpan> | annevk: http://simon.html5.org/dump/big5-with-urls.txt.zip |
| 13:56 | <zcorpan> | LANG=C grep -aEizB2 "(content(-type[[:space:]]*:|=[[:space:]]*[\"']?)[[:space:]]*text/html[[:space:]]*;|<meta[[:space:]])[[:space:]]*charset[[:space:]]*=[[:space:]]*[\"']?(big5|cn-big5|csbig5|x-x-big5|big5-hkscs)" web200904 > big5-with-urls.txt |
| 13:57 | <zewt> | i just died a little in my mouth |
| 13:58 | <annevk> | one internet awarded to zcorpan |
| 13:59 | <zewt> | as soon as grep seriously expects me to say [[:anything:]], it's time to switch to perlre |
| 13:59 | <zewt> | it's near the top of the ridiculous-syntax list |
| 13:59 | <zcorpan> | -z doesn't work with -P :( |
| 14:00 | <zewt> | i'd write a real script first, heh |
| 14:00 | <zewt> | anyway off to work |
| 14:00 | <zcorpan> | this was simpler |
| 14:44 | <annevk> | zcorpan: I'll have to look up HTML rendering for that |
| 14:45 | <annevk> | zcorpan: prolly 18px |
| 14:45 | <zcorpan> | annevk: the quiz is on twitter :-P |
| 14:45 | <annevk> | root margin doesn't collapse iirc |
| 14:46 | Ms2ger | thinks the best part of http://dev.w3.org/csswg/css3-tables-algorithms/Overview.src.htm is that it was written in MS Word |
| 14:47 | <zcorpan> | Ms2ger: LOL |
| 14:48 | <Ms2ger> | <meta name=Generator content="Microsoft Word 12 (filtered)"> |
| 14:48 | <annevk> | kind of insane that the return value of urllib2.urlopen() does not expose response codes |
| 14:48 | <annevk> | there might be a reason to have XMLHttpRequest in Python after all |
| 14:49 | <Ms2ger> | You should talk to lckl luke |
| 14:49 | <jgraham> | annevk: The built in urllib is kind of icky |
| 14:49 | <jgraham> | You probably want to use httplib2 or requests instead |
| 14:50 | <annevk> | I'm using urllib2 |
| 14:50 | <jgraham> | Well, like I said, that's your first problem |
| 14:51 | <jgraham> | (I mean it's not *unusable*, it's just that both the above mentioned libs are *better*) |
| 14:51 | <annevk> | isn't urllib2 the one written by Joe Gregorio and now finally included by Python? |
| 14:51 | <jgraham> | Not in any version of python I know about at least |
| 14:51 | <jgraham> | httplib2 is the one he wrote |
| 14:52 | <annevk> | so what is urllib2 then? |
| 14:52 | <jgraham> | Part of the standard library |
| 14:52 | <annevk> | hmm okay |
| 14:52 | <annevk> | stuff is confusing |
| 14:53 | <jgraham> | That partially relplaces urllib, which is also part of the standard library |
| 14:53 | <Ms2ger> | We need a urllib5, I guess |
| 14:53 | <jgraham> | Not python's finest moment |
| 14:53 | <jgraham> | Anyway, all the cool kids are using requests these days |
| 14:53 | <jgraham> | http://docs.python-requests.org/en/latest/index.html |
| 14:54 | <jgraham> | Although httplib2 is a fine choice also |
| 14:55 | <annevk> | not helping reduce my headache |
| 14:55 | <annevk> | but requests does look nice |
| 14:55 | <jgraham> | "pip install requests" -> no more headache |
| 14:55 | <jgraham> | (not medical advice) |
| 14:57 | <annevk> | ooh |
| 14:57 | <annevk> | but I can use the undocumented getcode() method too |
| 15:11 | <zcorpan> | http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1442 |
| 15:29 | <annevk> | this new strategy works btw |
| 15:30 | <annevk> | it'll take a while to get all the data downloaded, but I have some big5 pages now |
| 15:35 | <karlcow> | "You might be intrigued by the existence of two separate URL modules in Python - urllib and urllib2. Even more intriguing: they are not alternatives for each other. " — http://www.hacksparrow.com/python-difference-between-urllib-and-urllib2.html |
| 15:35 | <karlcow> | the support for Web standards in python std lib is kind of subpar in general |
| 15:36 | <Ms2ger> | Well, it's kind of subpar in browsers too |
| 15:36 | <annevk> | but then standards such as URL are subpar too |
| 15:37 | <karlcow> | and then if you thought there were not enough choices http://pypi.python.org/pypi/urllib3/ |
| 15:37 | <Ms2ger> | "The Interwebs: We're Subpar" |
| 15:37 | <karlcow> | ;) |
| 15:37 | <karlcow> | Ms2ger: hmm nice tag line :) |
| 15:37 | <Ms2ger> | Maybe something for MikeSmith's platform page :) |
| 15:38 | <karlcow> | http://urllib3.readthedocs.org/ |
| 15:39 | <karlcow> | https://github.com/shazow/urllib3 |
| 16:07 | <dglazkov> | good morning, Whatwg! |
| 16:09 | <jgraham> | dglazkov: 33% -> FAIL |
| 16:10 | <dglazkov> | good evening, jgraham? |
| 16:10 | <jgraham> | Ooh, much better :) |
| 16:11 | <dglazkov> | moderately crappy evening, jgraham! |
| 16:11 | dglazkov | aims to please |
| 16:11 | <jgraham> | :) |
| 16:12 | <Ms2ger> | Reasonable afternoon, dglazkov |
| 16:15 | <dglazkov> | reasonable indeed! The cafe downstairs is serving sardines and seaweed salad for breakfast. <3 |
| 16:22 | <jgraham> | Oh, is today at google btought to you by the letter "s"? |
| 16:22 | <jgraham> | *brought |
| 16:22 | <jgraham> | Which number is it? |
| 16:44 | <MikeSmith> | annevk: so apparently i18n WG was chartered to do RECs when charmod was published, but not any longer |
| 16:49 | <annevk> | MikeSmith: makes sense |
| 17:17 | <dglazkov> | jgraham: number 1, of course! :) |
| 18:09 | <dglazkov> | http://infrequently.org/2012/04/bedrock/ |
| 18:35 | <tantek> | slightlyoff - LOL at "Fault tollerence" [sic] - http://infrequently.org/12/eclipsecon/#23 |
| 18:36 | <tantek> | (found linked from your bedrock article that dglazkov dropped into the channel) |
| 18:52 | <annevk> | dglazkov: looks like the same message as earlier |
| 18:55 | <annevk> | dglazkov: it would be more interesting once someone outlined how say <textarea> would be implemented using these low-level primitives, including styling, text selection, etc. |
| 19:13 | <dglazkov> | annevk: yup |
| 19:14 | <dglazkov> | annevk: textarea is already pretty much done that way in WebKit -- just the long tail of one-off reconciliation tweaks remain. |
| 19:15 | <dglazkov> | annevk: a similar experiment for me is implementing innerHTML. |
| 19:23 | <zcorpan> | http://simon.html5.org/csspubquiz |
| 19:26 | <Ms2ger> | Semantic markup and twitter? Fascinating combination |
| 19:31 | <zcorpan> | all my quizzes so far are actual things i've stumbled upon, all but one in relation to the quirks spec |
| 20:56 | <Hixie> | someone asks if there's an equivalent of me but for JS -- who should I send them to for JS feedback? |
| 20:57 | <gsnedders> | There isn't. |
| 20:57 | <gsnedders> | es-discuss if you want a mailing list |
| 20:57 | <Ms2ger> | You? |
| 20:57 | <Hixie> | es-discuss@ what domain again? mozilla.org? |
| 20:57 | <gsnedders> | Hixie: Yeah |
| 20:57 | <Hixie> | thanks |
| 20:57 | <gsnedders> | If you want private feedback, join ECMA and TC39. |
| 21:01 | <Hixie> | anyone have an opinion on http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2012-January/034489.html ? |
| 21:02 | <Ms2ger> | I don't know anything about it, so I default to assuming dbaron is right |
| 21:03 | <Hixie> | yeah me too :-) |
| 21:23 | <rubys1> | Brendan is the closest thing to a BDFL for JS |
| 21:24 | <gsnedders> | Still a long way from it, though |
| 21:25 | <rubys1> | Hixie: got a minute? Apparently your proposal <http://krijnhoetmer.nl/irc-logs/whatwg/20120328#l-1010>; wasn't as controversial as I would have expected. |
| 21:27 | <Hixie> | rubys1: sure, sup |
| 21:27 | <rubys1> | a few questions, and perhaps we can wrap this up quickly. Do you have any thoughts on what template would be used? |
| 21:28 | Hixie | follows the link to work out what we're talking about |
| 21:28 | <Hixie> | template? |
| 21:28 | <Hixie> | you mean what boilerplate? as little as possible, i'd hope |
| 21:29 | <rubys1> | that's probably sufficient for my purposes right now. Making it look different than a WG draft was the spirit of the question. Sounds like that would be fine with you. Agreed? |
| 21:30 | <Hixie> | sure |
| 21:31 | <Hixie> | could even be more or less identical to the whatwg version |
| 21:31 | <Hixie> | though without the scripts, i guess |
| 21:31 | <rubys1> | Next question, would you want to keep this as is, or would you plan on changing it: http://www.whatwg.org/specs/web-apps/current-work/multipage/introduction.html#is-this-html5? ? |
| 21:32 | <Hixie> | it would presumably have to be clarified a bit |
| 21:32 | <rubys1> | Last question: |
| 21:33 | <rubys1> | when we last talked, I proposed http://dev.w3.org/html5/2dcontext-hixie/, given what you are intending, is the following OK as a URL: http://dev.w3.org/html/spec-hixie/ |
| 21:34 | <rubys1> | note html, not html5, not html-next, simply html. |
| 21:36 | <Hixie> | while that would be ideal in retrospect, i think it's probably best from a pragmatic position if we just leave all the html stuff and derivatives in the html5/ directory |
| 21:36 | <Hixie> | makes life easier when dealing with cvs tree fragments |
| 21:37 | <Hixie> | (spec-whatwg would probably be more in line with how the spec is described, too) |
| 21:37 | <Hixie> | (there's a number of specs in dev.w3.org/html5/... that are not what the w3c calls "html5" but fall under the buzzword "html5", e.g. websockets, iirc) |
| 21:38 | <Hixie> | (so i don't think it'd be especially controversial to keep using html5/...) |
| 21:40 | <rubys1> | I'm just gathering data at the moment. This may or may not be controversial; at the moment, I just want to know what is being proposed. In any case, I've got the answers I need to proceed. If you think of anything more, feel free to ping me. |
| 21:41 | <Hixie> | k |
| 21:41 | <Hixie> | rubys1: what would the purpose be of having a copy of the whatwg spec in dev.w3.org? |
| 21:41 | <Hixie> | rubys1: i don't have a problem doing it, but it does seem like it might be a bit confusing for people |
| 21:42 | <Hixie> | rubys1: (and we do have a lot of confusion surrounding this already, sadly) |
| 21:42 | <rubys1> | I thought it would be controversial because of exactly that reason. But since you proposed it, I followed through and asked around, and didn't find any pushback. In fact, some liked the idea. |
| 21:44 | <Hixie> | rubys1: oh i wasn't so much proposing it as saying that i would be happy to do it if you thought it would address some need. i'm just unclear on what need it solves. |
| 21:44 | <Ms2ger> | Hixie, so... Creating a new StorageEvent does set e.key to the empty string in Chrome and Gecko |
| 21:45 | <Hixie> | Ms2ger: yeah, i figured that might change things. seems better to just keep things consistent though, long-term. |
| 21:45 | <Ms2ger> | Works for me, just wanted to make sure you're aware |
| 21:46 | <Hixie> | yeah, i was. thanks for checking though. |
| 21:46 | <Hixie> | i think this is only the second or so issue that's arisen from that huge patch heycam did way back when we added nullable annotations to html |
| 21:47 | <Hixie> | pretty impressively low rate of errors |
| 21:47 | <Ms2ger> | I wish I had such a rate for my own specs :) |
| 21:48 | <Hixie> | that's all heycam; my own changes of that nature have way more bugs :-) |
| 21:48 | <Hixie> | that's why i didn't want to do it in the first place :-) |
| 21:49 | <zcorpan> | Ms2ger: speaking of spec bugs, do you consider http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1443 a spec bug? |
| 21:53 | <Ms2ger> | Fascinating case |
| 21:57 | <Ms2ger> | So the spec expects a 'foo--' element and a comment whose value is 'y</foo' as siblings? |
| 21:58 | <Ms2ger> | I'd say probably a spec bug, yes |
| 21:59 | <zcorpan> | well the spec says to copy over the children of the root element |
| 21:59 | <zcorpan> | the comment isn't a child of the root element |
| 21:59 | <zcorpan> | so it gets dropped on the floor per spec |
| 22:00 | <zcorpan> | but one might expect this to throw an exception |
| 22:01 | <Ms2ger> | I would |
| 22:01 | <Ms2ger> | Oh, that's even better |
| 22:02 | <Ms2ger> | Hixie, was that your intention for http://www.whatwg.org/specs/web-apps/current-work/multipage/the-xhtml-syntax.html#xml-fragment-parsing-algorithm ? :) |
| 22:02 | <Hixie> | hm? |
| 22:03 | <Hixie> | what's the question? |
| 22:03 | <Ms2ger> | http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1443 |
| 22:04 | <Ms2ger> | The spec parses "<foo-->" + input + "</foo-->" |
| 22:04 | <Hixie> | what browser should i be testing that in? chrome gives me a TypeError |
| 22:05 | <Ms2ger> | Gecko follows the spec |
| 22:06 | <Hixie> | won't you abort on step 5? |
| 22:06 | <Hixie> | with a SyntaxError? |
| 22:06 | <zcorpan> | no |
| 22:06 | <Hixie> | why not |
| 22:06 | <Hixie> | oh sorry |
| 22:07 | <zcorpan> | innerHTML prepends the string with the start tag and appends with end tag |
| 22:07 | <Hixie> | i thought foo == <html/> |
| 22:07 | <Hixie> | my bad |
| 22:08 | <Hixie> | i cannot honestly say i'd considered this case |
| 22:08 | <Hixie> | jesus |
| 22:08 | <Hixie> | who came up with this :-P |
| 22:09 | <zcorpan> | take a guess :-P |
| 22:09 | <Hixie> | me? :-P |
| 22:09 | <zcorpan> | nope :-) |
| 22:09 | <Hixie> | i think it would make sense to have the fragment parsing algorithm check for other siblings and throw |
| 22:09 | <Hixie> | possibly throwing a WhatWereYouThinkingError |
| 22:10 | <Ms2ger> | OperaQAError |
| 22:10 | <zcorpan> | heh |
| 22:10 | <zcorpan> | LOLzcorpanDontTestAllFuckingEdgeCasesOKError |
| 22:10 | <Hixie> | alternatively, between steps 3 and 4 have it check that you're not in the middle of a token |
| 22:10 | <Hixie> | and if so bail |
| 22:11 | <Ms2ger> | Alternatively, have annevk write an XML spec with a fragment parsing algorithm |
| 22:12 | <Hixie> | i'm happy to do either of the two changes i describe, or have anne take it over if he's not too busy with other specs |
| 22:12 | <Hixie> | either way, file a bug |
| 22:12 | <Hixie> | i agree we don't want what is there now |
| 22:12 | <Hixie> | PS. mark the bug "trivial" :-P |
| 22:13 | zcorpan | files |
| 22:33 | <gsnedders> | zcorpan: a && b returns a if a is false |
| 22:33 | <gsnedders> | zcorpan: (otherwise it returns b) |
| 22:34 | <gsnedders> | zcorpan: i.e., it returns a value x st ToBoolean(x) = ToBoolean(a) && ToBoolean(b) |
| 22:34 | <gsnedders> | Wait, no, I can't read. |
| 22:34 | <gsnedders> | That's not you! |
| 22:34 | <gsnedders> | Gah! |
| 23:50 | <Hixie> | does anyone implement startOffsetTime or can i rename it? |
| 23:51 | <smaug____> | can't find such thing in Gecko |
| 23:53 | <Hixie> | ok looks like nobody i tested implements it |
| 23:53 | Hixie | changes it to startDate |