| 00:41 | <Hixie> | interesting |
| 00:42 | <Hixie> | setTimeout() doesn't seem to throw if you pass it non-code non-string first arguments |
| 00:49 | <weinig> | Hixie: we recently changed that WebKit |
| 00:50 | <Hixie> | webkit returns undefined |
| 00:50 | <Hixie> | firefox actually just works |
| 00:50 | <Hixie> | (and doesn't do anything when the timeout comes) |
| 00:50 | <weinig> | well, I am not sure we threw, but we now call toString on whatever is passed if it is not a function |
| 00:50 | <weinig> | so setTimeout({ toString : function() { }, 0) now works as it does in firefox |
| 00:50 | <Hixie> | that doesn't seem to be the behaviour |
| 00:51 | <Hixie> | setTimeout(true, 0); should return a number if that was the case |
| 00:51 | <Hixie> | but it returns undefined |
| 00:51 | <weinig> | Hixie: this is a very recent change |
| 00:51 | <weinig> | Hixie: in the last week or so |
| 00:51 | weinig | checks to make sure it was actually landed |
| 00:51 | Hixie | updates to the latest build in case he last updated just before it landed or something |
| 00:52 | <weinig> | yeah, javascript:alert(setTimeout(true, 0)) returns a number in ToT |
| 00:52 | <weinig> | in shipping, we return undefined |
| 00:52 | <Hixie> | i never test shipping builds |
| 00:53 | <Hixie> | shipping is just another word for "obsolete" |
| 00:53 | <weinig> | Hixie: a trunk build from a week ago would do the same :) |
| 00:53 | weinig | was playing around with setTimeout and noticed this difference so fixed it |
| 00:53 | <Hixie> | yup, it is indeed changed |
| 00:53 | <Hixie> | cool |
| 00:54 | <weinig> | happy to help :) |
| 00:55 | <weinig> | weird coincidence though |
| 00:56 | <Hixie> | well in my case it's just a matter of going down a list and this was the next item |
| 00:59 | <Hixie> | you do the same with the second argument? |
| 01:40 | <Hixie> | can anyone think of any unusual objects in JavaScript that have a [[Call]] thingy defined but aren't what you would consider regular functions? |
| 01:40 | <Hixie> | ...and which have some detectable side-effect? |
| 01:42 | <olliej> | Hixie: huh? |
| 01:42 | <olliej> | Hixie: what do you mean? |
| 01:43 | <olliej> | Hixie: you mean standard runtime functions that aren't 'pure'? |
| 01:43 | <Hixie> | well e.g. select.options can be invoked as a function |
| 01:43 | <Hixie> | like select.options('a') |
| 01:43 | <Hixie> | as well as being a regular object in its own right |
| 01:44 | <Hixie> | with select.options.length, etc |
| 01:44 | <olliej> | Hixie: all functions are objects in their own right |
| 01:44 | <olliej> | Hixie: anyFunction.length |
| 01:44 | <olliej> | Hixie: will return the number of arguments to said function |
| 01:44 | <olliej> | Hixie: that said |
| 01:44 | <Hixie> | oh i should have said in the DOM, not in JS, my bad |
| 01:44 | <olliej> | Hixie: iirc /a/("foo") in js |
| 01:44 | <olliej> | Hixie: oh |
| 01:44 | <olliej> | only the dom |
| 01:45 | <Hixie> | well in js too is fine |
| 01:45 | <Hixie> | does /a/("foo") have any detectable side-effects? |
| 01:45 | <Hixie> | does it change state when you invoke it or something? |
| 01:45 | <olliej> | Hixie: evaluating a regex puts the results on the RegExp constructor |
| 01:46 | <Hixie> | come again? |
| 01:46 | <Hixie> | wait, really? |
| 01:46 | <olliej> | Hixie: yeah |
| 01:46 | <olliej> | i discovered this last week |
| 01:46 | <Hixie> | that's awesome |
| 01:46 | <olliej> | or maybe two weeks ago |
| 01:46 | <olliej> | that made me sad |
| 01:46 | <olliej> | Hixie: just a mo |
| 01:46 | <olliej> | i'll dig up the details |
| 01:47 | <Hixie> | /.+/('eueouue'); RegExp.lastMatch |
| 01:47 | <Hixie> | that's super |
| 01:48 | <Hixie> | hahaha yes! |
| 01:48 | <Hixie> | olliej: thanks, that's what i needed |
| 01:48 | <Hixie> | setTimeout(/.*/, 0, 'test'); affects RegExp.lastMatch |
| 01:48 | <olliej> | ah ha |
| 01:48 | <Hixie> | man in this API ridiculous or what |
| 01:48 | <olliej> | Hixie: hehe |
| 01:48 | <olliej> | it made me very sad |
| 01:48 | <olliej> | Hixie: when i discovered that |
| 01:48 | <Hixie> | heycam: i need a new thing in WebIDL which means "this interface is only for JS if you're even thinking of using this in another language please go have a cold shower right now" |
| 01:49 | <heycam> | hah |
| 01:49 | <Hixie> | maybe, say, as a [HereBeDragons] interface decorator |
| 01:49 | <heycam> | can i ask, what interface is this? |
| 01:49 | <Hixie> | WindowTimers |
| 01:50 | <heycam> | huh. how come? |
| 01:50 | <Hixie> | it has setTimeout and company on it, so that i can reuse them in workers without taking all of Window with it |
| 01:50 | <Hixie> | but if you mean "why is setTimeout so bad", consider that its signature is: |
| 01:50 | <Hixie> | long setTimeout(in any handler, [Optional] in any timeout, [Variadic] in any args) |
| 01:51 | <Hixie> | and the function's behaviour depends on whether handler has a [[Call]] method or not |
| 01:51 | <heycam> | so overloading a DOMString and a Callback interface isn't enough? |
| 01:52 | <Hixie> | what's the interface that defines a callback that the object /foo/ matches? |
| 01:52 | <Hixie> | (but which the string "foo" doesn't) |
| 01:53 | <heycam> | not sure what you mean |
| 01:53 | <Hixie> | how could overloading a DOMString and a Callback interface be enough? |
| 01:53 | <heycam> | currently i think you can overload and distinguish between DOMStrings and an object of an interface |
| 01:53 | <Hixie> | setTimeout() takes anything |
| 01:54 | <Hixie> | its first argument can be { toString: function() { return 'alert(0)'; } } |
| 01:54 | <Hixie> | and you'll get an alert |
| 01:54 | <Hixie> | it can also be window.alert |
| 01:54 | <Hixie> | you'll also get an alert |
| 01:54 | <heycam> | ok maybe the rules don't exactly fit, since when you have overloading any old type won't get stringified to match the DOMString |
| 01:54 | <heycam> | iirc |
| 01:54 | <heycam> | but i should check... |
| 01:56 | <Hixie> | setTimeout(handler, { toString: function() { return '20.1'; } }) is treated the same as setTimeout(handler, 20); |
| 01:56 | <Hixie> | if you give me the webidl to do that and i'll use it, but i think our time might be better spent just writing a few lines of prose |
| 01:57 | <heycam> | yeah |
| 01:57 | <heycam> | ok |
| 01:57 | <heycam> | you want a [Language=ECMAScript] or something? |
| 01:57 | <heycam> | or a term you can use in prose? |
| 01:58 | <Hixie> | actually i kinda think [HereBeDragons] would be pretty neat. but i could work with Language=ECMAScript too if you like. |
| 01:58 | <Hixie> | i just want something that prevents the idl from compiling |
| 01:58 | <heycam> | ? |
| 01:58 | <Hixie> | so that people who use it have to go figure out wtf is going on |
| 01:59 | <heycam> | then perhaps idl is not what you're looking for? |
| 01:59 | <Hixie> | possibly |
| 01:59 | <heycam> | well, there's a slight advantage to still using it; you get the long return value |
| 01:59 | <heycam> | and the association of arguments to the argument names |
| 02:00 | <heycam> | but i guess that's about it =) |
| 02:00 | <Hixie> | it also implies that the methods are on the object, etc |
| 02:00 | <Hixie> | i think i'll keep using it |
| 02:00 | <heycam> | yeah |
| 02:00 | <heycam> | ok |
| 02:00 | <Hixie> | don't worry about it |
| 02:00 | <Hixie> | they'll have to implement this anyway i guess |
| 02:01 | <Hixie> | so they'll soon find what a... poorly designed API it is |
| 02:01 | <heycam> | are you going to supply a big idl file of all the definitions (or a zip file of separate idl files)? |
| 02:01 | <Hixie> | (not my first choice of words) |
| 02:01 | <Hixie> | wasn't planning on it |
| 02:01 | <Hixie> | but people might want it |
| 02:01 | <Hixie> | it's pretty easy to make it |
| 02:01 | <Hixie> | just strip all the blocks with class=idl |
| 02:01 | <heycam> | that's what i do currently (with a bookmarklet) |
| 02:02 | <Hixie> | though a few things will fail, e.g. WindowProxy isn't defined in idl |
| 02:02 | <Hixie> | only in prose |
| 02:02 | <heycam> | hmm |
| 02:02 | <heycam> | because it's difficult to specify? |
| 02:02 | <Hixie> | (and, well, see above are setTimeout()!) |
| 02:02 | <Hixie> | I have no idea how I'd define WindowProxy in IDL |
| 02:03 | <Hixie> | it is in every sense indistinguishable from Window |
| 02:03 | <Hixie> | and you can't get a hold of the Window object anyway |
| 02:03 | <Hixie> | so... |
| 02:03 | <heycam> | if it's indistinguishable, and you've got a definition of Window, then it should be easy to come up with a definition for WindowProxy |
| 02:04 | <heycam> | (i.e., just copy it) |
| 02:04 | <Hixie> | all it does is forward every operation to the underlying Window |
| 02:05 | <Hixie> | for example, it's prototype changes when the underlying Window changes |
| 02:05 | <Hixie> | its |
| 02:06 | <heycam> | does that mean it's impossible to specify in idl though? |
| 02:06 | <Hixie> | i think it means that it's not useful to specify in idl |
| 02:06 | <heycam> | ok |
| 02:07 | <heycam> | so what i'm concerned about is whether it'd be possible to "compile" or process however all the idl for html5 |
| 02:07 | <heycam> | if there are dangling references to interfaces like WindowProxy... |
| 02:07 | <Hixie> | there are bits that need to be provided first |
| 02:07 | <Hixie> | but after they are provided, sure |
| 02:07 | <Hixie> | (such as WindowProxy) |
| 02:20 | <Hixie> | i wonder how to spec the whole "compile the string" thing |
| 05:07 | <hsivonen> | I wonder which one of jd's posts has resonated http://twitter.com/jbowtie/status/1691112133 |
| 05:19 | <ojan> | Hixie: ping |
| 05:20 | <Hixie> | here |
| 05:21 | <ojan> | Hixie: i'm trying to see if it's worth trying to add undo/redo events to webkit... |
| 05:21 | <ojan> | Hixie: seems like they are tied to the UndoManager though...so we'd need to hash all that stuff out before moving forward? |
| 05:21 | <Hixie> | pretty much |
| 05:21 | <Hixie> | though nobody has suggested a better solution yet |
| 05:24 | <ojan> | Hixie: to be sure i'm understanding the intent, onundo/onredo would fire when the user picks undo/redo from a menu, fires the undo/redo execCommands or hits the appropriate key-command, right? |
| 05:24 | <Hixie> | yup |
| 05:34 | <ojan> | Hixie: i don't quite follow some bits here... |
| 05:34 | <ojan> | in 6.11.4 |
| 05:34 | <ojan> | "If the entry immediately before the undo position is an undo object entry" |
| 05:35 | <ojan> | what else could it be? |
| 05:35 | <Hixie> | The entries are of two type: DOM changes and undo objects. |
| 05:35 | <Hixie> | click "undo transaction history" |
| 05:35 | <Hixie> | for more details |
| 05:36 | <ojan> | right, i saw that...so, undo/redo events don't fire for DOM changes? |
| 05:37 | <Hixie> | the undo/redo events are there just to let scripts perform the changes that need doing for the relevant objects they put on the list |
| 05:37 | <ojan> | ah |
| 05:37 | <ojan> | that's what i wasn't understanding |
| 05:37 | <Hixie> | the whole section is a mess imho |
| 05:37 | <Hixie> | i don't really like it |
| 05:38 | <Hixie> | (i don't have anything better to suggest though) |
| 05:38 | <ojan> | so, the thing you pass in to the add method is any arbitrary JS object of your choosing? |
| 05:38 | <Hixie> | right |
| 05:38 | <ojan> | i see |
| 05:38 | <ojan> | i'm starting to get the intention here now at least :) |
| 05:39 | <Hixie> | hehe |
| 05:39 | <ojan> | Hixie: i feel like it would be a ton simple to just have the object you pass in be required to have undo/redo methods |
| 05:40 | <ojan> | Hixie: gives more flexibility to the web developer in terms of when they setup the undo/redo state |
| 05:40 | <ojan> | (i.e. they can do it at the time of the change or at the time of the undo) |
| 05:40 | <ojan> | I guess you can do that already with the current API |
| 05:40 | <Hixie> | how would you handle the case where an undo operation has no matching redo? |
| 05:41 | <Hixie> | i recommend sending a more concrete proposal to the list |
| 05:41 | <ojan> | sure |
| 05:41 | <Hixie> | it really wouldn't take much to convince me to change the spec |
| 05:41 | <ojan> | have you considered undo/redo events for DOM changes? |
| 05:41 | <Hixie> | anything even semi-coherent would probably be convincing enough :-) |
| 05:42 | <Hixie> | i don't think i had use cases for those |
| 05:42 | <Hixie> | so probably not |
| 05:42 | <ojan> | ok, i'll try and draft an email then... |
| 05:42 | <Hixie> | (i'm not opposed though) |
| 05:42 | <ojan> | the use-case is for apps that want to capture the undo/redo and handle it themselves |
| 05:43 | <Hixie> | that's not a use case, that's a description of what it can do :-) |
| 05:45 | <ojan> | Hixie: as an example, the google rich text editor keeps its own undo stack. right now it doesn't do the right thing if you undo/redo from the menu because the JS code doesn't know that an undo happened. it can't catch and cancel the event. |
| 05:45 | <ojan> | Hixie: anyways, I'll put this stuff in an email |
| 05:46 | <Hixie> | k |
| 05:47 | <Hixie> | (why would an editor _want_ to do things itself, assuming an ideal implementation?) |
| 05:50 | <ojan> | Hixie: yeah, that's true, the google editor would use the browser's undo if it didn't suck so much |
| 05:53 | <ojan> | Hixie: if you have an editor where the backing store is not HTML, you want to be able to update your sense of the state when an undo/redo happens |
| 05:54 | <Hixie> | if you have an editor where the backing store is not HTML, surely you just want to catch any DOM changes |
| 05:54 | <Hixie> | not necessarily undo/redo ones in particular |
| 06:02 | <ojan> | yes, but you want to be able to know what happened, and in most cases, be able to cancel the event |
| 06:03 | <ojan> | in practice, mutation events are far too noisy and hard to work with |
| 06:03 | <Hixie> | it's possible that we need to fix mutation events |
| 06:05 | <hsivonen> | Hixie: posting them on the event loop or something more drastic? |
| 06:05 | <ojan> | Hixie: in practice, mutation events are useless for anything other than a high level sense that something changed |
| 06:06 | <Hixie> | hsivonen: i haven't studied the problem, so i don't know |
| 06:06 | <ojan> | Hixie: maybe that's a bit harsh, but i've never heard of them actually being used for anything else |
| 06:06 | <Hixie> | ojan: it would be good to tell www-dom⊙wo this, as that is where the DOM3 Events spec is being discussed |
| 06:06 | <Hixie> | and mutation events are on the table |
| 06:07 | ojan | goes to join yet another mailing list |
| 06:07 | <Hixie> | yeah not sure why that's not on public-webapps |
| 06:08 | <Hixie> | (in particular, not sure why that's elsewhere but widgets are on public-webapps) |
| 06:43 | <ojan> | Hixie: yeah, the more i look at mutation events for this, the more i think that they're just at the wrong granularity for the use-case in question. you actually want something that conveys the semantics of the user-action, not something the just tells you all the details of what changed after the fact |
| 07:03 | <ojan> | Hixie: more random questions... |
| 07:04 | <ojan> | dataTransfer only supports "Text" and "URL" as types |
| 07:04 | <ojan> | is that just because that's what IE does? or is there some reason not to also support HTML? |
| 07:07 | <Hixie> | dataTransfer supports any type |
| 07:07 | <Hixie> | where does it say it only supports Text and URL? |
| 07:09 | <ojan> | under setData/getData it only list Text and URL as options |
| 07:09 | <ojan> | i guess it doesn't forbid other options |
| 07:10 | <Hixie> | oh those are just two values that are supported for legacy reasons |
| 07:10 | <Hixie> | because the value is supposed to be a mime type |
| 07:10 | <ojan> | oh, i see...so to support HTML text we would just use the mime type directly? |
| 07:10 | <ojan> | i.e. text/html |
| 07:11 | <Hixie> | yup |
| 07:11 | <ojan> | Hixie: great. we'll attempt adding such to webkit. |
| 07:11 | <Hixie> | cool |
| 07:12 | <ojan> | thanks for talking through all my random questions :) |
| 07:12 | <Hixie> | np |
| 07:21 | <Hixie> | feedback i just got from someone not (as far as i know) affiliated with any existing solution for microdata: |
| 07:22 | <Hixie> | * they'd like to share event listing data on their city "what's on" guide web site "in a way that retains the event information in a meaningful way" |
| 07:23 | <Hixie> | asked why microformats don't work for them, they had several replies that apply equally to microformats and other technologies, and one that is specific to microformats: |
| 07:23 | <Hixie> | "There are no obvious parsing tools for people to actually do anything with the data (other than add an event to a calendar)" |
| 07:30 | <zcorpan> | Hixie: <li><p>If none of the values (if any) result in an access key being |
| 07:30 | <zcorpan> | + assigned, then the element has no <span>assigned access |
| 07:30 | <zcorpan> | + key</span>.</p></li> |
| 07:31 | <zcorpan> | Hixie: why not allow the UA to make up its own shortcut in that case? |
| 07:31 | <zcorpan> | Hixie: e.g. based on textContent |
| 07:31 | <zcorpan> | Hixie: or based on the keys available in the UA |
| 07:33 | <Hixie> | i guess we could... |
| 07:37 | <zcorpan> | Hixie: there's some magic required with focusing and accesskey that i don't quite remember |
| 07:38 | <zcorpan> | Hixie: e.g. for fieldset you should focus the first descendant form control or something |
| 07:38 | <Hixie> | send mail for these issues, i'll look at them in due course |
| 08:26 | <zcorpan> | maybe accesskeyLabel should be available to css 'content' |
| 08:28 | <zcorpan> | :has-accesskey::after { content:'(' accesskey ')' } |
| 08:31 | <hendry> | how are you supposed to express last modified time (mtime) and creation time (ctime) of a page in markup? |
| 08:33 | <Philip`> | hendry: <p>Last modified: 2009-05-04.</p> |
| 08:36 | <hendry> | Philip`: so no time element? :) |
| 08:37 | <Philip`> | hendry: What's the use case for using a special element here? :-) |
| 08:37 | <zcorpan> | "You would say just "n" for now, then, when the new feature is wildly deployed, " -- Hixie expects accesskey to be *wildly* deployed? |
| 08:38 | <zcorpan> | Philip`: maybe you would add the date to your calendar so you can compare next time you look at the page if it has been modified since you last read it |
| 08:39 | <hendry> | Philip`: i'm helping out moving ikiwiki templates to html5. every page has a ctime and mtime in there, and i was wondering how best to mark it up |
| 08:41 | <hendry> | Philip`: here is an example of the markup so far http://h.dabase.com/html5-tmp/html5/ |
| 08:48 | <Lachy> | Philip`, in hendry's example, the benefit of using the time element would be that the browser can, at the user's request, present the date in a more familiar format. (Though, this depends on future UAs implementing such a feature) |
| 08:50 | <Lachy> | although, such a feature would be more useful on sites that insist on using ambiguous localised date formats like 04/05/09 |
| 08:52 | <hendry> | Lachy: thanks, that's a good use case |
| 09:02 | <jgraham> | Lachy: Hopefully UAs won't implement such a feature |
| 09:02 | <jgraham> | Since I guess it will be more confusing than the status-quo |
| 09:03 | <jgraham> | (although I suppose some good implementation is possible like displaying the time as the author wrote it and giving the option to display it in some other format) |
| 09:04 | <Philip`> | They could display the localised time in a tooltip, perhaps |
| 09:04 | jgraham | notes that js also has magic $1-$9 variables on the RegExp object which point to the first 9 match groups in the last regexp called |
| 09:05 | <jgraham> | I shoul really document that on the web ecmascript page, I guess |
| 09:07 | <Lachy> | jgraham, that's one of the major problems solved by having the time element, so I hope they do implement it in some way. |
| 09:11 | <jgraham> | Lachy: Mixing up author supplied content and UA supplied content is likely to be really confusing though |
| 09:11 | <Philip`> | Lachy: That's a problem solved by having authors write dates like "2009-05-04" instead of "04/05/09", with no need for any markup |
| 09:14 | <Philip`> | Anyway there'll be too much legacy content saying <time datetime=2007-10-05>December 3rd 2009</time> for UAs to replace the textual content with something derived from the datetime |
| 09:17 | <Lachy> | Philip`, getting authors to write dates in pages only using unambiguous date formats isn't a viable solution in practice. Letting authors use whatever date format they like like in the page, while providing a way to mark up the date unambiguously and letting UAs expose that to users somehow is |
| 09:18 | <Lachy> | jgraham, I'm not proposing that UAs replace the text in the page with the unambiguous format by default, but rather provide a way for the user to access it on request |
| 09:19 | <Philip`> | Lachy: Why is getting authors to write dates in pages only using unambiguous date formats not a viable solution in practice? |
| 09:20 | Philip` | very rarely sees pages that say "01/02/03" so it seems like most people already use unambiguous date formats |
| 09:20 | <Lachy> | because we have ~15 years of experience showing us that authors still publish ambiguous date formats |
| 09:20 | <Lachy> | I see it relatively often |
| 09:22 | jgraham | wonders if authours who use ambiguous date formats would be likely to use <time> |
| 09:22 | <Lachy> | here's a page I recall seeing http://news.zdnet.com/2100-9595_22-293658.html scroll down to the comments |
| 09:25 | <Lachy> | if authors see practical benefits for users provided by UAs by using it, then they will |
| 09:26 | <Lachy> | but if <time> is just seen as useless metadata that doesn't help users at all in major browsers, then it's less likely |
| 09:27 | <Philip`> | They would also have to think the aesthetic advantages of writing mm/dd/yy outweigh the practical benefits for users of just writing yyyy-mm-dd with no markup |
| 09:28 | <Philip`> | Actually: They would also have to think the aesthetic advantages of writing mm/dd/yy, plus the practical benefits for users of whatever their UA does with <time>, minus the cost of adding and maintaining the extra <time> markup, outweigh the practical benefits for users of just writing yyyy-mm-dd with no markup |
| 09:29 | <Philip`> | Also they would have to know about and care about the issue, which I presume many people don't |
| 09:32 | <Lachy> | Philip`, from my experience, a lot of american sites do seem to prefer the asthetics of writing mm/dd/yy over the less ambiguous yyyy-mm-dd format |
| 09:32 | <olliej> | Lachy: yeah |
| 09:33 | <olliej> | Lachy: i've actually taken to writing dd-mmm-yyyy on everything |
| 09:33 | <olliej> | simply to avoid confusion |
| 09:33 | <Lachy> | and when I was doing web development a few years ago, many of the clients/designers demanded that I use dd/mm/yy (for Australian sites) despite my objectinos |
| 09:33 | <olliej> | Lachy: (NZ does dd/mm/yy so i keep getting confused) |
| 09:33 | <Lachy> | olliej, I know |
| 09:33 | <Lachy> | so does Australia |
| 09:34 | <olliej> | Lachy: yeah, you said above |
| 09:34 | <olliej> | (i hit enter just after your comment) |
| 09:34 | <Lachy> | and, IIRC, the UK (though I could be wrong about that) |
| 09:37 | <Rik`> | why mm/dd/yy exists ? |
| 09:42 | <olliej> | Rik`: because the US exists |
| 09:42 | <Rik`> | yeah but how did they come up with that ? |
| 09:42 | <Philip`> | Lachy: (It's dd/mm/yy in the UK) |
| 09:43 | <olliej> | Rik`: they went "how does the rest of the world do this? let's do it differently" |
| 09:43 | <olliej> | Rik`: you can't talk you don't even have the right names for your months ;) |
| 09:44 | <Rik`> | hum ? |
| 09:44 | <olliej> | Rik`: i'm joking :D |
| 09:44 | <olliej> | Rik`: referring to france not using english month names :D |
| 09:44 | <olliej> | for obvious reasons :D |
| 09:45 | <Philip`> | "The mm/dd/yy format is used in: Belize, Canada, Federated States of Micronesia, Kenya, Palau, Philippines, Puerto Rico, United States" |
| 09:45 | <Philip`> | (says Wikipedia) |
| 09:45 | <Rik`> | ohh, like the rest of the world except england, australia and new zealand ? :) |
| 09:46 | <olliej> | Rik`: hehe |
| 09:47 | <Rik`> | and we're not yet talking about miles and gallons :) |
| 09:48 | <olliej> | shudder |
| 09:48 | <olliej> | hatehatehate |
| 09:52 | Philip` | wonders whether those are the big gallons or the little gallons |
| 09:54 | <Lachy> | I suspect the US went with mm/dd/yy so that the numerical order matched the order when spoken as words, like May 4th 2009 |
| 09:55 | <Lachy> | whereas Australians tend to say 4th of May, 2009, and use dd/mm/yy |
| 09:58 | <Lachy> | wow, apparently the US also do crazy street address formats by putting the apartment number after the street number/name http://en.wikipedia.org/wiki/Endianness#Clarifying_analogy |
| 10:00 | <olliej> | Lachy: yeah |
| 10:00 | <olliej> | Lachy: that took me a while as well |
| 10:00 | <olliej> | Lachy: the US is a bizarre place |
| 10:01 | <olliej> | Lachy: oh, and i say 4th May, 2009 :D |
| 10:01 | <Lachy> | I'd better get used to it since I'm moving there in July |
| 10:02 | <olliej> | Lachy: where abouts? |
| 10:02 | <Lachy> | Mountain View, California |
| 10:02 | <olliej> | Lachy: ah ha |
| 10:03 | <olliej> | Lachy: we should meet up at some point |
| 10:03 | <Lachy> | ok |
| 10:54 | <annevk42> | http://www.webdirections.org/blog/ask-the-browser-makers-pete-lepage-from-the-internet-explorer-team/ -- see question 5 |
| 11:09 | <jgraham> | So, say I had a reasonably well structured word document and I wanted to convert it to HTML in a way that roughly preserved the structure but that also added links e.g. from the TOC and from certain terms to their defining instance (which would always be a subsection). WWhat tools could I use to help me achieve that? |
| 11:13 | <Lachy> | jgraham, depending on the techniques used to create the Word document, then I've found copying and pasting from Word into Dreamweaver to be a rather successful technique for maintaining structure |
| 11:14 | <Lachy> | though it requires the use of the Styles and Formatting task pane in word to do the styles, and the use of the some HTML specific styles (available from the customisation dialog), and there's a Word specific setting in dreamweaver that needs to be turned on to get paragraphs to work right |
| 11:14 | <jgraham> | Lachy: Let's assume for the sake of argument that I don't have dreamweaver :) |
| 11:15 | <jgraham> | (I could probably download a trial version at home or something but it's not quite convenient) |
| 11:15 | <Lachy> | what HTML editor do you have? |
| 11:16 | <Lachy> | http://www.google.com/search?q=word+to+HTML+converter |
| 11:16 | <jgraham> | Lachy: emacs |
| 11:16 | jgraham | thinks that might not have been the right answer |
| 11:18 | <Lachy> | then I can't help you with that. |
| 11:19 | <jgraham> | It seems like the oo.o HTML export + a considerable amount of scripting might be a good enough solution |
| 11:23 | <Lachy> | have you tried MS Word's own HTML saving feature? |
| 11:23 | <Lachy> | it has a clean HTML option |
| 11:23 | <jgraham> | Lachy: I don't have MS Word either |
| 11:23 | <Lachy> | ok |
| 11:23 | <Philip`> | You could copy-and-paste from Word into a JavaScript editing component |
| 11:23 | <Philip`> | Oh |
| 11:24 | <Lachy> | jgraham, I have Word. If you sent me the file, I could try doing that for you |
| 11:24 | <jgraham> | On the other hand, since the document in question is readily publically avaliable it is not a problem if someone else does these things for me and sends me the result |
| 11:25 | <jgraham> | Lachy: It is the ES5 spec :) http://wiki.ecmascript.org/lib/exe/fetch.php?id=es3.1%3Aes3.1_proposal_working_draft&cache=cache&media=es3.1:candidate-draft-tc39-2009-025.doc |
| 11:28 | <Lachy> | this is going to take a while. I have to do it on my Windows machine at home over VNC, which is itself being ssh-tunnelled over a wifi connection through my imac, so the connection is a little slow |
| 11:30 | <jgraham> | Lachy: It's not a problem. Getting a TML version of the document is more of a long-term solution to the issue that I spend too long searching for stuff in the spec :) |
| 11:30 | <jgraham> | *HTML |
| 11:32 | <Lachy> | ah, ok. Good, then it can wait till after I go have some lunch |
| 11:34 | <jgraham> | Lachy: Sure. Take as long as you like. Thanks for the help :) |
| 12:13 | <Lachy> | jgraham, http://lachy.id.au/temp/ecma-tc39.zip |
| 12:16 | <jgraham> | Lachy: That gives me a 403 |
| 12:20 | <Lachy> | fixed |
| 12:23 | <jgraham> | Lachy: Thanks |
| 12:24 | <jgraham> | (at a glance it doesn't seem like the Word output is substantailly better than the Oo.o output; both will need substantial cleanup) |
| 12:25 | <jgraham> | Or rather either since I don't intend to do both |
| 12:30 | <olliej> | is it just me or does the ES5 spec not document deletability/replacability of standard properties? |
| 12:33 | <jgraham> | olliej: What do you mean standard properties? Properties have the Writable:true attribute set unless otherwise specified |
| 12:34 | <jgraham> | s/Properties/Properties of built in objects/ |
| 12:34 | <jgraham> | (see section 15, pg 98 or PDF page 112) |
| 12:36 | <jgraham> | (oh and Configurable: true) |
| 12:42 | <olliej> | jgraham: yeah, cheers for that |
| 12:42 | <olliej> | jgraham: terminology changed from DontDelete to writable:false |
| 12:46 | <jgraham> | olliej: Indeed. There's no point in releasing a new spec without a few cosmetic changes to screw with the brains of people familar with the old spec ;) |
| 12:46 | <olliej> | jgraham: hehehe |
| 13:10 | <jgraham> | zcorpan: YT? |
| 13:33 | <Lachy> | looking at a couple of rental sites for Mountain View, they all seem to share a major usability problem! They don't state whether the prices are monthly, fortnightly or weekly! :-( |
| 13:34 | <Lachy> | oh, yeah, one of them does actually. Was just hard to find |
| 13:35 | <jgraham> | Lachy: Sweden has that underlying poblem writ large; it is assumed that everyone already knows all the rules so they don' bother to state them until you violate them |
| 13:38 | Lachy | realises that finding an apartment in the right area is hard for someone who knows absolutely nothing about US geography. |
| 13:39 | <olliej> | Lachy: typically monthly |
| 13:39 | <olliej> | Lachy: and they will almost cetainly be assuming a 12 month lease |
| 13:39 | <Lachy> | yeah, 12 month lease is normal |
| 13:40 | <jgraham> | Lachy: For where? e.g. here it is rather unusual or illegal or something |
| 13:40 | <Lachy> | olliej, what sort of price range is reasonable for that area? |
| 13:41 | <olliej> | Lachy: honestly not sure about MV |
| 13:41 | <olliej> | Lachy: i have a flatmate and our aparments rent is i think $US2400/mo (in cupertino) |
| 13:41 | <Lachy> | jgraham, illegal to offer 12 month leases? wtf? |
| 13:42 | <Lachy> | in Australia, leases are generally either 6 or 12 months |
| 13:42 | <jgraham> | Lachy: I'm not sure it is actually illegal. It's more that contracts aren't usually (legally?) for fixed periods but are continuous. And it is rather hard for landlords to kick you out unless you don't pay the rent |
| 13:43 | <Lachy> | jgraham, are there minimum lease terms imposed? |
| 13:44 | <olliej> | jgraham: leases are more about ensuring a minimum time |
| 13:44 | <olliej> | rather than a maximum |
| 13:44 | <jgraham> | Lachy: I think the only requirement in our contract is that we give 3 months notice if we want to leave |
| 13:44 | <jgraham> | So that is a kind of minimum |
| 13:45 | <Lachy> | wow, 3 months is an absurdly long notice period |
| 13:45 | <olliej> | yeah |
| 13:45 | <jgraham> | Yeah, it is pretty long |
| 13:45 | <Lachy> | but I think that's like Norway |
| 13:45 | <olliej> | when i wasn't in a lease, i only had to give 2 weeks notice |
| 13:46 | <Lachy> | in Australia, the term means that neither the tennant or landlord can break the lease before that period, and then after that it continues with a requirement for 2 weeks notice |
| 13:46 | <jgraham> | It seems to be the Swedish way. The landlord has no right to get rid of you (wwhich is good) but it is rather hard to leave (which is bad) |
| 13:47 | <Lachy> | (except when the tennant does something like significantly damages the property, then the landlord can evict the tennant before the end of the lease and keep the bond) |
| 13:48 | <Lachy> | olliej, is it usually a 1 month bond as well? |
| 13:48 | <olliej> | yeah |
| 13:48 | <Lachy> | ok |
| 13:48 | <olliej> | i may have had a bit more |
| 13:48 | <jgraham> | (We didn't have to pay any deposit or anything either) |
| 13:48 | <olliej> | due to my complete absence of a credit history |
| 13:49 | <Lachy> | I should check if Opera will be providing me with temporary accommodation so I can look for an apartment when I get there |
| 13:49 | <Lachy> | I'd rather not have to rent something before I get a chance to actually go over there and see the place |
| 13:49 | <jgraham> | (and the landlady seems to be part of some collective which negotiates the rent with the tennants association and then assigns the right rent to each apartment based on size, facilities, condition, etc.) |
| 13:50 | <jgraham> | (at least that's how I understand it) |
| 13:50 | <Lachy> | I like the deal I get in Norway by staying in one of Opera's apartments :-) |
| 13:50 | <Rik|work> | in France, you have to give a 3 months notice before leaving but you can arrange with the owner if you find someone to replace you |
| 14:18 | <gsnedders> | Germany is similar to France, but it is 1 month, IIRC |
| 14:19 | <gsnedders> | Wow. It is almost two hours by bus from Stirling to here |
| 14:19 | <olliej> | erk |
| 14:19 | <olliej> | bed time i think |
| 14:19 | gsnedders | shakes his head at olliej |
| 14:19 | <olliej> | gsnedders: ? |
| 14:20 | <gsnedders> | olliej: Your sleeping his hours… |
| 14:20 | <Lachy> | gsnedders, do you need to travel to Stirling? |
| 14:20 | <gsnedders> | s/his// |
| 14:20 | <olliej> | gsnedders: are awesome? ;D |
| 14:20 | <gsnedders> | Lachy: Well, yes, but only from Stirling to here; I'll be coming from London |
| 14:20 | <gsnedders> | olliej: :D |
| 14:20 | <olliej> | gsnedders: i hate x86 assembly |
| 14:20 | <Lachy> | gsnedders, where is "here"? |
| 14:20 | <Rik|work> | Stirling, like the engine ? |
| 14:20 | <olliej> | i have been up all night fighting it |
| 14:21 | <Rik|work> | olliej is a knight fighting evil while you sleep |
| 14:22 | <Lachy> | 2 hours doesn't seem too long though. Is that about 180 to 200km distance? |
| 14:22 | <gsnedders> | Lachy: St Andrews |
| 14:22 | <gsnedders> | Lachy: No |
| 14:22 | <Lachy> | how far is it? |
| 14:22 | gsnedders | looks up |
| 14:22 | <gsnedders> | 70km or so, I'd guess |
| 14:23 | gsnedders | has to do 6502 Assembly for a practical computing assessment |
| 14:23 | <gsnedders> | How useful! |
| 14:23 | <Lachy> | google maps says 83 |
| 14:24 | <gsnedders> | Yeah, 83km |
| 14:24 | <Lachy> | that must be an incredibly slow bus, travelling at an average speed less than 50km/h |
| 14:25 | <gsnedders> | The route it takes is unusually sane for a bus around here |
| 14:25 | <Lachy> | hmm, I guess it's passing through a fair few towns on the way, which would slow it down |
| 14:25 | <gavin_> | average speed of less than 50km for a bus in a city is not "incredibly slow" |
| 14:25 | <gsnedders> | A lot of towns. |
| 14:26 | <gsnedders> | gavin_: This is certainly not a city whatsoever. |
| 14:27 | <Lachy> | gsnedders, is the maximum speed limit for a bus there about 100km/h, outside of towns? |
| 14:27 | <gsnedders> | I think it is 60mph, so yeah. |
| 14:27 | <Lachy> | like on highways, etc. |
| 14:27 | <gsnedders> | 100 is 62.3km/h |
| 14:27 | <Lachy> | wtf? You measure in miles there?! |
| 14:27 | <gsnedders> | Oh sure. |
| 14:27 | <Lachy> | why? I though only the US was backwards enough to be using miles |
| 14:27 | <gsnedders> | We also measure fuel economy in miles per gallon (imperial gallons, not US gallons), and buy petrol in litres. |
| 14:28 | <olliej> | Lachy: i believe the highway speed limit is 60 or 65mph |
| 14:28 | <Lachy> | huh? |
| 14:28 | <olliej> | Lachy: and i wouldn't plan on relying on buses/public transport |
| 14:28 | <gsnedders> | olliej: We aren't talking about CA |
| 14:28 | <gsnedders> | olliej: We're talking about somewhere more awesome. :D |
| 14:28 | <olliej> | gsnedders: oh? |
| 14:28 | <olliej> | gsnedders: not hard |
| 14:28 | <olliej> | :D |
| 14:28 | <gsnedders> | olliej: Scotland, here. |
| 14:29 | <gsnedders> | olliej: Unless you want to find my house in CA :D |
| 14:29 | <olliej> | hehe |
| 14:29 | <Lachy> | olliej, I've been warned about the poor bus services in CA before. |
| 14:29 | <olliej> | erk |
| 14:29 | <gsnedders> | http://www.stagecoachbus.com/timetables/23pckttt18082008[1].pdf doesn't ever seem to be overly slow, anywhere, it just seems to be slightly slow for the entire route and it just adds up |
| 14:30 | <zcorpan> | jgraham: pong |
| 14:30 | <beowulf> | Lachy: what's wrong with miles? |
| 14:31 | <Lachy> | beowulf, the fact that they aren't metric, and doing maths with them is hard |
| 14:31 | <olliej> | gah |
| 14:31 | <olliej> | why am i still up |
| 14:31 | <olliej> | night all |
| 14:31 | <gsnedders> | Hm, Google Maps says only 16 hours and 18 minutes on foot. |
| 14:32 | <Lachy> | beowulf, try making sense out of why there's 12 inches/foot, 3 feet/yard, and 1760 yards/mile, or 5280 feet/mile. There's just no logic there. |
| 14:32 | <gsnedders> | Lachy: You actually know that? I don't. |
| 14:33 | <Lachy> | (and I had to look those last 2 values up, cause I've got no hope of remembering them) |
| 14:33 | <Lachy> | gsnedders, what's the difference between imperial and US gallons? |
| 14:34 | <gsnedders> | Lachy: They are a different amount. |
| 14:34 | <gsnedders> | Lachy: More detailed than that, I don't know and don't care. |
| 14:34 | Lachy | looks it up |
| 14:34 | <gsnedders> | (I have software to convert it to litres, so I don't care.) |
| 14:35 | beowulf | wonders when in real life Lachy needed to know how many feet where in a mile |
| 14:35 | beowulf | lives in Ireland, on one side of the border we measure in miles, on the other, km |
| 14:36 | gsnedders | ought to go to Ireland some time |
| 14:36 | <Lachy> | beowulf, luckily, I come from a sensible country that uses the metric system, so I don't need to know. But I might need to when I go to the US |
| 14:36 | gsnedders | is going to be taking a stupid route home from Linköping, via Berlin, London, Southampton, and finally Stirling before getting back to St Andrews |
| 14:37 | <Lachy> | gsnedders, so you did get the internship at Opera? |
| 14:37 | <gsnedders> | Lachy: Yes |
| 14:37 | <Lachy> | cool |
| 14:37 | beowulf | makes note to invite gsnedders to the next whatever camp to talk about html5 or something |
| 14:37 | <Lachy> | when? |
| 14:37 | <gsnedders> | Lachy: June 29th to August 28th |
| 14:37 | <gsnedders> | beowulf: Also: get me money to go there :P |
| 14:37 | <Lachy> | ok |
| 14:38 | <beowulf> | gsnedders: would sir prefer HSS or regular ferry? |
| 14:38 | <gsnedders> | HSS? |
| 14:38 | <gsnedders> | http://en.wikipedia.org/wiki/High-speed_Sea_Service? |
| 14:38 | <beowulf> | gsnedders: high speed ... service? the catamaran thing |
| 14:39 | <gsnedders> | Ah. I wouldn't really care :) |
| 14:39 | <beowulf> | :) |
| 14:39 | <gsnedders> | Provided I'm not meant to be doing anything else at the time (which is not much next year) I don't care how long it takes |
| 14:39 | <gsnedders> | Actually, more to the point, I'm not doing anything that involves me being anywhere specific, I can live fine with my laptop and do what is needed. |
| 14:40 | <jgraham> | Lachy: Mathematics with miles: 0.1 miles, 0.5 miles, 1.0 miles 100 miles. Cars on the motorway travel at an average of about 1 mile/minute |
| 14:40 | <beowulf> | Lachy: so you'd struggle if i told how how far my house is from here in fields? |
| 14:41 | <Lachy> | beowulf, is "fields" a standardised unit of measure? |
| 14:42 | <Dashiva> | Fields medal radii |
| 14:42 | <Lachy> | Dashiva, what? |
| 14:43 | <Dashiva> | I figured that would be standardized :) |
| 14:43 | <beowulf> | Lachy, the distance from one hedge to the other? |
| 14:43 | Lachy | assumes beowulf is now just being silly |
| 14:44 | <gsnedders> | Only 10:36 from train station to arriving by ferry in Dublin :P |
| 14:44 | <gsnedders> | Oh, and five changes :P |
| 14:45 | <Lachy> | oh, wow, apparently there are 128 ounces in a gallon (US). That also makes things complicated |
| 14:46 | jgraham | wonders why gsnedders is taking such a convoluted route home |
| 14:46 | <gsnedders> | 128 is a nice round number! |
| 14:46 | <Lachy> | is there a smaller unit than a fluid ounce? |
| 14:46 | <Lachy> | gsnedders, only if doing maths in binary |
| 14:47 | <jgraham> | Lachy: Sure. But I guess you mean a smaller imperial unit |
| 14:47 | <gsnedders> | jgraham: Well, I'm dropping into Berlin to see my sister (and my probable niece), then off to London to spend the night there, to Southampton for uni open day, then off to Stirling for conference on the same evening |
| 14:47 | <Lachy> | jgraham, yes |
| 14:47 | <gavin_> | you have a probable niece? |
| 14:47 | <gsnedders> | Yes. |
| 14:47 | <jgraham> | |niece> |
| 14:48 | <gsnedders> | She might be a nephew. |
| 14:48 | <Lachy> | jgraham, actually, I mean in US units, not imperial units |
| 14:48 | <beowulf> | Lachy: a dram? |
| 14:48 | <beowulf> | Lachy: not a dram then |
| 14:48 | <Lachy> | oh, I forgot about the pint. Better find out how much that is |
| 14:48 | beowulf | knows nothing of US units |
| 14:49 | <gsnedders> | Very approx. 500ml |
| 14:49 | <jgraham> | Lachy: A pint is the unit that ber comes in. That is all that you need to know |
| 14:49 | <jgraham> | *beer |
| 14:49 | gsnedders | thinks going to Dublin would be easier by air :) |
| 14:51 | <beowulf> | gsnedders: dublin is expensive coming from the uk, euro/sterling exchange rate and then the recession |
| 14:51 | <Lachy> | jgraham, beer should come in either a Schooner or a Middy |
| 14:52 | jgraham | laughs |
| 14:52 | <gsnedders> | beowulf: Peh. I have no money any way. What difference does it make? :) |
| 14:53 | <Lachy> | jgraham, why is that funny? |
| 14:54 | <jgraham> | Uh because they are such silly names? A schooner is a type of boat and, well, if you can't see why "middy" sounds funny there is no hope for you :) |
| 15:21 | <gsnedders> | Lachy: Avoid wearing glasses. |
| 15:25 | <Lachy> | gsnedders, my choices are: wear glasses, or avoid needing to look at anything that isn't close enough to see clearly |
| 15:26 | <gsnedders> | Lachy: I do the latter. |
| 15:26 | Lachy | notes that even the computer screen at arms length in front of me is a little blurry without glasses |
| 15:29 | jgraham | did the latter for rather a long time |
| 15:31 | gsnedders | does have glasses, just only wears them when he can't do the latter |
| 15:33 | <Lachy> | I'm getting new lenses when I go to Australia next week. I was really hoping the insurance would cover the cost :-( |
| 15:47 | gsnedders | isn't really convinced by the ending of Pride and Prejudice |
| 15:49 | <Lachy> | how does it end? |
| 15:50 | <gsnedders> | It just quickly says a lot trying to wrap up all the story arcs, instead of properly ending. |
| 15:51 | jgraham | isn't really convinced that finding only 16 / 1289370 pages with a <html> element is realistic and not, say, a bug in his script |
| 15:52 | <jgraham> | (As a collary, the 1289370 number might well be wrong) |
| 15:59 | <Lachy> | jgraham, are you searching for pages that use the start tag compared with those that don't? |
| 16:00 | <jgraham> | Lachy: No, I'm just testing the code that I wrote to look at the dotnetdotcom data. My conclusion is that the code is broken. This is not really surprising :) |
| 16:08 | gsnedders | thought jgraham was a brilliant programmer ;P |
| 16:14 | <jgraham> | Because gsnedders never has bugs |
| 16:15 | <gsnedders> | No, never. |
| 16:15 | <beowulf> | not closing your lists in an <a> is a bad idea, you get the same blocks within a element thing as from packet boundaries |
| 16:18 | <beowulf> | http://carisenda.com/sandbox/blocka/ |
| 16:44 | <Philip`> | jgraham: There should be something like 425K pages, not 1.3M |
| 16:45 | <Philip`> | jgraham: By the way, I hope you're planning to process all the ~20GB of data in Python in finite time :-) |
| 16:45 | <Philip`> | Uh |
| 16:45 | <Philip`> | *not planning |
| 16:46 | gsnedders | points at the parallelizm module |
| 16:47 | Philip` | points at the orders of magnitude by which Python is slower than fast languages |
| 16:47 | <Philip`> | and the non-corresponding orders of magnitudes more processing cores that you can run it on |
| 16:48 | gsnedders | notes he has been wondering about using C to do some processing with the 20GB of data |
| 16:48 | <jgraham> | Philip`: I was planning on seeing how hard it was to process in python and then hen that failed to process it using hadoop, perhaps using python for the map/reduce functions. But that may not be a sensible plan :) |
| 16:49 | <jgraham> | Also, since the process is mainly IO bound it's not clear why using python is a bad idea |
| 16:50 | jgraham | was following the approach in http://effbot.org/zone/wide-finder.htm which did OK: http://www.tbray.org/ongoing/When/200x/2007/10/30/WF-Results |
| 16:52 | <gsnedders> | jgraham: Shouldn't you be doing it in Perl if you're basing it on WF? |
| 16:52 | <jgraham> | gsnedders: No. :) |
| 16:53 | jgraham | has to go but wants to find out why he is doing everything wrong :) |
| 17:45 | gsnedders | wonders whether we really need to treat U+000C as whitespace |
| 17:45 | <hsivonen> | http://xformstest.org/klotz/shibumiscript/2009/04/tv-raman-on-xforms-and-screen-scraping.html |
| 17:46 | <hsivonen> | the post that seemed to be about HTML5 and XHTML2 seems to have comments increasingly about XForms--not XHTML2 |
| 17:59 | <Philip`> | jgraham: It shouldn't be IO bound for the current dotnetdotcom data - there's not even 4GB, so it should fit in RAM with no IO at all |
| 18:02 | <Philip`> | jgraham: Also, IO should be on the order of 100MB/s, which seems much faster than you can decompress and parse data |
| 18:29 | <Philip`> | jgraham: (By "IO" I mean raw disk IO) |
| 18:44 | <gsnedders> | If you have a text/html document within an object element, is it within a new browsing context? |
| 18:44 | <Philip`> | Why wouldn't it be? |
| 18:44 | gsnedders | claims ignorance |
| 18:47 | <gsnedders> | Does anyone serve XHTML as application/xml or text/xml? |
| 18:48 | <hsivonen> | gsnedders: mozillaquestquest uses text/xml, IIRC |
| 18:48 | gsnedders | was hoping he could just ignore such content, oh well |
| 18:48 | <hsivonen> | gsnedders: also, I believe there's MathML content that uses application/xml for IE XSLT hooks |
| 18:50 | <rubys> | what do browsers do with xhtml served as appliction/xml or text/xml? |
| 18:51 | <hsivonen> | rubys: dispatch on namespace, but at least old versions have issues with what interfaces the document object exposes |
| 18:51 | <rubys> | what does IE do? |
| 18:51 | <hsivonen> | rubys: also, I'd expect createElement not to be magic |
| 18:52 | <hsivonen> | rubys: shows a prettyprint tree or applies XSLT if available |
| 18:52 | <gsnedders> | rubys: It treats it as any other XML document, so you can use XSLT to transform it to HTML |
| 18:52 | <rubys> | Can the link to the XSLT be in an HTTP header? |
| 18:53 | <hsivonen> | unlikely |
| 18:53 | <hsivonen> | but I don't know |
| 18:55 | <rubys> | I will run a test to verify, but to be clear: webkit + presto + gecko all are thought to dispatch based on namespace when they encounter text/xml ? |
| 18:55 | <hsivonen> | rubys: yes, except the document object is created on MIME type |
| 18:56 | <hsivonen> | which is a bug per HTML5 |
| 18:56 | <hsivonen> | I don't know what the implementation reality on that point is today |
| 18:57 | <rubys> | I wasn't thinking of starting with HTML, but testing it out on separate SVGs. |
| 18:57 | <annevk5> | differs |
| 18:57 | <annevk5> | DOM behavior for e.g. createElement differs too |
| 18:57 | <hsivonen> | rubys: If you run tests, I suggest testing what interfaces you get on document and testing what document.createElement("foo") returns |
| 18:58 | <rubys> | the first thing I would likely try is to see how jquery reacts. :-) |
| 18:59 | <rubys> | but I like the idea of svg images being like feeds were a few years ago: displayed as raw xml to backlevel browsers and as something usable by modern browsers. |
| 19:00 | <hsivonen> | rubys: you find an XML tree better user experience than a download dialog? |
| 19:03 | <Philip`> | hsivonen: You could write a polyglot document that displays the same content via XHTML and via ASCII art in the XML tree |
| 19:05 | <hsivonen> | Philip`: should there be validation for that? :-) |
| 19:21 | <annevk42> | hmm, great, any form of pasting makes my chromium build crash |
| 19:21 | gsnedders | ought to do physics, but reading Anna Karenina is awfully tempting |
| 19:22 | Philip` | wonders how Larry sees hundreds of messages when loading whatwg.org/issues (presumably without manually clicking on all the links) |
| 19:22 | <Philip`> | since I just get the 15 folders, in Chrome 1.0 and Chrome 2.0 and Firefox 3.0 and Opera 9.6 |
| 19:22 | <annevk42> | I was trying to find out, but Chromium crashes on me |
| 19:22 | <hsivonen> | I replied to Larry |
| 19:23 | <annevk42> | (when I enter the address manually I get the same as Philip`) |
| 19:23 | <Philip`> | hsivonen: I noticed, just before I sent the reply I'd written myself :-) |
| 19:23 | <Philip`> | (saying basically the same thing) |
| 19:23 | <Philip`> | annevk42: Fortunately Chromium has lots of clever process isolation so that the browser won't crash |
| 19:24 | <annevk42> | you know, not working |
| 19:24 | <annevk42> | :) |
| 19:24 | <hsivonen> | has Chrome 2.0 already been pushed to users? |
| 19:24 | <Philip`> | hsivonen: No |
| 19:24 | <hsivonen> | Philip`: do you have separate VMs for different versions of Chrome? |
| 19:24 | <Philip`> | hsivonen: (I had to manually find and install the beta) |
| 19:25 | <Philip`> | hsivonen: No; I tested in Chrome 1.0, then upgraded to 2.0 and tested again |
| 19:25 | <hsivonen> | ah |
| 19:25 | Philip` | has no VMs at all |
| 19:26 | <Philip`> | mostly because VMware on Linux started panicking my kernel after I updated something |
| 19:26 | <Philip`> | and then I decided there were more useful uses of several gigabytes of disk space than a VM which I never use |
| 19:48 | <hsivonen> | Hixie: what's your take on http://mxr.mozilla.org/mozilla-central/source/layout/reftests/bugs/228856-2.html ? |
| 19:53 | <jgraham> | Philip`: My plan was to decompress the data upfront rather than on the fly. Possibly that isn't sensible (but I'm no sure how the compressed size of the data relative to the ram size matters since presumably it has o be decompressed at some point) |
| 19:55 | <Philip`> | jgraham: I think that isn't sensible, because of disk IO speed and RAM buffer sizes |
| 19:55 | <Philip`> | The decompressed data will get processed and discarded and garbage-collected very quickly |
| 19:55 | <Philip`> | so it won't take up much RAM in total |
| 19:56 | <Philip`> | which leaves all the rest available for buffering the disk |
| 19:56 | <jgraham> | Philip`: I'm not sure I understand. Maybe I don't understand gzip well enough |
| 19:57 | <Philip`> | jgraham: You decompress and process it streamingly, you don't decompress the entire thing into RAM and then process it all at once |
| 19:58 | <jgraham> | Philip`: But that limits you to a single process operating on the data, right? |
| 19:58 | <jgraham> | (i.e. you can't look at it in chunks) |
| 19:58 | <Philip`> | jgraham: If you've got a single gzipped file, then yes |
| 19:59 | <Philip`> | so I suggest doing some pre-processing to split it into several chunks, each of which can be gzipped independently |
| 19:59 | <Philip`> | (and also change it to a format that can be read more easily than null-terminated fields) |
| 20:00 | <Philip`> | and then you can parallelise by having each process process one of the chunks |
| 20:00 | <Philip`> | (so make them pretty small chunks, like a hundred megabytes or whatever) |
| 20:00 | <jgraham> | Philip`: All of that sounds like a sensible optimisation but essentially not that different to the (broken) code I already have |
| 20:01 | <Philip`> | Much more sensible than storing the uncompressed data on disk :-) |
| 20:01 | <jgraham> | Which processed the whole file in ~5 minutes. Although that might change a lot when it actually does the right thing |
| 20:01 | <jgraham> | Philip`: It's not really that big a file compared to the amount of free disk space I have :) |
| 20:02 | <Philip`> | jgraham: The point is that it's big compared to the amount of RAM you have |
| 20:02 | <Philip`> | and so it'll be slower to process |
| 20:03 | <Philip`> | I suppose unless you've got 4GB of RAM (or you restrict yourself to a subset of the data) you'll have to do some disk IO anyway, and if the OS does FIFO buffering then it'll be the worst possible case, which isn't so good |
| 20:04 | gsnedders | always gets FIFO and FILO mixed up |
| 20:04 | <jgraham> | FILO is a type of pastry |
| 20:04 | jgraham | is sure he has said that before |
| 20:05 | <Philip`> | gsnedders: It's easy to remember, because people never use the term FILO :-) |
| 20:05 | <Philip`> | (They're called stacks instead) |
| 20:05 | <gsnedders> | Philip`: Assuming the SQA has logic |
| 20:05 | <gsnedders> | (Unlikely) |
| 20:06 | <rubys> | LIFO is common, however |
| 20:06 | <Philip`> | jgraham: Is that ~5 minutes with some parallelisation? |
| 20:07 | jgraham | can find no evidence in the logs that he has actually made the pastry bun before so he is safe in the knowledge that people will only think him unfunny rather than unfunny and repetitive |
| 20:07 | <jgraham> | Philip`: Yes. 4 processes processing 50Mb chunks |
| 20:07 | <Philip`> | Pastry bun? :-p |
| 20:08 | <jgraham> | Argh |
| 20:09 | <Philip`> | jgraham: Hmm, seems to take me ~5 minutes to grep for a string in the 425K pages with no parallelisation |
| 20:09 | <Philip`> | (and it's all CPU bound, so it should parallelise embarrassingly) |
| 20:10 | <Philip`> | (Actually I suppose it's memory bound, so that's not really true; oh well) |
| 20:27 | <Hixie> | hsivonen: is the spec ambiguous? |
| 20:31 | <Philip`> | hsivonen: Did you see http://krijnhoetmer.nl/irc-logs/whatwg/20090424#l-503 ? |
| 20:31 | Philip` | wonders if he should report it via the bug tracker or something |
| 20:34 | <rubys> | Hixie: what's the status/outlook on RDFa? |
| 20:34 | <Philip`> | jgraham: (For comparison: It takes ~70 seconds to read all the pages and extract the HTTP headers, with 4 processes in parallel; and ~400 seconds to parse all the HTML (via SAX2) and extract all the <meta> elements) |
| 20:35 | <Hixie> | rubys: just had a meeting with the last group of people i know of who had input they wanted to give on the issue, having lunch now, plan to start going down the use cases this afternoon |
| 20:36 | <rubys> | so you will have a proposal by the end of the... week? month? Don't need a firm date, just a guess. |
| 20:37 | <Hixie> | there are a number of use cases each of which will likely get their own proposals |
| 20:37 | <Hixie> | i expect the easiest ones will have proposals by tonight |
| 20:38 | <Hixie> | and the more complex ones likely by the end of the week |
| 20:38 | <rubys> | thanks! |
| 20:39 | <Hixie> | np |
| 20:45 | <annevk42> | "Estimated date for last e-mail: 2009-11-01" fun |
| 20:47 | <Hixie> | it was april 2010 last week :-) |
| 20:47 | <annevk42> | it's making a nose dive atm |
| 20:48 | <jgraham> | Philip`: OK, now I have some idea what to aim for :) |
| 20:49 | <Hixie> | annevk42: it's gonna flatline this week as i go back to microdata stuff :-) |
| 20:50 | <Philip`> | jgraham: (This is on a 2.4GHz quad Core 2) |
| 21:05 | <gsnedders> | Where is the estimate? |
| 21:06 | <Philip`> | http://tomayko.com/ - <meta http-equiv='X-UA-Compatible' content='YOUR MOM'> |
| 21:07 | <annevk42> | hehe |
| 21:13 | <annevk42> | W3C QA blog needs karl back; this latest post doesn't even provide some simple pointers |
| 21:15 | <annevk42> | Hixie, if you could reply to my email about sync fetch that'd great btw |
| 21:16 | <Philip`> | http://philip.html5.org/data/ua-compatible-contents.txt |
| 21:17 | <annevk42> | Hixie, especially if you have any tips on how to share an algorithm between a sync and async fetch at the same time (where the algorithm would only dispatch events in the async case, but would share a bunch of common that do not affect the event loop (apart from completely killing the network request; i.e. the current task)) |
| 21:17 | <Hixie> | annevk42: do you want other events, timeouts, etc, to fire during the sync request? |
| 21:18 | <annevk42> | nope |
| 21:18 | <Hixie> | k |
| 21:18 | <annevk42> | it would be sync very much like alert() is sync I guess |
| 21:18 | <annevk42> | but I still need it to follow redirects, throw on network errors, etc. (and dispatch events if invoked as async) |
| 21:19 | <Hixie> | i don't think i'd recommend reusing the 'fetch' algorithm then |
| 21:19 | <Hixie> | i'd just write a new one |
| 21:19 | <Hixie> | that happens to have bits in common |
| 21:20 | <Hixie> | not really sure what to suggest, tip wise |
| 21:20 | <Hixie> | how can i help you? |
| 21:22 | <annevk42> | from http://dev.w3.org/2006/webapi/XMLHttpRequest/#send step 9 and 10 are basically the part that the UA needs to do while fetching regardless of sync/async |
| 21:22 | <annevk42> | now I was thinking that if fetch qeued some kind of task I could just wait for that task to complete in the sync case |
| 21:22 | <annevk42> | and for async return early |
| 21:23 | <Hixie> | the task will not complete if you're waiting |
| 21:23 | <Hixie> | the tasks only ever run when the event loop runs |
| 21:23 | <Hixie> | this is really just a regular event loop pattern |
| 21:24 | <annevk42> | I see |
| 21:24 | <annevk42> | writing out everything two times seems like the wrong approach though |
| 21:26 | <annevk42> | I suppose writing a sync fetch isn't too hard; hmm |
| 21:28 | <annevk42> | does XHR delay the load event? |
| 21:31 | <annevk42> | so you'd have ... fetch resource; return send(); follow loading algorithm; ... or ... sync fetch resource; follow loading algorithm; return send() |
| 21:33 | <annevk42> | this still requires some kind of event loop for the sync case in the UA of course, but I guess that's always required, because otherwise you'd get stuck |
| 21:33 | <annevk42> | but it doesn't have to be explicit in the spec |
| 21:42 | <Philip`> | http://philip.html5.org/data/ua-compatible-headers.txt |
| 21:43 | <Philip`> | I like the first one |
| 21:44 | <gsnedders> | IE=8;FF=3;OtherUA=4: X-UA-Compatible: IE=8;FF=3;OtherUA=4 |
| 21:44 | <gsnedders> | WTF? |
| 21:45 | <annevk42> | Hixie, did the above make some amount of sense? |
| 21:47 | <Philip`> | gsnedders: *shrug* |
| 21:47 | <Hixie> | annevk42: sorry, was afk. reading. |
| 21:48 | <Hixie> | annevk42: you don't need an event loop in the sync case if you just avoid queueing tasks in that case |
| 21:48 | <Hixie> | annevk42: i would just have on algorithm, and every now and then say "if sync... else if async ..." |
| 21:49 | <annevk42> | Hixie, yeah, that is "follow loading algorithm" :) |
| 21:49 | <annevk42> | I suppose I could do it for fetching as well |
| 21:51 | <annevk42> | defining the Web sometimes feels like building a house of cards |
| 21:53 | Philip` | wobbles the table |
| 22:27 | <Hixie> | oops |
| 22:27 | <Hixie> | i just XSSed myself |
| 22:27 | <Hixie> | <iframe src="data:text/html,<script>document.write(location)</script>"></iframe> doesn't do quite what you might think it does |
| 22:47 | gsnedders | wonders how it does that |
| 23:39 | <annevk5> | Hixie, the attribute having the wrong name in the spec does seem like a problem |
| 23:40 | <Hixie> | yeah your e-mail is in the bucket |
| 23:40 | <Hixie> | i'm assuming there's something more wrong with it |