| 02:07 | <jarek> | Hi |
| 02:08 | <jarek> | why 'no-display' and 'no-content' values for 'overflow' property are marked in red here: |
| 02:08 | <jarek> | http://www.w3.org/TR/css3-box/ |
| 02:09 | <jarek> | I mean: http://www.w3.org/TR/css3-box/#overflow |
| 02:11 | <jarek> | MDN does not seem to be mentioning them anywhere |
| 02:38 | <kennyluck> | jarek, I strongly suggest you not read that spec. I haven't seen it being discussed for a year or so. |
| 02:39 | <kennyluck> | If a statement is marked in red, that means there are issues. |
| 07:11 | <MikeSmith> | hsivonen: http://bugzilla.validator.nu/show_bug.cgi?id=875 (filed today) is yet another report caused by the misleading error+elaboration message for subtypes of the input element |
| 07:12 | <MikeSmith> | I'm wondering what you think of the idea of just allowing all the input attributes on all subtypes in the schema, and moving all the error reporting to Assertions.java |
| 07:13 | <MikeSmith> | and special-casing the elaboration message |
| 07:14 | <MikeSmith> | to output the same elaboration message from all input types |
| 07:15 | <MikeSmith> | maybe the elaboration message for input could be a <dl> list where each <dt> is an attribute name and the <dd> is a list of input types for which that attribute is allowed |
| 07:25 | <hsivonen> | MikeSmith: it least the elaboration needs to be type-specific. Maybe even the error message could be intercepted to avoid having to resort to Assertions.java |
| 07:25 | <MikeSmith> | OK |
| 08:07 | <hsivonen> | who is "james" in http://www.w3.org/2011/11/02-jquery-minutes.html ? |
| 08:07 | <hsivonen> | jgraham? |
| 08:13 | <hsivonen> | huh? Does Windows Server edition not have an H.264 Media Foundation decoder or why does <video> not work in IE9 there? |
| 08:13 | <hsivonen> | (per slides from the above minutes) |
| 08:19 | <hsivonen> | roc: did you notice that drawing DOM stuff to canvas was among the top requests paul_irish and ycats got from Web authors? |
| 08:19 | <roc> | no |
| 08:19 | <roc> | link? |
| 08:21 | <franksalim> | this one? http://paulirish.com/2011/what-feature-would-improve-the-web/ |
| 08:25 | <hsivonen> | franksalim: yes |
| 08:25 | <hsivonen> | roc: also slide 11 of https://docs.google.com/present/view?id=ajdqczcmx5pv_148ggbxbfg2&pli=1 |
| 08:25 | <roc> | easy to ask for |
| 08:25 | <roc> | not so easy to fix the security problems with |
| 08:27 | <hsivonen> | roc: right |
| 08:31 | <roc> | hsivonen: do you know anything about dvcs.w3.org? |
| 08:31 | <roc> | we actually do have a way to render DOM elements to canvas |
| 08:32 | <roc> | by the way |
| 08:32 | <roc> | canvas.drawImage(new Image("data:text/xml,<svg ...><foreignObject ...>...</foreignObject></svg>"), 0, 0); |
| 08:33 | <hsivonen> | roc: I have pushed a test to dvcs.w3.org once |
| 08:33 | <roc> | did you have to do anything special to get credentials? |
| 08:33 | <roc> | my W3C password, that works for reading member emails etc, doesn't work for dvcs |
| 08:33 | <roc> | AFAICT |
| 08:33 | <hsivonen> | roc: you need to be a participant in a WG and use the same credentials you use to see behind the Member-only paywall |
| 08:34 | <roc> | hmmm |
| 08:34 | <hsivonen> | roc: odd. you are a participant in the HTML WG after all |
| 08:34 | <roc> | am I? |
| 08:34 | <roc> | I don't know |
| 08:34 | <roc> | or does my ability to post to public-html indicate that I am? |
| 08:34 | <roc> | I'll email dbaron |
| 08:34 | <hsivonen> | roc: you are |
| 08:34 | <hsivonen> | http://www.w3.org/2000/09/dbwg/details?group=40318&public=1 |
| 09:03 | <heycam> | hsivonen, yes that was jgraham |
| 09:18 | <hsivonen> | heycam|away: thanks |
| 09:45 | <MikeSmith> | roc: if you're still around, I can help with dvcs.w3.org perms |
| 09:45 | <roc> | cool |
| 09:45 | <roc> | my username is 'rocallah' |
| 09:45 | <MikeSmith> | OK |
| 09:45 | <MikeSmith> | w |
| 09:46 | <MikeSmith> | which repo? |
| 09:49 | <roc> | hg/audio |
| 09:49 | <MikeSmith> | OK |
| 09:50 | <MikeSmith> | roc: please try it now |
| 09:55 | <roc> | here goes |
| 09:55 | <roc> | great, that worked!! |
| 09:55 | <MikeSmith> | super |
| 09:59 | <roc> | thanks |
| 10:03 | <MikeSmith> | no problem |
| 10:11 | <roc> | http://robert.ocallahan.org/2011/11/drawing-dom-content-to-canvas.html |
| 10:11 | <roc> | paul_irish: ^^^ |
| 15:02 | <manu`> | Heads-up, Data Driven Standards Community Group launches at W3C: http://manu.sporny.org/2011/data-driven-standards/ |
| 15:02 | <manu`> | and the link to the group: http://www.w3.org/community/data-driven-standards/ |
| 15:07 | <jgraham> | manu`: I agree entirely with the idea that this methodology is a good one and should be adopted, but I don't quite see what the value of a community group is in this situation |
| 16:00 | <AryehGregor> | smaug____, do you think we can get rid of Range.detach()? http://www.w3.org/Bugs/Public/show_bug.cgi?id=14591 |
| 16:08 | <smaug____> | hmm |
| 16:08 | <smaug____> | it is indeed useless |
| 16:08 | <smaug____> | or, not quite |
| 16:09 | <smaug____> | but anyway quite strange method |
| 17:14 | <dglazkov> | good morning, Whatwg! |
| 17:25 | <manu`> | jgraham: re: http://www.w3.org/community/data-driven-standards/ the advantage, imho, of a community group is to create a place where people can gather - one list that discusses how to technically achieve the goal and document how it is done. |
| 17:26 | <manu`> | We'd like to start building tools that allow folks to just plug in regular expressions that are run on a monthly/quarterly basis... we're thinking of using CommonCrawl to do the first sets of crawls |
| 17:26 | <manu`> | and then utilizing someone like 80legs.com for the next few crawls. |
| 17:27 | <manu`> | We're planning on doing a crawl to see where and how RDFa and Microdata is being used/abused... |
| 17:28 | <manu`> | I've been trying to get something like this going for about 6 months now, but every time we try to partner w/ someone, it falls through... and it's become increasingly difficult to wrangle everyone involved.... so, CG seemed like a good place to gather and promote the ideas. |
| 17:29 | <manu`> | I think we'll use the wiki to gather services and document how to write crawls for each service... probably have a github account in time for crawling template storage... so that you can just copy the crawling template (map/reduce job) to an Amazon Elastic Map Reduce instance and hit "GO!" |
| 17:30 | <manu`> | make sense? |
| 17:49 | <gsnedders> | brucel: Yeah, the postoffice brokenness is definitely new :\ |
| 17:53 | <gsnedders> | manu`: There's a few options. One is just to use data from dotdotbot, though that's a fairly old dump. |
| 17:53 | <gsnedders> | manu`: The other option is just to crawl pages from some public index like dmoz |
| 17:54 | <gsnedders> | I mean, yeah, you won't get quite the amount of data you could from Google's index, but probably still enough to be reasonable |
| 17:54 | <gsnedders> | I guess the problem is finding the pages that actually use RDFa/Microdata, though, really |
| 18:07 | <manu`> | gsnedders: do you know how much data dotdotbot has? or dmoz? CommonCrawl has tens of terabytes of data (5 billion pages?) 80legs claims to crawl the entire crawl-able web every 3 months (but does not store the data - they just process as they crawl) - there is also WebGrep - but they don't support regexes, just static string matching. |
| 18:08 | <gsnedders> | manu`: dotdotbot has just 13GB, but for most purposes it's a representive sample of the web as a whole (which probably means you'll find almost no RDFa in it). |
| 18:08 | <gsnedders> | manu`: dmoz just process as they go, but you can get all the URLs and crawl yourself enough for a representitive sample |
| 18:09 | <gsnedders> | The problem with RDFa/Microdata is going to be making sure you get a representitive sample of them and not of the web as a whole. |
| 18:09 | <gsnedders> | (or rather the subset of the web that uses RDFa/Microdata) |
| 18:09 | <manu`> | yes, true... but seeing as how nobody really has any public data yet... we have to start somewhere. |
| 18:09 | <erlehmann> | i need a kissology schema |
| 18:09 | <erlehmann> | does any one have one? |
| 18:09 | <manu`> | one of the things we're considering is running a test on each crawling service... to see how representative each one are. |
| 18:10 | <manu`> | s/one are/one is wrt to the other/ |
| 18:10 | <erlehmann> | i'd like this to work using RDFa <http://daten.dieweltistgarnichtso.net/src/wer-kuesst-wen/?json=internet-elite.json>; |
| 18:10 | <erlehmann> | ;) |
| 18:11 | <erlehmann> | but i cannot seem to find any good graphical RDF aggregators |
| 18:11 | <gsnedders> | manu`: The other alternative is just browser extentions and process what users visit |
| 18:12 | <manu`> | erlehmann - I remember somebody doing something like that w/ FOAF some time ago, but I can't remember the URL for it... |
| 18:12 | <erlehmann> | manu`, i remember lish daelnar doing the same thing with pico -w |
| 18:12 | <erlehmann> | sexchart.org/sexchart.9.43 |
| 18:12 | <erlehmann> | http://sexchart.org/sexchart.9.43 |
| 18:13 | <manu`> | gsnedders: yes, that would be a good alternative as well - but I'm suspect as to how easy it will be to have people install software to crawl the web on our behalf.... I think it's a good idea, just that writing the software and getting a community built around it seems to be more difficult than utilizing these large web crawls. |
| 18:13 | <erlehmann> | but the point is that a graphical RDFa aggregator could show a social graph quite nicely |
| 18:13 | <erlehmann> | why are the large crawls needed? |
| 18:14 | <manu`> | erlehmann: absolutely, it would be nice to have... one of our engineers tried using gource to do a visualization of some RDF data we had... it was a good idea that was never finished. |
| 18:15 | <manu`> | re: large crawls - representative sample... we're not just concerned about RDFa/Microdata... I think it would be good for the HTML5 spec... allowing features to be killed off more easily (see the latest <time> fiasco - which I agreed w/ the removal of time, but w/o data it's hard to make a case for /any/ removal) |
| 18:16 | <manu`> | re: http://sexchart.org/sexchart.9.43 - that made my eyes bleed. |
| 18:51 | <erlehmann> | time is removed? |
| 18:51 | <erlehmann> | oh noes. |
| 18:51 | <erlehmann> | i have to subscribe to the newsletters more often |
| 18:52 | <erlehmann> | manu`, the problem with relationship graphs is that every single person i know that can do more than i can do does not want to code a tool that may or may not ruin their supposedly monogamous relationship or show embarassing ex partners. |
| 18:52 | <erlehmann> | so i do not really have any help with coding. |
| 18:53 | <erlehmann> | but lots of people saying “one should be able to dispute RDF assertions” |
| 18:53 | <erlehmann> | facepalm m( |
| 19:02 | <matjas> | why does `foo & bar` (missing semicolon) render as an ampersand? |
| 19:04 | <bga_> | &<i />amp <- hack :) |
| 19:06 | <zewt> | &amp;amp;amp;amp;amp;amp; |
| 19:06 | <bga_> | %) |
| 19:07 | <matjas> | zewt: that renders as `&amp;amp;amp;amp;amp;`, which makes sense, since there’s a semicolon following the first |
| 19:07 | <matjas> | i’m just trying to understand the other case |
| 19:08 | <zewt> | i know, heh |
| 19:08 | <zewt> | matjas: i'd just guess web-compatibility |
| 19:09 | <bga_> | matjas may be its same as < tr > |
| 19:09 | <bga_> | ^ its not tag |
| 19:13 | <matjas> | http://www.whatwg.org/specs/web-apps/current-work/multipage/tokenization.html#tokenizing-character-references |
| 19:14 | <matjas> | bga_: the weird thing is that `foo & bar` is actually valid (even in HTML4) |
| 19:14 | roc | wonders if paul_irish reads IRC |
| 19:34 | <kennyluck> | matijsb, no it's not. It's a parse error. |
| 19:38 | <matjas> | kennyluck: that’s what i thought after reading http://www.whatwg.org/specs/web-apps/current-work/multipage/tokenization.html#tokenizing-character-references, but validator.nu and http://validator.w3.org/check (in HTML 4.01 strict mode) don’t complain at all |
| 19:39 | <matjas> | oh wait, my bad, validator.nu does complain |
| 19:39 | <matjas> | /ignore me! |
| 19:39 | <matjas> | but in HTML4 it seems to be valid, unless that’s a bug in the validator |
| 19:40 | <kennyluck> | I know nothing about HTML 4.01 so you might be right about the HTML 4.01 part. |
| 19:44 | <matjas> | http://validator.nu/?doc=data%3Atext%2Fhtml%3Bcharset%3Dutf-8%2C%3C%21DOCTYPE+html+PUBLIC+%22-%2F%2FW3C%2F%2FDTD+HTML+4.01%2F%2FEN%22+%22http%3A%2F%2Fwww.w3.org%2FTR%2Fhtml4%2Fstrict.dtd%22%3E%3Ctitle%3ETest%3C%2Ftitle%3E%3Cp%3Efoo%2520%26amp%2520bar&parser=html4&showsource=yes validator.nu’s HTML 4.01 validator says it’s a parse error too |
| 19:44 | <matjas> | must be a bug with http://validator.w3.org/check |
| 19:47 | <AryehGregor> | What's the best way to do new content attributes where we want values of "true", "false", and "inherit"? Like contenteditable or spellcheck? |
| 19:48 | <kennyluck> | What about "true", "false", and "auto"? I am thinking about dir=auto. |
| 19:49 | <AryehGregor> | I'm just wondering which way would be preferred for new attributes. |
| 19:50 | <Philip`> | manu`: http://philip.html5.org/data/dotbot-20090424.txt for dotbot |
| 19:50 | <Philip`> | manu`: (http://philip.html5.org/data/pages-using-rdfa.txt is seemingly RDFa-using pages from that) |
| 19:52 | <Philip`> | manu`: I think dmoz had like 5M distinct URLs when I last checked; average page size (if you have an upper limit of maybe 1MB) is about 30KB if I remember correctly, so it's like 150GB to download all those pages, which isn't particularly problematic |
| 19:54 | <Philip`> | manu`: With the dotbot data (~0.5M pages) it only took me a few minutes to parse and analyse every page, on a single quad-core machine, so no need for fancy cloud-based map-reduce processing until you have maybe an order of magnitude more pages or want instant results |
| 19:56 | <Philip`> | matjas: "&" without semicolon is listed in http://www.whatwg.org/specs/web-apps/current-work/multipage/named-character-references.html#named-character-references so it gets parsed (but I think it's always a syntax error if you use any of the ones without semicolons) |
| 19:57 | <matjas> | Philip`: mind = blown, I had missed that, thanks! |
| 19:57 | <matjas> | i also tested with raquo but that has an semicolon-less entry as well |
| 19:58 | <Philip`> | matjas: (The details of the list are due to compatibility requirements - semicolons are required wherever the spec can get away with it) |
| 20:12 | <manu-db> | Philip`: Thanks for the link... looking at it now. |
| 20:14 | <manu-db> | Philip`: Do you think that the dotbot and dmoz sample sets are large enough to give a decent representation of usage on the Web? |
| 20:15 | jgraham | wonders if it is worth the effort of trying to kill <details> |
| 20:15 | <jgraham> | Or at least putting it on ice for now |
| 20:16 | <jgraham> | (because I am not convinced that it is possible to implement well at this stage) |
| 20:16 | <jgraham> | (due to styling problems) |
| 20:17 | <paul_irish> | roc: i do read IRC :) thx for the awesome post.. badassjs already wrote it up http://badassjs.com/post/12473322192/hack-of-the-day-rendering-html-to-a-canvas-element-via |
| 20:17 | <scor> | Philip`: this data is from 2009? |
| 20:18 | <roc> | paul_irish: great, thanks! |
| 20:19 | <roc> | paul_irish: it may be possible to work around Webkit's data: URI bug by using a BlobBuilder to construct the SVG image and getting a Blob URI for it |
| 20:21 | <Philip`> | manu-db: Depends what you're measuring usage of - if you assume the data is a uniform random sample of the web, and you determine that N pages have some property, you can easily calculate the error bars on N, and I've completely forgotten the details but I think it's reasonably accurate when N is at least a few dozen, for samples of this scale |
| 20:22 | <manu-db> | Philip`: Right - college level statistics and all... but the assumption being made is that the data in dmoz and dotbot is a uniform random sample of the Web... and I'm not convinced that it is. |
| 20:22 | <Philip`> | manu-db: Yeah, they're definitely not |
| 20:23 | <manu-db> | to put it another way - we've done analysis on the Sindice database and a few others and found that even those large data sets have some bias. |
| 20:23 | <Philip`> | At least the dotbot data might be a uniform sample of some defined population (i.e. the set of pages they crawled), though not necessarily a useful population |
| 20:23 | <Philip`> | There's an infinite number of web pages so it's impossible to even have the concept of a uniform sample |
| 20:24 | <manu-db> | right... so, right now, we're looking at doing a 80+TB crawl - it'll cost us around $100 USD ... if the numbers that we find there match the dmoz / dotbot data - then we can assume that dmoz / dotbot have an acceptable randomized sample of the Web... |
| 20:24 | <jgraham> | I was going to say, what do you mean by a fair sample in this case? |
| 20:25 | <manu-db> | "fair sample" - I don't know... right now, I'm just wondering how much deviation there will be among the crawlers for the same question... like: "how may <time> elements are there in the data set?" |
| 20:25 | <Philip`> | There are various problems like dmoz having a zillion nytimes.com pages at some point in the past (if I remember correctly), which can be improved by e.g. limiting number of pages per domain |
| 20:25 | <manu-db> | (and then you divide the occurences by number of pages, etc. |
| 20:26 | <manu-db> | jgraham: I think one of the problems is that we don't know what a "fair sample" looks like... there is no metric for determining a fair sample... |
| 20:27 | <manu-db> | for the reasons that Philip` states... I tend to shrug when people ask that question. I don't have a good answer - thus the need for the Data-Driven Standards work... |
| 20:27 | <Philip`> | Probably the most useful sample is the set of pages visited by users in a day, multiplied by the importance they assign to each visit |
| 20:28 | <Philip`> | A more feasible approximation is the set of pages visited by users in a day |
| 20:28 | <jgraham> | Indeed. I can't even begin to imagine what you would say. If you show usage of <time> in 1% of all pages but those 1% are all wordpress blogs that will be upgraded with the next security release, is that a significant number of people or not? |
| 20:28 | <manu-db> | I'm sure you could factor their page rank in there somewhere. |
| 20:28 | <Philip`> | dmoz probably has a lot of bias towards old sites (because they were entered a long time ago) that nobody visits nowadays |
| 20:29 | <jgraham> | manu-db: You might be able to get that data more directly e.g. by getting a browser maker to add element counters to their data collection tools |
| 20:29 | <manu-db> | jgraham - yes, answering questions like that is difficult... usually getting data just creates more questions that you want to ask the data... |
| 20:29 | <Philip`> | dotbot probably has a lot of bias towards deep database-driven sites with large numbers of pages |
| 20:29 | <Velmont> | jgraham: Well, -- themes are not really always that easily upgraded. |
| 20:29 | <Philip`> | (and most pages are visited very rarely) |
| 20:29 | <jgraham> | (the user-determined average) |
| 20:29 | <manu-db> | jgraham: good idea - but getting browser vendors to move on stuff like this is a very long and painful process... we haven't had much luck with it in the past. |
| 20:30 | <jgraham> | Right, there's probably a bunch of reasons that's a bad idea |
| 20:30 | <Velmont> | jgraham: Also, I have <time> on all pages of universitas.no which has quite a lot of pages (a news paper). It's a fish in the sea, but I guess many othershave it.:] |
| 20:30 | <jgraham> | But it is the only thing that comes close to the rather-reasonable definition of "fair" (i.e. usage weighted) that Philip` gave |
| 20:30 | <Philip`> | If you're measuring stuff like <time>, then that's extremely rare and very recent, so I expect it'll be very hard to find and depend hugely on the sample method, whereas usage of something like longdesc has probably not changed much for a decade so it doesn't matter so much where you look |
| 20:31 | <jgraham> | Velmont: It was only an example |
| 20:31 | <Philip`> | scor: It is |
| 20:31 | <jgraham> | (but does indicate another problem with getting data from browser vendors which is that it is useless without URLs and they shouldn't hand those over) |
| 20:34 | <Philip`> | I suppose the other problem is that even when you have pretty accurate and detailed and multiply-reproduced data about e.g. longdesc, people ignore it |
| 20:34 | <manu-db> | You could artificially create some sort of relevance rank by applying the Google page rank of a domain to the URLs that you scan of that domain... but even that is pretty hand-wavingly band... that could give you a relevance value for the markup for a particular page. |
| 20:35 | <manu-db> | In any case - I don't think the questions we're going to be asking are that detailed at first. |
| 20:35 | <manu-db> | We just want to know "How many sites will break if we remove X?" |
| 20:36 | <manu-db> | and then at least we have data... that we can argue endlessly about the significance of those sites. |
| 20:36 | <zewt> | be nice if google could do things like dom-level queries, heh |
| 20:36 | <zewt> | google xpath |
| 20:36 | <manu-db> | the thing that concerns me is that we don't even have the basic set of data right now. |
| 20:37 | <manu-db> | zewt: So, I was looking into the map-reduce stuff and if there was a Python HTML5 DOM (which there is), you could do those types of queries on the Common Crawl data set. |
| 20:37 | <Philip`> | I don't think you want to use Python - Java is about a hundred times faster for this |
| 20:38 | <Philip`> | (where "this" includes HTML5 parsing) |
| 20:38 | <jgraham> | Philip`: I think you mean "I don't think you'd want to use a python *parser*) |
| 20:38 | <jgraham> | s/)/"/ |
| 20:38 | <manu-db> | Well... you can use Java too... I just try not to unless absolutely necessary. |
| 20:38 | <bga_> | :) |
| 20:38 | <bga_> | use pure C! |
| 20:38 | <Ms2ger> | You can use C++, as soon as hsivonen finishes his standalone parser :) |
| 20:39 | <Ms2ger> | (Along with HTML5 becoming a recommendation?) |
| 20:39 | <manu-db> | unfortunately, Hadoop is written in Java... which is what Amazon's Elastic Map Reduce crap uses... so, no C++ love there. |
| 20:41 | <Velmont> | The h264 js decoder, would be nice to see a theora js-based decoder, seeing as theora has lots less complexity. Then maybe I could retire using java applet cortado for showing videos to legacy browsers. |
| 20:41 | <Philip`> | manu-db: Annoyingly, trying to prove a negative ("there aren't many significant pages that will break if we remove X") seems massively harder than a positive (which you can prove by demonstrating there are N affected pages in this sample) :-( |
| 20:42 | <Philip`> | (For a negative, people will always argue your sample may be missing many significant pages, and will probably be right) |
| 20:43 | Philip` | isn't entirely sure it's worth the effort of trying to do the former |
| 20:44 | <manu-db> | Philip`: I don't quite follow, mind elaborating? |
| 20:46 | <Ms2ger> | Absence of proof isn't proof of absence, and stuff like that |
| 20:48 | <Philip`> | If you say "our search found nobody using <time> so we can safely remove it", someone will say "but this major site over here uses it", or "your sample is 3 months old and there was a load of publicity 2 months ago that will have encouraged thousands of people to use it", or "you ought to try looking in .gov sites because I have a hunch they might use it", etc |
| 20:51 | <Philip`> | (whereas if you say "this list of three hundred sites uses <time>, and based on the sample size there's probably at least a thousand times that many in the full collection that was sampled, so it's too expensive to remove it", then nobody will disagree) |
| 20:51 | <manu-db> | right |
| 20:52 | <Philip`> | (so the latter is easy and can produce usable results to help ensure compatibility in language design, but the former seems to end up frequently in endless discussions about the methodology) |
| 20:55 | <manu-db> | yes, that's true. However, having data across 80TB of data is better than not... especially if we can understand how randomized these sample sets are... |
| 20:55 | <manu-db> | I'm not saying that you won't have people saying that you sampled the wrong data set... |
| 20:56 | <manu-db> | but by having a pretty solid data set and methodology, you can convince the more rational people among us about a trend. |
| 20:56 | <manu-db> | where solid data set >= 80TB of data or 5 billion pages |
| 20:56 | <manu-db> | and methodology == the same question asked across dotbot, CommonCrawl, and 80legs.com gave roughly the same answer. |
| 20:57 | <manu-db> | (not saying that is easy to do... but it sounds better than what we're doing right now) |
| 20:57 | <manu-db> | (and it seems technically achievable for a very small investment of time and cash) |
| 20:58 | <Philip`> | Is the plan to update the data set regularly? (I'd imagine it's more useful to have one that's e.g. 10% of the size but updated every 3 months, so you can follow trends over time and detect usage of recent features) |
| 20:58 | <erlehmann> | manu-db, why not use html5lib? |
| 20:58 | <erlehmann> | for python? |
| 20:58 | <erlehmann> | i do not see what python dom stuff could do better. i feel dumb. |
| 20:58 | <Philip`> | (The dotbot data is kind of uselessly outdated now) |
| 20:58 | <manu-db> | Philip`: I think CommonCrawl updates their data twice a year... 80legs updates their data every 3 months. |
| 20:59 | <manu-db> | Philip`: It would be fun to see how much the dotbot data deviates from the frequently updated sample sets... |
| 21:01 | <manu-db> | erlehmann: Yes, you could use html5lib - except that some people say that it's slow (which translates into lots of $$$ on an Amazon Elastic Map Reduce Job on multiple terabytes of data) |
| 21:02 | Philip` | even has data saying it's slow :-) |
| 21:03 | <erlehmann> | oh |
| 21:05 | <jgraham> | It is slow |
| 21:05 | <jgraham> | This is not really an opinion :) |
| 21:05 | <devfil> | AryehGregor: hi, I'm trying to use your execCommand implementation but looks like it doesn't work on firefox :/ |
| 21:06 | <AryehGregor> | devfil, I'll need a lot more details than that to debug the issue. First, what URL are you looking at? editor.html is *not* meant to be actually usable in practice at this point. |
| 21:06 | <Philip`> | (...Or at least I did have data - it's somewhere in the IRC logs, I'm just not sure where) |
| 21:07 | <AryehGregor> | (I'm about to leave for a while, but I should be back in an hour or so, so just be patient -- or continue this discussion by e-mail) |
| 21:07 | <devfil> | AryehGregor: I'm using nicEdit but instead of calling window.execCommand I'm calling myExecCommand, it works in chrome |
| 21:07 | <AryehGregor> | devfil, it will work somewhat in recent Chrome and Firefox, fail in some cases even in them, and fail horribly in other browsers. It's really meant for testing, so I don't expect it to be reliable in other contexts. |
| 21:08 | <devfil> | AryehGregor: yes, I know |
| 21:08 | <devfil> | AryehGregor: I'm using firefox 7.0.1 |
| 21:08 | <AryehGregor> | devfil, try Firefox 9.0a2 or later. |
| 21:08 | <AryehGregor> | Probably won't matter, though. |
| 21:09 | <AryehGregor> | Also try giving me a test case or describing the exact problem. |
| 21:09 | <AryehGregor> | AFK, BBL. |
| 21:15 | <kennyluck> | How different is recent Chrome from WebKit now? I wonder |
| 21:16 | <dglazkov> | kennyluck: what do you mean? |
| 21:18 | <kennyluck> | dglazkov, well AryehGregor mentioned execCommand work somewhat in recent *Chrome* and Firefox. This makes me wonder how far is Chromimum from the WebKit trunk at the moment. |
| 21:20 | <dglazkov> | kennyluck: http://src.chromium.org/viewvc/chrome/trunk/src/DEPS tells you the current WebKit revision being used (see "webkit_trunk" variable value), and the first entry on http://trac.webkit.org/ will give you the latest WebKit revision. |
| 21:20 | <devfil> | AryehGregor: only the first time it doesn't work |
| 21:20 | <dglazkov> | kennyluck: so, right now it's about 70 revisions. |
| 21:20 | <TabAtkins> | kennyluck: Chrome pulls from the trunk, though. We don't fork, though a given release branches based on a particular trunk revision. |
| 21:21 | <kennyluck> | AryehGregor, so when you so "it fail horribly in other browser" I guess you don't include Safari running with the WebKit trunk? |
| 21:22 | <jgraham> | kennyluck: I expect AryehGregor can't run safari |
| 21:24 | <kennyluck> | TabAtkins, good to know. Thanks. |
| 21:58 | <karlcow> | http://twitter.com/0penweb |
| 22:06 | <AryehGregor> | kennyluck, Safari counts the same as outdated Chrome as far as I'm concerned. I'm talking about IE and Opera, and maybe mobile browsers. |
| 22:06 | <AryehGregor> | Also, I could run Safari for Windows on a VM if I cared enough. |
| 22:07 | <AryehGregor> | But it doesn't matter, Chrome works the same for my purposes. |
| 22:22 | AryehGregor | wonders why while (var foo = bar()) isn't allowed, while for (var foo = bar(); ...) is |
| 22:23 | <gsnedders> | AryehGregor: because it's a var statement, and for/for-in are a special case. |
| 22:23 | <AryehGregor> | Why have a special case instead of just allowing it in any similar place? |
| 22:24 | <smaug____> | AryehGregor: it is not similar place |
| 22:24 | <gsnedders> | AryehGregor: Because it's semantics are different in the for/for-in case. |
| 22:25 | <AryehGregor> | Hmm. I guess for (a; b; c) { d; } is logically the same as a; while (b) { d; c; }, so only the middle part is comparable. |
| 22:26 | <AryehGregor> | Huh, var a; a = [1, 2] && a[0]; doesn't work. Apparently it evaluates the whole expression, including the assignment, before it actually assigns? |
| 22:26 | <AryehGregor> | That's counterintuitive. |
| 22:26 | <AryehGregor> | Not like C, at least. |
| 22:27 | <Philip`> | Surely being counterintuitive is like C |
| 22:27 | <zewt> | C is one of the most intuitive languages you'll find |
| 22:27 | <AryehGregor> | C is amazingly intuitive. |
| 22:27 | <AryehGregor> | It's really, really simple. |
| 22:27 | <zewt> | ^5 |
| 22:27 | <gsnedders> | AryehGregor: logical and is evaluated by GetValue(LHS) and then GetValue(RHS) |
| 22:28 | <gsnedders> | IIRC |
| 22:28 | <AryehGregor> | And GetValue() of an assignment is the RHS of the assignment, without actually doing the assignment, I guess? |
| 22:28 | <Philip`> | Doesn't seem particularly intuitive when there's trivial stuff like "a = a++" where it's impossible to know what it'll do |
| 22:28 | <zewt> | AryehGregor: that's parsed as a = ([1,2] && a[0]), not (a = [1,2]) && a[0] |
| 22:29 | <AryehGregor> | zewt, oh! |
| 22:29 | <Philip`> | Also not intuitive: aliasing |
| 22:29 | <AryehGregor> | So I could do (a = [1, 2]) && a[0], and that would work? |
| 22:29 | <gsnedders> | AryehGregor: Yes |
| 22:29 | <zewt> | if you felt the need, heh |
| 22:29 | <AryehGregor> | Okay, now that makes sense. |
| 22:29 | <AryehGregor> | Although maybe I'll just stick with being verbose and having some extra function calls. |
| 22:30 | <zewt> | Philip`: having used C for a decade and a half or so, I've never felt the need to write "a = a++" :) |
| 22:30 | <AryehGregor> | Okay, so my reproducible crash in Chromium, which is a regression, has not had anyone pay attention to it in more than a week despite the fact that I provided detailed reproduction instructions and a crash ID? Seriously? http://code.google.com/p/chromium/issues/detail?id=101791 |
| 22:31 | <zewt> | it's just a fairly isolated (and rare, for that language) language ambiguity |
| 22:31 | <gsnedders> | zewt: Well, isn't that technically undefined? |
| 22:31 | AryehGregor | pokes dglazkov and TabAtkins |
| 22:32 | <zewt> | by spec I'm not sure, but not that I'd defend that in particular--undefined things are based--it's just fairly isolated, in my experience |
| 22:32 | <zewt> | also are bad |
| 22:32 | <zewt> | (based? fingers on autopilot, apparently) |
| 22:33 | <gsnedders> | zewt: It ES (to take an example of where that's defined), it's a no-op if a is a Number |
| 22:37 | <TabAtkins> | AryehGregor: I don't have bug editting privileges on there, but I'll poke someone. |
| 22:37 | <AryehGregor> | TabAtkins, thanks. |
| 22:37 | <gsnedders> | *In |
| 22:58 | <AryehGregor> | TabAtkins, can you reproduce the crash? |
| 22:58 | <TabAtkins> | AryehGregor: Yup, and I put my own crash id in the bug report. |
| 22:58 | <AryehGregor> | Thanks. |
| 22:58 | <TabAtkins> | AryehGregor: https://bugs.webkit.org/show_bug.cgi?id=71737 |
| 22:59 | <AryehGregor> | "You are not authorized to access bug #71737. |
| 22:59 | <AryehGregor> | " |
| 22:59 | <TabAtkins> | Oh, sorry, it's marked as a security bug. |
| 22:59 | <AryehGregor> | You should be able to CC me. |
| 22:59 | <TabAtkins> | AryehGregor: Still ayg⊙an? |
| 23:00 | <AryehGregor> | Yes, should be. |
| 23:00 | <TabAtkins> | k, you're cc'd |
| 23:00 | <AryehGregor> | Thanks. |