| 00:27 | <ajnewbold> | gsnedders: ahoy |
| 06:29 | <jwalden> | "We think that developers will have an easier time building interoperable sites on top of IE8’s strong platform work (like [...] cross-domain requests (XDR) [...])." |
| 06:29 | <jwalden> | the sheer audacity |
| 06:47 | <BenMillard> | jwalden, you're talking about the IE blog here? http://blogs.msdn.com/ie/archive/2008/12/03/compatibility-view-improvements-to-come-in-ie8.aspx |
| 06:48 | <jwalden> | yeah |
| 06:48 | <jwalden> | I added a comment to about that effect, too, except less exasperatedly-worded |
| 06:48 | jwalden | neologizes a bit, and a bit more |
| 06:51 | <BenMillard> | "When users install Windows 7 Beta or the next IE8 update, they get a choice about opting-in to a list of sites that should be displayed in Compatibility View." |
| 06:53 | <BenMillard> | "Users who choose to get the list receive it via Windows Update packages, just like IE security updates." |
| 06:54 | <BenMillard> | "[...] the presence of a <META> tag / HTTP header “wins” [...]" |
| 06:55 | <BenMillard> | "In some cases, for a new browser, developers have to spend time to add a tag or header to make their sites compatible." |
| 06:56 | <BenMillard> | it's good that Microsoft are discussing this in detail in advance |
| 06:58 | <BenMillard> | oh, it names some sites: "[...] high-volume sites like facebook.com, myspace.com, bbc.co.uk, and cnn.com with pages that weren’t working for end-users with IE’s new standards compliant default." |
| 07:00 | <BenMillard> | "The most important thing we can do now is deliver better interoperability for a better web [...]" |
| 07:15 | <jwalden> | it's mostly just this-is-how-far-we'll-bend-over-backwards-for-sites-that-relied-on-our-bugs plus a little see-we're-trying-to-do-things-that-look-monopolistic |
| 07:15 | <jwalden> | which is all well and good, as far as that goes |
| 07:15 | <jwalden> | er |
| 07:16 | <jwalden> | look-not-monopolistic |
| 07:19 | hsivonen | wonders what the sites using XDR will be interoperable with |
| 07:29 | <hsivonen> | hmm. the opt-in list will be fun for QA |
| 07:30 | <hsivonen> | (both at Microsoft and at the sites affected) |
| 07:33 | <zcorpan> | i guess it means authors will have to make sure it works in both IE8 standards mode and IE8 compat mode even if they only want to target standards mode |
| 07:35 | <BenMillard> | zcorpan, perhaps such sites are expected to provide "a <META> tag / HTTP header"? |
| 07:38 | <zcorpan> | BenMillard: does it override the user's opt-in list? |
| 07:42 | <BenMillard> | zcorpan, apparently "[...] the presence of a <META> tag / HTTP header “wins” [...]" |
| 07:43 | <BenMillard> | zcorpan, you can read the whole thing: http://blogs.msdn.com/ie/archive/2008/12/03/compatibility-view-improvements-to-come-in-ie8.aspx |
| 07:55 | <BenMillard> | krijn, could simple mentions of domain names like example.com and subdomain.example.com get linkified in the logs? |
| 08:54 | <MikeSmith> | hsivonen: very cool to see your HTML5 Parsing in Gecko news |
| 09:20 | <gsnedders> | Philip`: I will be, provided I'm on a NXEC train (and not some temp. hired one without wifi) |
| 09:48 | <krijn> | BenMillard: probably, yes :) |
| 09:48 | <krijn> | Got a nice regex for that? |
| 09:49 | <BenMillard> | krijn, I don't |
| 09:49 | <krijn> | Me neither |
| 09:50 | <BenMillard> | krijn, TLDs are standardised, and they are preceeded by a dot when used in a simple domain |
| 09:50 | <BenMillard> | so if you have a list of valid TLDs, maybe you can use that as the start of a RegEx? |
| 09:51 | <BenMillard> | supporting ccTLDs could be a real chore, but .com and .org and simple ones like that would still help |
| 09:51 | <krijn> | Imho it's easier to select the foo.com, right click, Go to web address |
| 09:51 | <krijn> | But that's just me :) |
| 09:51 | <krijn> | Auto linking example.com doesn't make sense |
| 09:52 | <krijn> | And that's probably the most used one |
| 09:53 | <BenMillard> | krijn, it's your choice, of course |
| 09:53 | <krijn> | Trying to convince you a bit, since I'm lazy |
| 09:53 | <zcorpan> | i think it'd result in a fair amount of mislinking |
| 09:53 | <krijn> | Yeah |
| 09:53 | <jwalden> | you could make a regex from http://mxr.mozilla.org/mozilla-central/source/netwerk/dns/src/effective_tld_names.dat?raw=1 |
| 09:54 | <krijn> | I don't think it's worth it |
| 09:54 | <jwalden> | which would be as accurate as it gets these days |
| 09:54 | <jwalden> | I agree |
| 09:54 | <jwalden> | :-) |
| 09:54 | <BenMillard> | fair enough, was just an idea :) |
| 09:54 | <krijn> | Great idea btw! ;) |
| 09:54 | <BenMillard> | krijn, did you see I made demos for some other IRC log pages?: http://projectcerbera.com/!dev/irc-logs/ |
| 09:55 | <krijn> | Yeah, very colorish :) |
| 09:55 | <BenMillard> | krijn, as before, take as much/little as you want from them |
| 09:55 | <BenMillard> | having a consistent header and footer across all IRC log pages, and better navigation between pages, seemed like an improvement |
| 09:56 | <krijn> | I agree |
| 09:56 | <krijn> | This would certainly double the readers :) |
| 09:58 | <BenMillard> | lol, wow that'd be 8!! :P |
| 09:59 | <krijn> | http://krijnhoetmer.nl/stats/usage_200812.html#TOPURLS |
| 09:59 | <krijn> | It's one of my top urls though |
| 10:00 | <krijn> | (Could mean the rest is even more shitty, but I'll try not to think about that case) |
| 10:01 | <BenMillard> | krijn, wow it looks like individual pages of logs are more popular than any other page on your site! |
| 10:01 | <BenMillard> | krijn, so the /irc-logs/ area must dominate your traffic? |
| 10:01 | <krijn> | Yeah, it does |
| 10:02 | <krijn> | But I think that has to do with forwarding /irc-logs/htmlwg/ to the most recent day |
| 10:02 | <krijn> | And that /irc-logs/htmlwg/ is linked from w3.org |
| 10:02 | <krijn> | html-wg even |
| 10:02 | <krijn> | From http://www.w3.org/html/wg/ I mean |
| 10:02 | <BenMillard> | nice! I hadn't realised that linked to your logs. |
| 10:03 | <krijn> | Yay for my PR :p |
| 10:03 | <BenMillard> | I added navigation between pages after seeing Dan Connolly say what I'd been thinking for some time: http://krijnhoetmer.nl/irc-logs/microformats/20081125#l-122 |
| 10:03 | <BenMillard> | s/say/said/ |
| 10:03 | <krijn> | Yeah, prev/next day links would be handy |
| 10:04 | <annevk3> | should be quite easy to make too... |
| 10:04 | <krijn> | Not for me ;) |
| 10:07 | <krijn> | Could do some loops, currentday - 1, see if that logfile exists.. If yes, make a link, if not, check currentday - 2, etc |
| 10:07 | <krijn> | Same for nextday |
| 10:07 | <krijn> | With a max of 10 loops or something |
| 10:11 | <BenMillard> | krijn, sounds like that would handle changes in the month, too |
| 10:11 | <krijn> | Yeah, that's what it is for, mostly |
| 10:12 | <BenMillard> | neato |
| 10:12 | <krijn> | Or I could list all logfiles, sort them, pick the current one - 1, and the current one + 1 |
| 10:12 | <krijn> | (Which is what I do on the homepage) |
| 10:18 | <yecril71> | I would rather recommend a DOM transformation to convert a set of check boxes to a drop-down list. |
| 10:18 | <yecril71> | That, of course, relies on innerHTML to work. |
| 10:19 | <yecril71> | Structured approaches work better with structured languages than regular expressions. |
| 10:20 | <BenMillard> | yecril71, where is that being done? |
| 10:20 | <yecril71> | Regular expressions are for simple validation, and for languages that have no public DOM. |
| 10:21 | <yecril71> | Date: Tue, 2 Dec 2008 07:48:15 +0000 (UTC) |
| 10:21 | <yecril71> | Subject: Re: [whatwg] [WF2] action="mailto:" - encoding spaces |
| 10:22 | <BenMillard> | oh, an e-mail on a list, not what krijn and I were talking about. ok :) |
| 10:22 | <krijn> | Not really, no :) |
| 10:22 | <krijn> | A mailing list is so outdated! |
| 10:23 | <krijn> | (And hard to keep up with, as well) |
| 10:23 | <BenMillard> | krijn, I'm off for dinner now, then I'll resume neatening up my idea for the "day" log. Kudos for being so open to suggestions. |
| 10:23 | <krijn> | Np, sorry I'm not using everything immediately :) |
| 10:24 | <yecril71> | Absolutely, it is an antique way of getting things done, and extremely inefficient. |
| 10:26 | <yecril71> | I would still keep replying on the list but my remarks have been criticised as unintelligible and unconstructive. |
| 10:26 | <yecril71> | So Hixie told me I should rather interrupt your discussions on IRC. |
| 10:34 | <jgraham> | IRC: where being unintelligible and unconstructive is considered normal |
| 10:37 | jgraham | notes that listing "The body element" under "obsolete elements" is a bit confusing |
| 10:39 | annevk3 | notes that it has been noted already |
| 10:41 | jgraham | notes that he expected it had already been noted but thought it was worth noting again just in case |
| 10:47 | krijn | scratches his notes |
| 10:49 | annevk3 | notes he would have noted that he expected jgraham would have noted it already, but wanted to spare people on the amount of notes |
| 10:49 | <krijn> | Duly noted |
| 10:50 | <hsivonen> | thanks for the kind comments on the Gecko build |
| 10:52 | <jgraham> | Hmm am I misreading the spec... it looks as if </br> should create a <br> start tag |
| 10:52 | <zcorpan> | jgraham: that's correct |
| 10:53 | <roc> | hsivonen: context? |
| 10:53 | <hsivonen> | roc: http://hsivonen.iki.fi/html5-gecko-build/ |
| 10:53 | <jgraham> | zcorpan: Did that change recently? |
| 10:53 | <zcorpan> | jgraham: no |
| 10:53 | jgraham | wishes for a way to find out when a section of the spec last changed |
| 10:54 | <roc> | cool |
| 10:54 | <roc> | ooh, you're supporting Hixie's SVG and MathML inclusion? |
| 10:54 | <roc> | that'll put the cat among the pigeons |
| 10:55 | <hsivonen> | roc: thanks. yes, SVG and MathML are supported |
| 10:55 | <doublec> | amazon S3 came up as an example during the 'video element and cross domain' debate as a service that doesn't support access controls |
| 10:55 | <doublec> | it turns out it does provide the functionality to add custom headers |
| 10:55 | <doublec> | so as long as the headers don't need dynamic data, you could use it to host cross domain files |
| 10:56 | <roc> | intriguing |
| 10:56 | <annevk3> | Access-Control might need dynamic data if the video is personalized based on a cookie |
| 10:56 | <doublec> | yes, true |
| 10:56 | <annevk3> | but if it's simple storage Access-Control-Allow-Origin:* would suffice |
| 10:57 | annevk3 | doesn't really know what S3 allows for |
| 10:57 | <doublec> | only setting a header and string value when uploading the file, which gets served as an http header when a client downloads it |
| 10:58 | <annevk3> | so you can't really have personalized video hosting there anyway |
| 10:58 | <doublec> | correct |
| 11:03 | zcorpan | tests hsivonen's gecko build |
| 11:03 | <zcorpan> | seems to work :) |
| 11:04 | <zcorpan> | hsivonen: http://simon.html5.org/articles/mobile-results is rendered in quirks mode |
| 11:06 | <zcorpan> | or is document.compatMode lying? |
| 11:07 | <hsivonen> | zcorpan: oops. it always renders in the quirks mode. |
| 11:07 | <zcorpan> | heh |
| 11:08 | <hsivonen> | zcorpan: blog post fixed. thanks |
| 11:17 | <yecril71> | MSDN says: The value of a form control is encoded so that spaces and special characters are converted to their ASCII equivalents. |
| 11:17 | <zcorpan> | pages seem to render surprisingly well in quirks mode |
| 11:18 | <yecril71> | Guess what the ASCII equivalent for a space should be, given that a space is a valid ASCII character :-) |
| 11:19 | <zcorpan> | hsivonen: some pages suffer from the FOUC |
| 11:19 | <zcorpan> | e.g. csszengarden |
| 11:20 | <yecril71> | The example is oPhrase=My%20favorite%20color%20is%20plaid so that must be %20. |
| 11:20 | <yecril71> | In particular, converting a space to + is not mentioned at all. |
| 11:22 | <Lachy> | hsivonen, in the HTML5 gecko build, there's a strange bug with the handling of numeric character references |
| 11:23 | <Lachy> | hsivonen, see the To: field on this post for an example: http://lists.w3.org/Archives/Public/public-html/2008Dec/0074.html |
| 11:24 | <Lachy> | hsivonen, note how it says: "HTML WG <w3.or..." whereas the other email addresses, which are also in the source as char refs, display proplerly |
| 11:27 | <zcorpan> | Lachy: if i paste the source into the live dom viewer then it doesn't happen |
| 11:42 | <yecril71> | IE7, however, converts a space to + and encodes + as %2B as required. |
| 11:47 | <Lachy> | hsivonen, here's a minimified demo showing the bug http://tinyurl.com/5z6phg |
| 11:49 | <Lachy> | hsivonen, note that the behaviour seems to change depending on the length of the source code before it. i.e. if you remove one line of source code, it's likely to display a different set of character references wrongly |
| 11:54 | <zcorpan> | Hixie: at least 1 web developer thinks <iframe seamless> is a good thing :) http://friendlybit.com/html/html-includes/ |
| 12:17 | <yecril71> | I wonder if it is appropriate for the user agent to display a HTML document upon navigating to a mailto: URI. |
| 12:18 | <yecril71> | I think Lynx does that. |
| 12:20 | <roc> | Firefox can do that |
| 12:20 | <yecril71> | I also thing inappropriate appearance directives for FORM controls should be ignored. |
| 12:20 | <roc> | a Web app can register a page as the mailto: protocol handler |
| 12:21 | <MikeSmith> | yecril71: my Lynx launches some curses-based interactive app for actually sending mail |
| 12:21 | <MikeSmith> | or maybe it's built into Lynx |
| 12:26 | <annevk3> | http://www.bluishcoder.co.nz/2008/12/srt-subtitles-with-html5-video.html |
| 12:26 | <annevk3> | fail |
| 12:27 | <annevk3> | (not the post) |
| 12:28 | <MikeSmith> | annevk3: the jquery.srt doesn't work as expected? |
| 12:28 | <annevk3> | MikeSmith, the approach to subtitles was supposed to be subtitles embedded in the video stream, rather than some ECMAScript hack |
| 12:29 | <MikeSmith> | I see |
| 12:29 | <annevk3> | hopefully it will work long term, but I'm sceptic |
| 12:33 | <krijn> | Subtitles/captioning embedded in videos doesn't work, people want them to be indexable right now as well.. |
| 12:34 | <hsivonen> | annevk3: I had a glass half-empty feeling, too :-/ |
| 12:35 | <krijn> | http://www.minvws.nl/en/video/ also uses SRT afaik |
| 12:36 | <hsivonen> | for the record, I think SRT is the way to go--but as a built-in feature without JS |
| 12:37 | <hsivonen> | when it comes to captioning, SRT vs. anything fancy is even more drastic than 80% of use cases with 20% the effort |
| 12:37 | <krijn> | Something like <video subs=foo.srt> you mean? |
| 12:38 | <hsivonen> | krijn: or SRT-equivant data in the media stream |
| 12:38 | <krijn> | Or subs=foo (for mookid) |
| 12:39 | <hsivonen> | SRT vs. fancy is probably closer to 99% of use cases for 1% of effort |
| 12:39 | <hsivonen> | s/equivant/equivalent/ |
| 12:41 | <Lachy> | SRT may be ok for some simple captioning/subtitling, but it does have a lot of limitiations |
| 12:42 | <hsivonen> | Lachy: right, I didn't say 100% of use cases :-) |
| 12:43 | <Lachy> | Joe Clark was advocating open captions on ALA recently, based on the limitations (especially fonts) of closed caption formats |
| 12:43 | <Lachy> | another alternative is the bitmap subtitle formats, such as those used on DVD and Blu-ray |
| 12:44 | <hsivonen> | Joe Clark may have unusually strong sensitivities to getting particular fonts displayed |
| 12:44 | <Philip`> | Would plain transcripts (not temporally linked to the video at all) satisfy many of the use cases, with much reduced effort? |
| 12:44 | <Lachy> | he has a point about using fonts that are optimised for readability on screen |
| 12:45 | <hsivonen> | Lachy: and OSs don't ship with any screen-readable fonts? |
| 12:45 | <Lachy> | hsivonen, I don't know. |
| 12:46 | <jmb> | Philip`: they may well do. although, there's loads of non-verbal stuff in video which should probably be conveyed somehow |
| 12:46 | <hsivonen> | Philip`: well, the TV and cinema people have assumed you need at least the SRT level of features |
| 12:47 | <Philip`> | Lachy: Does he only care about accessibility of professionally-produced videos, and not care about other people producing videos when they don't know anything about fonts and don't have any installed and decide to go with Comic Sans because it looks nice and friendly? |
| 12:47 | <Lachy> | anyway, I think the ultimate solution would be to use text-based subtitle format with the ability to embed custom fonts within the video container |
| 12:47 | <Lachy> | Philip`, for amature videos, I'm sure he's aware that there isn't much chance of getting people to use anything much more complex than SRT, if they'll even use that |
| 12:48 | <Philip`> | hsivonen: The TV and cinema people have a vested interest in not producing a solution that removes the video component :-) |
| 12:48 | <hsivonen> | for now, I'd settle for SRT rendered in white with black outline near the bottom of the frame rendered in my browser-defalt sans-serif font |
| 12:50 | <Lachy> | hsivonen, sure, but it would also need to support basic positioing, so that it can be moved to the top to avoid overlapping other on-screen text |
| 12:50 | hsivonen | notes that Joe Clark relies on the user's computer having fonts for his blog instead of sending huge Microsoft Publisher-created bitmaps |
| 12:51 | <Lachy> | that's because using bitmaps like that would be less accessible than using even poor quality fonts |
| 13:16 | <MikeSmith> | I notice that the Moz and Webkit default stylesheets use -moz-padding-start: and -webkit-padding-start: instead of just padding: ... wondering why |
| 13:18 | <MikeSmith> | also -moz-margin-start and -webkit-margin-start, and -moz-border-start |
| 13:19 | <mstange> | MikeSmith: https://developer.mozilla.org/En/CSS/-moz-padding-start |
| 13:19 | <MikeSmith> | mstange: thanks |
| 13:20 | <annevk3> | yeah, those properties should just be standardized |
| 13:21 | <annevk3> | otoh, if all those properties end up in browsers and tools and such, authors might get confused quite a bit |
| 13:22 | <yecril71> | The encoding algorithm for mailto links, as specified, consists of two steps. |
| 13:22 | <yecril71> | The first step is encoding the control values normally, and the second step is to transform all occurrences of + to %20. |
| 13:23 | <yecril71> | This is equivalent to making an exception in the first algorithm. |
| 13:25 | <BenMillard> | MikeSmith, your document could convert "-start" to "-left" and "-end" to "-right" in such properties, then remove the vendor prefix, and maybe say somewhere it assumes dir="ltr"? |
| 13:27 | <MikeSmith> | BenMillard: the default browser stylesheets don't actually use -end |
| 13:27 | <MikeSmith> | would it be wrong to just convert it to plain "padding" ? |
| 13:28 | <BenMillard> | MikeSmith, moz-padding-start is equivalent to padding-left when dir="ltr", afaict. |
| 13:28 | <BenMillard> | s/moz-padding-start/-moz-padding-start/ |
| 13:28 | <MikeSmith> | I see |
| 13:30 | <yecril71> | It seems "jump to step" would be more readable if the steps were both numbered (for easy location) and labeled (for remembering what they do). |
| 13:32 | <MikeSmith> | BenMillard: so I guess I'll have my script convert those to padding-left and maybe put a note somewhere explaining |
| 13:33 | <annevk3> | rtl people could claim you're biased |
| 13:35 | <MikeSmith> | annevk3: hmm, so maybe I'll just keep it -vendor-padding-start , with "vendor" in ital |
| 13:36 | <MikeSmith> | hmm, yeah, that seems better |
| 13:36 | <BenMillard> | annevk3 & MikeSmith, the document is written in English. :) |
| 13:37 | <annevk3> | BenMillard, what it describes should still be sort of i18n compatible |
| 13:37 | <annevk3> | BenMillard, just like we care about accessibility and such :p |
| 13:37 | <annevk3> | ;) |
| 13:37 | <MikeSmith> | annevk3: btw, what about "-appearance" ? any move to standardize that? |
| 13:37 | <annevk3> | that's in a standard, but the standard is not that clear |
| 13:37 | <annevk3> | css3-ui |
| 13:40 | <MikeSmith> | OK |
| 13:40 | <BenMillard> | annevk3, versions which are translated into languages with other writing directions could translate the CSS samples accordingly. |
| 13:40 | <annevk3> | the CSS samples are UA default style sheet rules... |
| 13:40 | <hsivonen> | Lachy: thanks for the minimized test case |
| 13:41 | <annevk3> | so not really samples |
| 13:42 | <hsivonen> | hmm. that's a cross-language bug |
| 13:42 | <hsivonen> | fortunately |
| 13:43 | <BenMillard> | annevk3, I'm not sure that distinction is helpful when a reader sees something which look like CSS sample CSS but which uses propeties they don't know about...but it's Mike's call. :) |
| 13:43 | <BenMillard> | s/CSS sample CSS/a CSS sample/ |
| 13:45 | <annevk3> | it's not a sample |
| 13:56 | <annevk3> | "Look ma! No namespace declarations!" :p |
| 13:56 | <annevk3> | -- http://hsivonen.iki.fi/test/svg-and-mathml-in-html.html |
| 14:02 | <MikeSmith> | hmm, looking for a default Opera stylesheet but not finding one |
| 14:02 | <annevk3> | there's not really such a thing |
| 14:02 | <MikeSmith> | ah, OK |
| 14:02 | <MikeSmith> | I see other stuff in /usr/share/opera/styles |
| 15:23 | <annevk3> | "The net result of the 3.0 generalizations is that Python 3.0 runs the pystone benchmark around 10% slower than Python 2.5. Most likely the biggest cause is the removal of special-casing for small integers. There’s room for improvement, but it will happen after 3.0 is released!" |
| 15:23 | <annevk3> | guess we better not port html5lib just yet |
| 15:23 | <annevk3> | -- http://docs.python.org/dev/3.0/whatsnew/3.0.html |
| 15:24 | <rubys> | unicode is much better in python 3 |
| 15:25 | <annevk3> | sure |
| 15:25 | <rubys> | in any case, I'd like to see hsivonen's work be used to bootstrap a C/C++ implementation with a Python binding |
| 15:25 | annevk3 | doesn't like the new print either |
| 15:25 | <annevk3> | I'd love that too |
| 15:25 | <rubys> | yea, the new print sucks, but isn't a big deal |
| 15:25 | <rubys> | I mostly use print for debugging |
| 15:26 | jgraham | is very happy about better unicode handling |
| 15:27 | <jgraham> | although html5lib has some bugs that only occur on UCS2 or UCS4 builds |
| 15:27 | <tthorsen> | I think the new print looks ok, but I haven't tried it. What's the problem with it? |
| 15:27 | <rubys> | parens are required |
| 15:27 | tthorsen | likes parens |
| 15:28 | <jgraham> | I have a hard time complaining about two extra characters |
| 15:28 | <jgraham> | I guess it may be a small usability loss |
| 15:28 | <annevk3> | I don't :) but rubys is right, it's not crucial |
| 15:29 | <rubys> | In ruby, p is a function that can be used without parens, and equates roughly to python's print(repr(...)). Very handy for debugging. |
| 15:29 | <jgraham> | rubys: Can't most ruby functions be used without parens? |
| 15:29 | <rubys> | yes |
| 15:30 | jgraham | generally likes parens around functions |
| 15:31 | <jgraham> | Or at least I found it weird when I tried Haskell. Although that could be other things too |
| 15:31 | <rubys> | the one case where it is handy is functions that return None (or nil, or void or whatever) can be used as statements. |
| 15:41 | <rubys> | hsivonen: take a look at http://rails.intertwingly.net/blog/index.html using your html5 parser enhanced minefield |
| 15:42 | <rubys> | The whatwg logo is borked |
| 16:00 | <hsivonen> | rubys: That bug has been fixed on trunk. It was a layout layer problem. |
| 16:00 | <rubys> | cool |
| 16:01 | <hsivonen> | rubys: https://bugzilla.mozilla.org/show_bug.cgi?id=459817 |
| 16:03 | <rubys> | oh, cool, I'm the testcase! :-) |
| 16:07 | <Philip`> | Python 2.x lets you say "print x," to prevent it printing a newline at the end; can you do something similar in Python 3.0 if print is using standard function call syntax instead? |
| 16:08 | <tthorsen> | print(x, end=""), I think |
| 16:09 | <Philip`> | That's quite horrid |
| 16:09 | Philip` | prefers it when programming languages are updated to make it easier to write things, not harder |
| 16:09 | <rubys> | http://docs.python.org/dev/3.0/whatsnew/3.0.html#print-is-a-function |
| 16:12 | gsnedders | should probably care about supporting Python 3 now |
| 16:12 | <gsnedders> | does html5lib work at all? |
| 16:13 | <jgraham> | gsnedders: No |
| 16:13 | <tthorsen> | Philip`: I did not know about the "print x," in 2.X. That's pretty weird syntax - even if it is easy to write. I think the new syntax is easier to understand for people who are not Python wizards |
| 16:13 | <jgraham> | That is pretty low on my priority list though, after "stop hunderds of testcases failing" |
| 16:16 | jgraham | made a start on the ferry but has some way to go |
| 16:16 | <jgraham> | Like before it was over a thousand testcases |
| 16:17 | <Philip`> | tthorsen: It isn't any easier for those people to write - they'll do "print('hello')" and not want a newline, then they'll see that it adds a newline, and then they'll either give up or they'll have to look in the documentation, which will then tell them "write 'print "hello",' to suppress the newline" or "write 'print("hello", end="")' to suppress the newline" |
| 16:17 | <Philip`> | so it's equally a pain both ways |
| 16:17 | <jgraham> | sys.stdout.write("hello") works I guess |
| 16:18 | <rubys> | only if you import sys :-) |
| 16:18 | <Philip`> | but with the new way, when you're feeling very frustrated at the lack of a decent debugger and are trying to sprinkle a load of print commands throughout your code to work out what's going on, you'll have to type more characters on every line and you will get even more frustrated :-) |
| 16:19 | <tthorsen> | yeah. I was thinking about reading code that others had written. If I saw "print x," somewhere I would probably not even notice the ',' and I would certainly not guess that it was for suppressing the newline |
| 16:20 | <jgraham> | import sys; p = sys.stdout.write |
| 16:20 | <tthorsen> | If I was in charge, print would not output newlines at all. I don't mind putting a \n at the end of my strings. |
| 16:20 | <tthorsen> | s/newlines/automatic newlines/ |
| 16:21 | <Philip`> | The rationale at http://www.python.org/dev/peps/pep-3105/ seems to be concerned entirely with relatively uncommon advanced requirements, and totally ignores the basic "I just want to print some text" use case |
| 16:22 | <Philip`> | tthorsen: Recent versions of Perl attempt to solve that problem by having 'print' which does no newlines, and 'say' which adds one automatically |
| 16:22 | <rubys> | That's Python for you: one size fits all. If you want a language which cottons to use cases (at the expense of consistency), consider Perl or Ruby. |
| 16:23 | <annevk3> | I very much like significant whitespace though |
| 16:25 | <Philip`> | I think the main reason I use Python nowadays is ctypes, because I can't anything that works that well in Perl |
| 16:25 | <Philip`> | and if I'm writing a little bit of code that requires ctypes, I might as well write the rest of the program in Python because that's easier than mixing languages |
| 16:25 | <tthorsen> | annevk3: I'd actually prefer c-style { and }. The whitespace blocks are okay until you want to refactor a 3000 line program. Moving and reindenting big blocks of code feels dangerous when the indentation actually matters. |
| 16:25 | <annevk3> | I don't write 3000 line programs |
| 16:26 | <Philip`> | $ wc -l html5parser.py |
| 16:26 | <Philip`> | 2169 html5parser.py |
| 16:26 | <Philip`> | It's close :-p |
| 16:27 | hsivonen | wonders if PyDev supports all the eclipse refactoring goodness for Python |
| 18:21 | <Philip`> | When html5lib's sanitizer says things like re.match("^(\s*[-\w]+\s*:\s*[^:;]*(;|$))*$", style) isn't that, like, totally wrong, because it needs escaped backslashes or needs to be r"..."? |
| 18:22 | <Philip`> | Oh, maybe not - it seems Python treats "\w\s" identically to "\\w\\s" |
| 18:23 | <Philip`> | (Seems fragile to rely on that, though) |
| 18:25 | <Dashiva> | I never saw a reason not to use raw strings for regexps |
| 18:32 | <Philip`> | For today's exciting "you have two problems" quiz: How do you make re.match('^(\s*[-\w]+\s*:\s*[^:;]*(;|$))*$', 'x: y; ' * 20) not take a probably-exponential amount of time to run? |
| 18:34 | <Philip`> | Bonus points if the solution is still adequately fast when the input is ('x: y; ' * 20)+'x' |
| 18:34 | <Dashiva> | Could you put \s inside the [^:;]? |
| 18:35 | <Dashiva> | Not sure what you're matching exactly |
| 18:36 | <Philip`> | Dashiva: It's trying to sanitize style attribute values |
| 18:37 | <Philip`> | so it's matching "x: y; x: y" and "x: y; x: y;" (with various whitespace, and limitations on 'x' and 'y') |
| 18:37 | <Philip`> | and if it fails to match when it reaches the end, it apparently does a huge amount of backtracking |
| 18:38 | <Dashiva> | This part here seems like the problem as far as I can tell: \s*[^:;]* |
| 18:38 | <Dashiva> | Otherwise there isn't much to backtrack on |
| 18:39 | <Dashiva> | Although that wouldn't stop it from trying, would it... |
| 18:40 | <Philip`> | Hmm, that seems a good point - that's where it's going to get the exponential number of matches from |
| 18:41 | <Dashiva> | Could you just remove the \s part? Since it'd be part of the class anyway |
| 18:42 | <Philip`> | Yep, that's what I tried and it seems to make it go fast |
| 18:43 | <hsivonen> | http://delicious.com/url/d01372fa73da4044019a8ce733974a5e |
| 18:43 | <hsivonen> | tagged lol, omg, wtf |
| 18:45 | <MikeSmith> | "it is becoming increasingly difficult to distinguish W3C specs from Onion articles" |
| 18:47 | <jcranmer> | Philip`: regular expressions are not NP-hard |
| 18:47 | <jcranmer> | well, PCRE are |
| 18:50 | <Philip`> | jcranmer: That doesn't stop implementations sometimes taking exponential time to execute them |
| 18:51 | <jcranmer> | it depends on whether or not you want to refer to subgroups or just match |
| 18:52 | <Philip`> | jcranmer: Just matching, but in an implementation that does backtracking rather than compiling to a DFA or whatever silly thing computer scientists say you should do |
| 18:53 | <jcranmer> | nondeterministic fininte automotan implicitly constructing the DFA |
| 18:59 | <jcranmer> | Philip`: simple way to do it is this |
| 19:00 | <jcranmer> | get rid of the outermost * and just match multiple times :-) |
| 20:17 | <Philip`> | Hmm, Python's regexp engine is dumber than Perl's |
| 20:18 | <Philip`> | re.match('^(a|a)*$', ('a' * 21) + 'b') takes forever, whereas (("a" x 1000)."b") =~ /^(a|aa)*$/ is instantaneous |
| 20:18 | <Philip`> | Uh |
| 20:18 | <Philip`> | I meant (("a" x 1000)."b") =~ /^(a|a)*$/ |
| 20:20 | <Dashiva> | That's because perl cheats |
| 20:21 | <Philip`> | "Cheat" is just a term that means your opponent is being cleverer than you |
| 20:48 | <gsnedders> | Dashiva: Then all optimizations are cheats |
| 20:48 | <Dashiva> | Not all of them, just the ones that are cheats |
| 20:49 | <Hixie> | shepazu: specifically number 2 in http://lists.w3.org/Archives/Public/public-html/2008Oct/0127.html |
| 20:50 | <Hixie> | shepazu: (that e-mail is sorted in the order of what would most help html5, from most helpful to least helpful) |
| 20:50 | <Philip`> | Hixie: Did you mean those last two lines in #html-wg? |
| 20:50 | <Hixie> | er yes |
| 20:50 | <Hixie> | thanks |
| 20:52 | <annevk3> | rubys, if Decimal is serialized to a string, does it include the m at the end? |
| 20:52 | annevk3 | wonders if there's some way to detect the difference between a decimal and float |
| 20:56 | <annevk3> | some example in http://intertwingly.net/stories/2008/09/20/estest.html suggests not |
| 21:04 | <rubys> | annevk3: the current thinking is to *not* serialize it with an m |
| 21:05 | <rubys> | annevk3: you there? |
| 21:05 | <annevk3> | yeah |
| 21:05 | <annevk3> | had some issues |
| 21:06 | <rubys> | just wondering why you want to tell the difference between decimal and number? |
| 21:06 | <annevk3> | seems it would be annoying if you ever wanted to extend e.g. JSON |
| 21:06 | <rubys> | JSON is very specific: 3.1m would be a syntax error |
| 21:06 | <annevk3> | hence "extend" |
| 21:06 | <rubys> | extending JSON would be very hard given that there are a number of implementations in a number of languages. |
| 21:07 | <annevk3> | but maybe it's ok, I haven't really thought long about it |
| 21:07 | <rubys> | the thinking is that perhaps someday there would be a "use decimal" and the JSON parser would produce decimal quantities when parsing JSON |
| 21:07 | <annevk3> | you'd think that be true for XML too, and yet they shipped XML 1.0 5ed a few days ago |
| 21:07 | <Philip`> | annevk3: Just write it as a JSON string, and the consumer can decide that they want to parse it as a (decimal) number |
| 21:08 | <rubys> | Philip`: +1 |
| 21:08 | <annevk3> | true |
| 21:08 | <rubys> | The question is: if somebody puts 1.1 in JSON, do they mean 1.100000000000000088817841970012523 (which is what you get with number) or 1.1 (which is what you get with decimal)? |
| 21:09 | <Hixie> | annevk3: JSON has no defined error handling, so extending it would be a huge amount of work |
| 21:10 | <annevk3> | and yet people want it in all kinds of APIs |
| 21:28 | <annevk3> | Hixie, http://www.whatwg.org/specs/web-apps/current-work/#textFieldSelection never happened "When we define HTMLTextAreaElement and HTMLInputElement we will have to add the IDL given below to both of their IDLs." |
| 21:36 | <Hixie> | there's an XXX in both of those sections about it |
| 21:37 | <annevk3> | k |
| 23:51 | <olliej> | annevk3: congrats |
| 23:57 | <jwalden> | I ask <http://blogs.msdn.com/ie/archive/2008/12/03/compatibility-view-improvements-to-come-in-ie8.aspx#9173873>;; the response is http://blogs.msdn.com/ie/archive/2008/12/03/compatibility-view-improvements-to-come-in-ie8.aspx#9175737 |
| 23:57 | <jwalden> | absolutely ridiculous |
| 23:58 | <jwalden> | although it's not clear that's ms responding |
| 23:58 | <Hixie> | it's somewhat true, and a better situation than nothing being interoperable at all... |
| 23:59 | <Philip`> | jwalden: It seems clear that's not MS responding, because it doesn't say "[MS]" in their name |