| 00:08 | <Hixie> | yeah I'm not sure I follow the use case for the scheduled notifications |
| 00:09 | <Hixie> | i'm baffled by why some people seem to want range.insertNode() to not insert the node into the range, but only when the range is collapsed |
| 00:10 | <annevk> | Hixie, calendar apps? |
| 00:11 | <Hixie> | yeah but see what roc said |
| 00:12 | <annevk> | presumably if that happened the app would update the notification? |
| 00:12 | <roc> | either the app is running in a browser context or it isn't |
| 00:12 | <roc> | if it is, it doesn't need the timed events |
| 00:13 | <roc> | if it isn't, there's a problem with staleness |
| 00:16 | Philip` | attempts to write feedback on arc(), but then realises he already did it a month ago |
| 01:31 | <andersca> | Hixie? |
| 01:37 | <Lachy_> | jgraham, yt? |
| 01:45 | <Hixie> | andersca: here |
| 02:11 | <andersca> | Hixie: just sent mail to the whatwg list |
| 02:16 | <Hixie> | cool |
| 02:16 | <Hixie> | what's the priority? |
| 02:18 | <Hixie> | andersca: does the fallback thing help? |
| 02:19 | <andersca> | Hixie: I don't think so, since it requires the namespace to have the same domain as the manifest |
| 02:20 | <Hixie> | ah indeed |
| 02:20 | <andersca> | Hixie: also, it would be nice to be able to handle the mt[0-9].google.com case too |
| 02:20 | <andersca> | I'd say the priority is "medium" |
| 02:20 | <Hixie> | my guess would be that i would just make the onlien whitelist a list of uri prefixes |
| 02:21 | <Hixie> | and you'd list the 10 google.com hosts in the file |
| 02:21 | <andersca> | yeah, that sounds good |
| 02:21 | <Hixie> | feel free to experiment with doing that and mail your experiences to the list :-) |
| 02:21 | <andersca> | will do, thanks! |
| 07:22 | <BenMillard> | today I decided to skim some recent IRC logs and public-HTML messages |
| 07:22 | <BenMillard> | I think ignoring almost all public-HTML e-mail was the right choice for me, albeit a tough one |
| 07:23 | <BenMillard> | I also think ignoring nearly everything from this channel is the right choice for me |
| 07:23 | <BenMillard> | not because either are of poor quality or whatever |
| 07:24 | <BenMillard> | it's simply that the time I'd spend tracking it all wouldn't make me more useful or productive |
| 07:25 | <BenMillard> | in more positive news, Mozilla are now funding me to do more work along the lines of Collections of Interesting Data Tables: http://blog.hecker.org/2008/05/28/mozilla-foundation-activities-week-ending-20080523/ |
| 07:31 | <othermaciej> | BenMillard: that's cool |
| 07:36 | <BenMillard> | othermaciej, thanks :) |
| 07:36 | <BenMillard> | this was the proposal I gave them: http://projectcerbera.com/web/study/html5-research |
| 08:16 | <hsivonen> | Dashiva: paragraphs in the Bible can span over verse boundaries but if you were using normal CSSish counters, it would be natural in the markup sense to make verses siblings of each other without interrupting paragraphs |
| 08:18 | <BenMillard> | Hixie, ^ about Mozilla funding. Current thoughts are: http://projectcerbera.com/blog/2008/05#day29 |
| 08:19 | <Hixie> | hello |
| 08:19 | <Hixie> | yeah i saw the announcement -- someone mentioned it in #htmlwg. congratulations! |
| 08:19 | Hixie | looks at your blog |
| 08:20 | <Hixie> | i'd love to see data on the img alt stuff, jgraham would be a great person to coordinate with on that, he has some good ideas of what to look for |
| 08:20 | <Hixie> | (though it's going to be tough to draw conclusions about what is most likely to get authors to do the most accessible thing from _any_ data, sadly) |
| 08:22 | <BenMillard> | hixie, cheers. I'll probably start by adding the 30 or so tables I didn't get add to the first study |
| 08:22 | <BenMillard> | and then analyse how frequently header+header relationships and <td scope> are used |
| 08:24 | <Hixie> | cool |
| 08:24 | <BenMillard> | jgraham and I previosuly coordinated on data table stuff, so doing that in parallel could work |
| 08:24 | <Hixie> | awesome |
| 08:24 | <Hixie> | one thing i'd caution against, looking at this list, is to not do anything that could be automated somehow -- if there's something that could be automated that you think needs doing, just let me know and i'll get some google resources to do it |
| 08:25 | <Hixie> | probably a better use of all our resources |
| 08:25 | <Hixie> | the most valuable thing you can bring to these studies is the human judgement aspect |
| 08:25 | <BenMillard> | hixie, yeah I can't compete with Google (or Philip` or gnsedders) when it comes to total numbers |
| 08:26 | <Hixie> | also one thing i can really help with is getting you lists of sample urls that have some markup feature |
| 08:27 | <BenMillard> | that would be another string to my bow :) |
| 08:28 | <BenMillard> | although I think it's more interesting to work from the other direction; finding types of content and then check the markup |
| 08:28 | <Hixie> | indeed |
| 08:28 | <BenMillard> | both ways are useful, though |
| 08:28 | <BenMillard> | I intend to spend maybe a week writing down every URL I come across in daily life to help target "normal" websites |
| 08:29 | <Hixie> | i could also get you a list of pages with high page rank or something, if you want |
| 08:29 | <hsivonen> | BenMillard: congratulations for the grant |
| 08:30 | <BenMillard> | hsivonen, thanks! |
| 08:30 | <BenMillard> | Hixie, these are all very cool offers and would be really helful |
| 08:30 | <BenMillard> | but right now, I need to make some websites to cover the money I've lost not making anything for the past 2 months! |
| 08:30 | <Hixie> | :-) |
| 08:30 | <Hixie> | feel free to contact me whenever |
| 08:31 | <Hixie> | give me a few days to get you data, it takes time to go through billions of files |
| 08:32 | <BenMillard> | I'll add a link to these logs to my blog, then when I return from making sites I can refresh my memory with all these options |
| 08:32 | <BenMillard> | it's really quite exciting! |
| 09:01 | <bosky101> | hi. does anyone remember the site/url where someone(i think it was ian hixie) gave information of css classnames heuristics based on a large crawl |
| 09:03 | <hsivonen> | bosky101: http://code.google.com/webstats/ |
| 09:05 | <bosky101> | brilliant! thanks |
| 10:58 | <annevk> | hmm, the weather is terrible |
| 10:58 | <Philip`> | I disagree |
| 10:58 | <Hixie> | yeah, hasn't been above 25C all week |
| 10:58 | <Hixie> | i had to wear a coat today! |
| 10:59 | <annevk> | until now it was good, now it rains badly |
| 11:06 | <Hixie> | It feels so weird, writing all these "should"s |
| 11:06 | <roc> | It's freezing here |
| 11:06 | <roc> | down to 9C |
| 11:06 | <Hixie> | i guess it's coming on to winter in your neck of the woods |
| 11:07 | <roc> | definitely |
| 11:08 | <annevk> | freezing at 9C, hah |
| 11:08 | <Philip`> | Today is quite sunny - it should reach 20C if we're lucky |
| 11:08 | <annevk> | a few years back it reached -10 / -15 in Oslo at some points |
| 11:09 | <Hixie> | it reached -20C while i was there, for a few hours at night, irrc |
| 11:09 | <Hixie> | iirc |
| 11:09 | <roc> | I was joking. When I lived in Pittsburgh it reached -20C a few times |
| 11:09 | <Hixie> | -20C is stupid cold |
| 11:09 | <annevk> | brr |
| 11:09 | Hixie | hugs the bay area |
| 11:09 | <Hixie> | we have two seasons here |
| 11:09 | <Hixie> | spring and summer |
| 11:09 | <hsivonen> | Is it warm on the inland side of the Bay? |
| 11:09 | <hsivonen> | San Francisco is *cold* |
| 11:09 | <roc> | yes |
| 11:09 | <roc> | yes it is |
| 11:10 | <Hixie> | SF doesn't count as the bay area, insofar as weather goes |
| 11:11 | <roc> | I'm actually amazed it doesn't get colder in Oslo than Pittsburgh |
| 11:11 | <roc> | I thought those Norse hackers were good because they couldn't go outside for six months |
| 11:12 | <annevk> | gulfstream |
| 11:12 | <hsivonen> | If I create a fresh document instance in JS and add <script> elements, will they run? if so, in what context? |
| 11:12 | <annevk> | well, it does get pretty dark in the winter most of the day :) |
| 11:12 | <hsivonen> | If I importNode script elements to the current HTML document, will they run? |
| 11:12 | <Hixie> | hsivonen: yes, and in that document's |
| 11:12 | <Hixie> | hsivonen: importNode will work if the <Script>s haven't already run |
| 11:12 | <hsivonen> | Hixie: is there a way to tell a document object to be safe and not run scripts? |
| 11:13 | <Hixie> | not before html5, no |
| 11:13 | <Hixie> | <iframe sandbox> will do that though |
| 11:13 | <Hixie> | in new uas |
| 11:13 | <hsivonen> | Hixie: how does your live dom viewer handle script security from user input? |
| 11:13 | <roc> | the stupid irony of the Bay Area is that they have great weather but most of the population gains nothing due to late sleep patterns and/or insane work habits |
| 11:13 | <Hixie> | hsivonen: that domain has no cookies |
| 11:14 | <Hixie> | roc: hah, so true |
| 11:14 | Hixie | saw dawn yesterday :-/ |
| 11:15 | <annevk> | Hixie's working hours work for me :) |
| 11:15 | <annevk> | gives a reasonable Europe overlap |
| 11:15 | <Hixie> | that's partly why i do it |
| 11:15 | <Hixie> | at least, that's my excuse |
| 11:15 | <Hixie> | i could do it as easily if i got up early :-P |
| 11:17 | <hsivonen> | I find it bad for my health that people start having interesting discussions when it is late in the evening here |
| 11:17 | <hsivonen> | this time works much better for me |
| 11:17 | <Hixie> | heh |
| 11:18 | <Hixie> | hm |
| 11:18 | <Hixie> | the notification API has one problem |
| 11:18 | <Hixie> | if you open multiple instances of a calendar app |
| 11:19 | <Hixie> | and they all fire notifications for each event... |
| 11:19 | <annevk> | why would you do that? |
| 11:20 | <Hixie> | do what? |
| 11:21 | <Philip`> | They could use localStorage to synchronise between all the tabs |
| 11:21 | <Hixie> | seems unlikely that they'd bother |
| 11:21 | <Hixie> | i guess we can just tell uas to coallesce identical messages |
| 11:22 | <annevk> | open multiple tabs of the same app |
| 11:23 | <Philip`> | If I get two identical email messages, close enough together that their notifications will overlap, I'd probably want to receive two notifications, else I might miss the second message |
| 11:23 | <Hixie> | i do that sometiems with calender, either by mistake or to compare things |
| 11:23 | <Hixie> | if you get two identical messages, why would you care about the second one? |
| 11:24 | <roc> | "Missile launched!" |
| 11:24 | <Philip`> | *two different email messages but with identical subject lines so the subset of data used in the notifications is identical |
| 11:24 | <Hixie> | hmm |
| 11:26 | <Philip`> | (Get one notification, switch to tab and read email message, get second notification but the UA coalesces it with the first so I can't tell there's been another message) |
| 11:26 | <hsivonen> | do iframes let the parent document remove the documentElement and create a new one? |
| 11:26 | <Hixie> | hsivonen: in principle, yes, in practice it depends on the UA |
| 11:27 | <Hixie> | Philip`: how about just having it say "x2" "x3" "x4" ? |
| 11:27 | <hsivonen> | Hixie: do you happen to remember if Gecko/WebKit/Opera allow removal and reinsertion of the document element? |
| 11:27 | <Hixie> | no idea |
| 11:27 | <roc> | Hixie: yeah, that's like Unix kernel messages |
| 11:27 | <hsivonen> | ok |
| 11:27 | <hsivonen> | thanks |
| 11:28 | <roc> | Error: not a typewriter (4 times) |
| 11:28 | <hsivonen> | I'm trying to figure out if I should bother to make an HTML5 live DOM viewer using GWT and the Validator.nu parser |
| 11:28 | <roc> | sick |
| 11:29 | <hsivonen> | roc: why sick? isn't it cool compiler tech? |
| 11:29 | <Hixie> | sure, go for it :-) |
| 11:29 | <hsivonen> | I find that my code is one regexp removal away from GWT compat |
| 11:29 | <roc> | those things are not mutually exclusive |
| 11:29 | <Philip`> | If the coalescence only happened to notifications from different tabs, I suppose that'd avoid the missed-email problem except in quite rare cases |
| 11:30 | <Philip`> | so it probably wouldn't be a real problem then |
| 11:31 | <roc> | the 4x/4 times feature is nice for abusive code and for buggy code too |
| 11:31 | Philip` | hopes the notification thing would include UI for disabling notifications from abusive sites |
| 11:32 | <Philip`> | (as well as the UI for enabling it first) |
| 11:32 | <roc> | nah |
| 11:32 | <roc> | well |
| 11:33 | <roc> | in Firefox the obvious thing would be to start by showing it in an infobar top-of-tab |
| 11:33 | <annevk> | i believe the design is that to make it system wide you first need to grant the site access somehow |
| 11:33 | <roc> | and give you options to make it go away or promote to desktop level |
| 11:33 | <annevk> | what roc said |
| 11:34 | <Philip`> | I'm thinking about what happens when unknowingly grant permission and then the site abuses the feature and you want to stop it |
| 11:34 | <Hixie> | actually i have this whole plan for how to deal with abuse |
| 11:34 | <Hixie> | the spec describes it, littered with "should"s |
| 11:34 | <roc> | one of the most interesting things we're doing on the Web is figuring out ways to let apps do everything they want without trust and without driving users mad with approve/deny decisions |
| 11:35 | <Hixie> | indeed |
| 11:36 | <Philip`> | You just a generic Cancel/Allow dialog box with a "[ ] remember this decision" checkbox |
| 11:36 | <Philip`> | s//need/ |
| 11:36 | <roc> | no |
| 11:36 | <roc> | no no no |
| 11:36 | <Philip`> | Alas :-( |
| 11:38 | Hixie | puts Philip` in the brig for 24 hours for even joking about suggesting that |
| 11:43 | <annevk> | so is Android just software or also hardware? |
| 11:43 | <zcorpan> | Hixie: since "textarea" isn't in the list of tags that escape foreign lands, whether to do case fixup or not has nothing to do with clashing with html since it'll be in the svg namespace anyway (re http://www.w3.org/mid/Pine.LNX.4.62.0805230003340.12911⊙hdc ) |
| 11:44 | <Hixie> | annevk: android is a software platform |
| 11:44 | <annevk> | so where is the hardware from? |
| 11:45 | <Hixie> | (in fact, it's an open source linux-based OS distribution, though the source hasn't been released yet aiui) |
| 11:45 | <Hixie> | the hardware comes from any hardware manufacturer, most likely those in the Open Handset Alliance: http://www.openhandsetalliance.com/oha_members.html |
| 11:46 | <Hixie> | look under "Handset Manufacturers" |
| 11:46 | <Hixie> | though typically handsets are sold through network operators ("Mobile Operators" on that list) |
| 11:47 | <Hixie> | (i mean, typically for any platform, not just android) |
| 11:48 | <annevk> | I meant the hardware used in the demos, but I guess your answer works :) |
| 11:49 | <Hixie> | not sure about particular demos, they probably mentioned it in the relevant talks though |
| 11:51 | <hsivonen> | so Microsoft wasn't able to break the coupling of Java-the-language and Java-the-class-library, but now Google is doing it with GWT and Android |
| 12:00 | <roc> | IBM also did it |
| 12:00 | <hsivonen> | true |
| 12:00 | <hsivonen> | actually, the only desktop Java app I run regularly is Eclipse |
| 12:00 | <hsivonen> | Swing lost |
| 12:02 | <hsivonen> | with Eclipse, though, the full J2SE class library is still there |
| 12:02 | <hsivonen> | not so with GWT which doesn't even have proper java.io |
| 12:02 | <roc> | I was thinking of other things |
| 12:02 | <hsivonen> | oh |
| 12:13 | annevk | looks at the notifications API |
| 12:13 | <Hixie> | don't yet |
| 12:14 | <Hixie> | it'll be done momentarily |
| 12:15 | <annevk> | in the IDL it's listed under "// modal user prompts" |
| 12:15 | <annevk> | i guess modal should be removed there |
| 12:16 | <Hixie> | valid |
| 12:17 | <Hixie> | ok the spec is regenned |
| 12:18 | <annevk> | ah yeah, it differs a lot :) |
| 12:19 | <annevk> | it says "of abused" instead of "or abused" |
| 12:19 | <Philip`> | Hixie: http://www.whatwg.org/specs/web-apps/current-work/multipage/ is broken |
| 12:20 | <Hixie> | it happens |
| 12:20 | <annevk> | also, should it be phrased in a way that does not require it to be visual? "show" makes me think it's visual only |
| 12:21 | <Hixie> | it'll fix itself the next time it regens |
| 12:21 | <Hixie> | happens every 24h |
| 12:21 | <annevk> | is "sandboxed notifications browsing context flag" a a new <iframe sandbox> feature? |
| 12:22 | <Hixie> | reload |
| 12:26 | <annevk> | still has the typo |
| 12:27 | <annevk> | (second paragraph, s/of/or/) |
| 12:30 | <Hixie> | oops, missed your comment above |
| 12:30 | <Hixie> | was meant to be "if" not "or" :-) |
| 12:31 | <annevk> | they can be annoying without being abused too! |
| 12:31 | annevk | doesn't care about working of such things though |
| 12:34 | <Hixie> | hehe |
| 12:34 | <annevk> | wording, I mean, meh |
| 12:37 | <Philip`> | "By default no origin should be flagged as such" - that's abuse of RFC2119 language |
| 12:39 | <virtuelv> | Hixie: isn't showNotification a bit overengineered? |
| 12:39 | <Hixie> | how so? |
| 12:39 | <virtuelv> | I'd say that content and callback is sufficient |
| 12:39 | <virtuelv> | leave the title to the UA |
| 12:39 | <Hixie> | (it's a lot simpler than, say, the version gears has) |
| 12:39 | <virtuelv> | and drop the subtitle |
| 12:40 | <virtuelv> | void showNotification(in DOMString msg, in Function callback); is what Opera has in the widget spec |
| 12:40 | <Hixie> | for gmail we want notifications like: |
| 12:40 | <Hixie> | | Canvas line style comments |
| 12:40 | <Hixie> | | Philip Taylor |
| 12:40 | <virtuelv> | Hixie: try to remain compatible with Growl? |
| 12:40 | <Hixie> | | The spec says: "The lineCap attribute |
| 12:40 | <Philip`> | "user agents may allow users to whitelist flags or groups of flags as being trusted notification sources" - should that be s/flags/origins/g? |
| 12:40 | <Hixie> | | defines the type of endings ... |
| 12:41 | <Hixie> | Philip`: er yes. |
| 12:41 | <virtuelv> | Hixie: while you're at it: window.getAttention() |
| 12:42 | <Hixie> | what would that do? |
| 12:42 | <virtuelv> | typically, try to grab attention to the window, without grabbing focus |
| 12:42 | <virtuelv> | example: change tab color, flash task bar, etc |
| 12:42 | <virtuelv> | no arguments, no return value |
| 12:42 | <Hixie> | that seems far too abuseable |
| 12:43 | <virtuelv> | abuseable, how? |
| 12:44 | <Philip`> | Hixie: "all subdomains and ports of *.example.org" - that should be "all subdomains and ports of example.org", or "all ports of any domain matching *.example.org" or something |
| 12:44 | <Hixie> | i don't want every porn site i go to to be flashing its task bar button at me |
| 12:44 | <Hixie> | Philip`: yeah... hmm |
| 12:44 | <Philip`> | Hixie: "Then, mail.example.com and calendar.example.com would both be able to show notifications" - s/.com/.org/ |
| 12:44 | <Hixie> | oops |
| 12:45 | <annevk> | apparently IE has it, from https://bugzilla.mozilla.org/show_bug.cgi?id=293077 |
| 12:46 | <Hixie> | IE has (had?) window.focus() |
| 12:46 | <Hixie> | which i even worse |
| 12:46 | <annevk> | oops, yeah |
| 12:46 | <annevk> | it does |
| 12:46 | <virtuelv> | Hixie: but you want them to display "Free porn" popups? |
| 12:46 | <annevk> | I think Opera has that too and Gmail abuses it which is annoying |
| 12:47 | <Hixie> | virtuelv: as defined, they can't, since hte popups are limited to the rendering area of the tab. |
| 12:47 | <Hixie> | virtuelv: (unless i click the button to trust the site) |
| 12:48 | <Hixie> | also, the real question is when would you want to grab attention without having something to grab attention about? if you have somethign to tell the user, use a notification. |
| 12:48 | <Hixie> | it'll do the getAttention() thing if the user trusts your site |
| 12:48 | <virtuelv> | Hixie: for instance a web-based IRC client that notifies you whenever someone mentions your name |
| 12:48 | <virtuelv> | like XChat does, for instance |
| 12:49 | <Philip`> | Hixie: "a manner consisted with the platform conventions" - s/consisted/consistent/ |
| 12:49 | <Hixie> | why wouldn't you want it to show a notification with the line that was said |
| 12:49 | <Hixie> | oops, fixed, thanks |
| 12:50 | <Philip`> | Hixie: Should the first "should" in "Otherwise, the notification should be rendered in a manner consisted with the platform conventions for system-wide notifications, but it should be rendered within ..." be "should not"? |
| 12:50 | <Hixie> | no? |
| 12:50 | <Hixie> | why would we ever recommend against using platform conventions? |
| 12:51 | <Philip`> | It seems weird to say it should be consistent with platform conventions (which are always outside the browser window) and also that it should be inside the window |
| 12:51 | <Hixie> | i guess i could remove the first bit |
| 12:52 | <Philip`> | The latter "should" should be more important than the first one, so I'm not sure the first one is needed at all |
| 12:52 | <Hixie> | k |
| 12:52 | <virtuelv> | Hixie: because that line is far too intrusive |
| 12:52 | <virtuelv> | I just need to know that something has happened |
| 12:53 | <Hixie> | i'd want to know what happened |
| 12:53 | <virtuelv> | for instance, when someone mentions my name in a different tab in xchat, someone has mentioned my name |
| 12:53 | <Hixie> | i hate that i have to go look at my irc client to establish that they just said "nn" |
| 12:53 | <annevk> | what happens if only whitespace was passed? |
| 12:53 | <Hixie> | anyway |
| 12:53 | <Hixie> | that's moot |
| 12:53 | <Hixie> | because like i said |
| 12:53 | <Hixie> | getAttention() is far too abusable |
| 12:54 | <Hixie> | annevk: you get a blank notification with just the favicon and title of the page, i guess |
| 12:54 | <Philip`> | Hixie: "if a site contains a gadget of a mail application in a sandboxed iframe and that frame triggers a notification upon the receipt of a new e-mail message, that notification would be displayed on top of the gadget only." - the last bit seems wrong, since the notification would be displayed on the top-level browsing context instead of just the gadget |
| 12:55 | <Philip`> | Hixie: s/burried/buried/ |
| 12:56 | <virtuelv> | what is the point of a notification if it can't escape the viewport for the page it originated from, btw? |
| 12:56 | <Hixie> | no, if it's sandboxed it'll be displayed on the iframe itself |
| 12:56 | <Hixie> | (fixed buried) |
| 12:57 | <Hixie> | virtuelv: the point is to show the user that the site supports notifications so that it can opt into showing them system-wide |
| 12:57 | <virtuelv> | I realise that you can trust the notification, but I'd need to know that something has happened in the tab to switch to it, I don't actually need the notification |
| 12:57 | <Hixie> | s/it/he/ |
| 12:57 | <Hixie> | see the spec, even normal (not trusted) notifications make the tab caption get highlighted in some way |
| 12:57 | <Philip`> | Hixie: Is the sandboxed iframe a top-level browsing context? |
| 12:58 | <virtuelv> | Hixie: which is what getAttention() would do as well |
| 12:58 | <Hixie> | Philip`: no, how could it be? |
| 12:58 | <Hixie> | virtuelv: but there's no way to opt in with getAttention() |
| 12:58 | <Hixie> | virtuelv: and it's just a subset of showNotification() |
| 12:58 | <Philip`> | Hixie: It could have been defined that way without me noticing :-) |
| 12:59 | <virtuelv> | Hixie: more precisely, it's the empty subset of showNotification |
| 12:59 | <Hixie> | Philip`: ah :-) |
| 12:59 | <Hixie> | virtuelv: eh? |
| 12:59 | <Philip`> | Hixie: In that case, the spec just says "it should be rendered within the top-level browsing context of the browsing context associated with the script execution context of the script that invoked the method" and says nothing about it being displayed just on the sandboxed iframe |
| 13:00 | <Hixie> | look at the paragraph immediately above that one |
| 13:00 | <virtuelv> | empty: no title, subtitle, text or callback |
| 13:00 | <virtuelv> | or think of it as a subset of focus() that doesn't try to grab the window focus |
| 13:00 | <Philip`> | Hixie: Oh, right, I missed/forgot that bit |
| 13:01 | <virtuelv> | I believe most WM's have a method called getAttention these days |
| 13:01 | <Hixie> | virtuelv: that would show a notification with the page title and favicon (and maybe domain name, and maybe time, and some buttons). |
| 13:01 | <Hixie> | virtuelv: most WMs assume that the user is running trusted code. The Web is different. |
| 13:01 | <Hixie> | anyway i should go to bed |
| 13:01 | <Hixie> | thanks for the review Philip` |
| 13:02 | <virtuelv> | Hixie: and that is more information than I would need to show |
| 13:02 | <Hixie> | virtuelv: i disagree :-) |
| 13:03 | <virtuelv> | Hixie: you have reinvented alert() asynchronously in this case, albeit a tab-modal one |
| 13:03 | <annevk> | it's not modal |
| 13:04 | <Philip`> | Hixie: Did you see my first comment? |
| 13:04 | <Philip`> | ('"By default no origin should be flagged as such" - that's abuse of RFC2119 language') |
| 13:04 | <Hixie> | Philip`: nope, missed that one too. will fix. |
| 13:05 | <Hixie> | virtuelv: yup, pretty much. except it's not modal, and is transient, and isn't locked to the page. |
| 13:05 | <Philip`> | Hixie: "the user agent could just shown one" - s/shown/show/ |
| 13:06 | <Hixie> | thanks, will fix that too |
| 13:06 | <Philip`> | Hixie: Why remove duplicate notifications that are from the same browsing context? That case seems most likely to be intentional (e.g. two identical emails) since there's only one instance of the application in that browsing context |
| 13:07 | <ralphm> | Hi! I'm toying with adding link elements pointing to XMPP end points, but I am not sure what rel values to use. |
| 13:07 | <Hixie> | Philip`: because having two notifications that are the same is dumb and ugly :-) |
| 13:08 | <ralphm> | Example: I have a page that is a representation of a resource, and want to offer an XMPP Publish-Subscribe endpoint that could be subscribed to, to receive updates when the resource changes. |
| 13:09 | <Hixie> | annevk: btw that bug is what caused me to start speccing this in the first place (it's the top e-mail in the dom-focus folder) |
| 13:09 | <ralphm> | Any ideas? |
| 13:09 | <Hixie> | i'm afraid i don't really know what an xmpp end point is |
| 13:09 | <Hixie> | but what about rel="xmpp-end-point"? |
| 13:10 | <ralphm> | Hixie: you do know about XMPP itself? I was actually thinking of rel="alternate feed" |
| 13:11 | <Hixie> | don't use two keywords |
| 13:11 | <Hixie> | it'll create two links |
| 13:12 | <ralphm> | yeah, intentionally. I read the description of the different link rels here: http://www.whatwg.org/specs/web-apps/current-work/multipage/section-links.html#feed |
| 13:12 | <Hixie> | oh, i see, you want to use the existing "alternate feed" definition |
| 13:12 | <ralphm> | yes |
| 13:12 | <Hixie> | so it's just an RSS page? |
| 13:12 | <Hixie> | except not RSS? |
| 13:12 | <ralphm> | it is not a page |
| 13:13 | <Hixie> | then don't use rel="alternate feed" :-) |
| 13:13 | <ralphm> | it would point to e.g. xmpp:pubsub.example.org?;node=test |
| 13:13 | <ralphm> | why not? |
| 13:13 | <Hixie> | (sounds complicated) |
| 13:13 | <Hixie> | rel="alternate feed" points to a syndication feed document |
| 13:14 | <Hixie> | which happens to contain the same content as the current document |
| 13:14 | MikeSmith | says, cool to see ralphm on #whatwg |
| 13:14 | <ralphm> | well, that XMPP resource provides that, except not as a document as such |
| 13:14 | <annevk> | ralphm, I wouldn't use feed, feed is for RSS/Atom |
| 13:15 | <annevk> | ralphm, just mint xmpp-end-point or something |
| 13:15 | <ralphm> | you would get Atom entry documents sent over XMPP asynchronously whenever the resource changes |
| 13:15 | <Hixie> | lordy |
| 13:15 | mpt | tries to imagine what the "I trust this Web site to show me notifications outside the browser window" would look like |
| 13:16 | <Hixie> | mpt: a little icon that looks a bit like the windows "restore window" button icon, at the top left of the toast notification thingy |
| 13:17 | <mpt> | Or maybe like the "Undock" button for palettes in various art programs |
| 13:17 | <ralphm> | I would say the end result for the user would be similar, it is a feed, a client (given XMPP powers) would be able to subscribe to the feed and give the user updates, except without polling. |
| 13:18 | <ralphm> | so that's why I thought 'alternate feed' would be appropriate, although I understand it was not designed with XMPP in mind, per se. |
| 13:18 | <Hixie> | mpt: right |
| 13:19 | <Hixie> | ralphm: what's the use case for this? just changing to a push model instead of pull? |
| 13:20 | <ralphm> | Hixie: for this particular example, yes |
| 13:21 | <Hixie> | wouldn't that mean you couldn't have more than 65000 subscribers? |
| 13:21 | <ralphm> | also, it could be a feed for just one particular thing, like one blog item or one person's profile page |
| 13:21 | <ralphm> | Hixie: huh what? |
| 13:21 | <ralphm> | why? |
| 13:21 | <Hixie> | ralphm: isn't there a limit to how many clients can connect to an xmpp server at once? |
| 13:22 | <Hixie> | i guess they don't each need their own outgoing port |
| 13:22 | <ralphm> | Hixie: XMPP has a distributed server model, it scales pretty well |
| 13:23 | <ralphm> | also, the attempt to subscribe could point to a repeater of the same notifications, so that eventually scales infinitely |
| 13:23 | <ralphm> | and way better than Atom/RSS polling |
| 13:25 | <Hixie> | i guess it's just hte idea of sending html embedded in xml embedded in xml that has me worried |
| 13:25 | <Hixie> | anyway |
| 13:25 | <Hixie> | the main problem with using "feed" is that current era blog readers wouldn't be able to handle the xmpp: protocol |
| 13:25 | <Hixie> | and so you'd still want them to get the http: one |
| 13:26 | <ralphm> | yes, I would provide one of those too |
| 13:26 | <Hixie> | but the UA would in practie not know which to send to the blog reader |
| 13:26 | <Hixie> | so i'd recommend just minting a new pair of keywords like "xmpp-feed" and "main-xmpp-feed" |
| 13:26 | <ralphm> | also, we tend to design for the future, I hope |
| 13:27 | <Hixie> | sadly designing for the future without taking the present into account tends to lead to systems that never get adopted |
| 13:27 | <ralphm> | yeah, like annevk suggested. Sure I can do that, but I am wondering if we should define a new rel if the semantics are mostly identical, except for mapping to a different protocol scheme |
| 13:28 | <hsivonen> | ralphm: are you the person who discussed XMPP integration with annevk and me at XTech after lightning talks? |
| 13:28 | <ralphm> | hsivonen: yeah |
| 13:28 | <Hixie> | ralphm: well, it's not really the semantics that matter, it's how they are handled in the wild |
| 13:28 | <hsivonen> | ralphm: ok. (good to connect the threads) |
| 13:30 | <ralphm> | Hixie: a quick scan of web browser reveals that ie, safari seem to ignore the link (seems ok to me) and ff does show it, but nothing happens when you act on it (e.g. clicking on the feed icon) |
| 13:30 | <Hixie> | ralphm: if you do want to use feed/alternate feed, i recommend finding a dozen or so different blog readers, and testing them with a variety of different ways of linking the feeds, as picked from a big sample of blogs (e.g. the last 50 blogs mentioned on digg and reddit, and feeds from sites like facebook and twitter) and seeing what they do when you add xmpp: to the mix |
| 13:30 | <Hixie> | ralphm: well you need to test the browsers in conjunction with various blog reader software and blog reader sites |
| 13:31 | <Hixie> | ralphm: and you need to test not just how xmpp: works, but how the fallback to http: works when xmpp: is specified but not supported |
| 13:31 | <Hixie> | ralphm: since if that doesn't work very reliably, people just won't use it |
| 13:32 | <Hixie> | anyway, i have to go to bed now |
| 13:32 | <Hixie> | it being 25 to 6 :-) |
| 13:32 | <ralphm> | Hixie: yeah, that seems fair. Night and thanks |
| 13:32 | <Hixie> | np |
| 13:32 | <Hixie> | nn |
| 13:40 | <ralphm> | If anyone else has ideas or suggestions, that'd be most welcome. We want to draft a specification for auto discovery of XMPP entities and publish-subscribe topics within the XMPP Standards Foundation. |
| 14:29 | <gavin_> | annevk: is there a way for me to followup on a bug I filed on Opera a long time ago? |
| 14:30 | <annevk> | nr? |
| 14:30 | <gavin_> | I don't remember |
| 14:30 | <gavin_> | and I can't find any email |
| 14:30 | <annevk> | e-mail? |
| 14:30 | <annevk> | address |
| 14:30 | <gavin_> | not sure of that either... gavin.sharp⊙gc, or maybe gavin/gsharp⊙mc |
| 14:32 | <annevk> | bit busy, will look |
| 14:32 | <MikeSmith> | Hixie: why was the name "Doctype" chosen for Google Doctype? |
| 14:32 | <gavin_> | thanks |
| 14:34 | <hsivonen> | now instead of having to compete against w3cschools and Zeldman, I have to compete against Google itself for the search term "doctype" |
| 14:35 | <hsivonen> | in fact, Google Doctype pushed me off the first result page on google.fi |
| 14:35 | <annevk> | gavin_, ok, found something, it's marked fixed |
| 14:36 | <hsivonen> | still there on google.com |
| 14:36 | <annevk> | gavin_, it's about setting Opera as the default browser |
| 14:37 | <gavin_> | yeah, that's the one |
| 14:37 | <gavin_> | cool |
| 14:37 | <gavin_> | is it in shipped builds? |
| 14:38 | <gavin_> | I suppose I should test |
| 14:38 | <annevk> | gavin_, should be in 9.5 builds i guess |
| 14:38 | <gavin_> | alright, thanks |
| 14:38 | <annevk> | gavin_, according to my data it was fixed over a year ago :) |
| 14:39 | <gavin_> | yeah, I reportred it a awhile ago and kinda forgot |
| 14:39 | <gavin_> | it was causing trouble with our profile importer |
| 14:39 | <annevk> | hereby thanks for the report :) |
| 14:42 | <hsivonen> | Hixie: is your Live DOM Viewer under an Open Source license? |
| 15:05 | <Philip`> | hsivonen: Hixie hasn't sued me for copying and modifying it |
| 15:17 | <hsivonen> | Philip`: not good enough :-( |
| 15:53 | <Lachy> | Our presentation went well this morning |
| 15:55 | <annevk> | thanks for telling us that! |
| 15:55 | <annevk> | :p |
| 15:55 | <hsivonen> | do Gecko/WebKit/Opera support DOM Level 3 setUserData? |
| 15:56 | gsnedders | hopes he didn't look too stupid |
| 15:59 | <Philip`> | gsnedders: That's why you should avoid ever having photos published :-p |
| 15:59 | <gsnedders> | Philip`: I should take a photo of you when you don't realise next time I see you :) |
| 16:11 | <annevk> | Philip`, did you add a regression testcase for the html5lib thing you fixed? |
| 16:13 | <Philip`> | annevk: Yes (test_stream.py / test_newlines2) |
| 16:14 | <Philip`> | (The Ruby implementation doesn't seem to be affected, as far as I can tell) |
| 16:21 | <gsnedders> | PLY doesn't seem able to easily cope with http-parsing. :( |
| 16:23 | <gsnedders> | comment = "(" *( ctext / quoted-pair / comment ) ")" would be really horrible |
| 16:24 | <gsnedders> | It has no sort of repetition outwith inside tokens |
| 17:18 | <Philip`> | "If the WG wants to recommend UI for browsers, it should do so in a seperate document, so as to not confuse what is required for interoperability." - how does that relate to the recent notification API, whose definition is pretty much all about UI? |
| 18:19 | <gsnedders> | Oh dear. |
| 18:19 | gsnedders | concludes there isn't any lexer in any language he knows that can be used for HTTP while keeping quite close to ABNF |
| 18:22 | <Philip`> | Is there not an ABNF lexer implementation? |
| 18:22 | <gsnedders> | No |
| 18:22 | <gsnedders> | But we've been over this before :) |
| 18:22 | <Philip`> | Have we? |
| 18:23 | Philip` | 's memory is failing |
| 18:23 | <gsnedders> | Yeah, when I was on the train going outh. |
| 18:23 | <gsnedders> | *south |
| 18:23 | <gsnedders> | There's http://simpleparse.sourceforge.net/ which does EBNF |
| 18:23 | <Philip`> | Does http://www.coasttocoastresearch.com/home.php not exist or work or something? |
| 18:24 | <gsnedders> | There was some issue with it |
| 18:24 | Philip` | goes away |
| 18:25 | <gsnedders> | Ah yeah. |
| 18:25 | <gsnedders> | I couldn't work out how the hell you used it |
| 18:27 | gsnedders | works it out! |
| 19:17 | <hsivonen> | takkaria: http://html4all.org/mailman/archives/list_html4all.org/2008-May/000892.html |
| 19:18 | <Dashiva> | I wonder what he imagines those marks would be used for if there was no implementation |
| 19:38 | <Philip`> | Dashiva: They are used for semantics, I guess |
| 19:41 | <gsnedders> | Hixie: Could you possibly look at <http://simplepie.org>; from a Google IP and say if it looks as if it's been hacked — apparently it does from Y! IPs (done by IPs, and not UA, seemingly) |
| 19:45 | <Philip`> | gsnedders: http://66.102.9.104/search?q=cache:tAPdA3FmK6cJ:simplepie.org/ |
| 19:46 | <gsnedders> | Philip`: Sure, but that's six days old. I have no idea how old this seeming hack is, and I can't find any trace of it in the sp.o code |
| 19:47 | <Philip`> | "After the neotropical steroid, designer of the clearer prilosec it intoxicated interstate bubbly and recapitulative; satirically mottled." - surely you can't fix the hack and destroy such poetry |
| 19:47 | <Philip`> | Ah |
| 19:49 | <gsnedders> | http://66.102.9.104/search?q=cache:tAPdA3FmK6cJ:simplepie.org/&hl=en&strip=1 — that shows it all :\ |
| 19:51 | gsnedders | can't find the cause at all :\ |
| 19:54 | <Philip`> | Checked the web server configuration too? |
| 19:55 | <Philip`> | It looks like most of the links in the page are to other similarly compromised sites |
| 19:57 | <gsnedders> | Philip`: I don't have access to that |
| 20:01 | <takkaria> | yay, I got on html4all |
| 20:01 | <takkaria> | I must be part of the cabal now |
| 20:05 | gsnedders | wonders if he is |
| 20:06 | <takkaria> | RB even managed to link to the wrong post of mine |
| 20:06 | <gsnedders> | http://html4all.org/pipermail/list_html4all.org/2007-September/000424.html |
| 20:06 | <smedero> | gsnedders: For what it is worth, I just had someone on the Y! internal network look at simplepie.org and they don't see the spammy version. I can email you the frontpage HTML source if you like... but I don't see any of the residuals of one of the wordpress spam hacks. |
| 20:07 | <takkaria> | it makes a good point |
| 20:07 | <Philip`> | shepazu: http://google.com/search?q=cache:Yof6TxEwk3YJ:schepers.cc/ - your site seems a bit spammy |
| 20:07 | <gsnedders> | smedero: I just emailed the guy I've been dealing with, having removed what I thought might be the cause |
| 20:08 | <Philip`> | shepazu: Same problem on http://www.svg-whiz.com/ which looks like yours |
| 20:08 | <shepazu> | Philip`: spammy? |
| 20:10 | <Philip`> | shepazu: "Couldn't you behoove on polygon duet, so you could squib outshoots? Her senegas stand-up as scrape as surpassing eugenias. "Transitionally, I'm bussing out", equidistant diethylpropion hcl 75mg tuberous sclerosis 2." and so on |
| 20:11 | <Philip`> | (plus links to other sites that have the same issue when viewed via the Google cache) |
| 20:11 | <shepazu> | Philip`: I'm not seeing that... |
| 20:12 | <Philip`> | shepazu: Using the Google cache link? |
| 20:12 | <gsnedders> | smedero: Could you get them to look once more? |
| 20:13 | <smedero> | yeah I just nudged him... though now his status says "lunch!". :/ bums. |
| 20:14 | <shepazu> | Philip`: suggestions? |
| 20:19 | smedero | heads to lunch as well |
| 20:21 | <zcorpan_> | krijnh: your linkification regexp fails on your own url at http://krijnhoetmer.nl/irc-logs/wai-aria/20080528 |
| 20:22 | <Philip`> | shepazu: No idea - ask gsnedders when he's worked out why he has the same problem :-) |
| 20:22 | <shepazu> | thanks for the pointer |
| 20:22 | Philip` | writes a script to see how big this network is |
| 20:23 | <Philip`> | Oh, that doesn't work since Google blocks my cache-spider |
| 20:24 | <annevk> | is my site affected in some way? my page rank dropped from 7 to 5 for no apparent reason |
| 20:24 | <Philip`> | shepazu: http://www.vectoreal.com/ was in the list before Google started hating me, though |
| 20:24 | <annevk> | "To highlight the insanity of the WhatWG" classic |
| 20:25 | shepazu | wishes he had more time to figure this crud out |
| 20:27 | <Philip`> | shepazu: It looks like a pretty widespread problem :-( |
| 20:28 | <shepazu> | Philip`: to be honest, I don't understand what the issue is... |
| 20:28 | <shepazu> | why is archive showing spam when the site itself isn't? |
| 20:29 | <shepazu> | for the record, though, I am totally against squibbing offshoots |
| 20:29 | <Philip`> | shepazu: It probably detects Googlebot (via IP or something) and modifies the response content |
| 20:29 | <annevk> | clever |
| 20:30 | <shepazu> | is this something I can or need to do something about? |
| 20:30 | <Philip`> | I think it replaces everything between the first <p> and last </p> |
| 20:30 | <gsnedders> | Philip`: In the sp.o case it doesn't |
| 20:30 | <annevk> | shepazu, you should look in your templates whether there's some hack in there and probably change passwords and all |
| 20:31 | <shepazu> | ah, ok, thanks anne |
| 20:31 | <Philip`> | shepazu: It seems to be that your site/server is hacked in some way, which is not good, but I haven't got a clue how to fix it :-/ |
| 20:31 | <hsivonen> | are all those sites running a vulnerable version of WordPress? |
| 20:31 | <annevk> | if they're very clever they also hacked the new password thingie to e-mail them |
| 20:31 | <annevk> | so a clean install might be safest |
| 20:32 | <Philip`> | gsnedders: Yes it does |
| 20:32 | <gsnedders> | Philip`: OK, maybe it does. But that, I don't think, is what is causing it. |
| 20:32 | <hsivonen> | IIRC, Technorati stopped indexing old WP installs due to something like this |
| 20:32 | <gsnedders> | hsivonen: sp.o is running the latest release |
| 20:32 | <Philip`> | gsnedders: "<div class="blogimage"><img src="/images/feature_feed.png" alt="[Feed Icon]" /></div> <p><strong><em>SimplePie is a very ..." --> "<div class="blogimage"><img src="/images/feature_feed.png" alt="[Feed Icon]" /></div> "I suppose laterad", barracked corpus." |
| 20:32 | <hsivonen> | ah. ok |
| 20:32 | <Philip`> | shepazu: Is svg-whiz.com using Wordpress? |
| 20:33 | <shepazu> | I did have my blog hacked recently, but svg-whiz doesn't have a blog, it's just HTML |
| 20:33 | <Philip`> | Aha |
| 20:33 | <gsnedders> | hsivonen: WP isn't even touch the home page |
| 20:33 | <Philip`> | Better question: Your site is hosted on Dreamhost, isn't it? |
| 20:33 | Philip` | thanks bsmedberg for noticing that :-) |
| 20:33 | <gsnedders> | Mine is, and I'm oh-so-tempted to blame them :P |
| 20:34 | <annevk> | oh crap, are all dreamhost sites affected? |
| 20:34 | <annevk> | or was this from some time ago? in which case i'm not |
| 20:35 | <Philip`> | annevk: The Google cache things are from the past few days, so it can't have been that long ago |
| 20:36 | <gsnedders> | It still happens for some people at Y!, I know |
| 20:36 | <Philip`> | shepazu: In that case, I think you don't need to do anything, since it's all Dreamhost's fault |
| 20:36 | <shepazu> | evil dreamhost! |
| 20:37 | <shepazu> | (though I do like them in general) |
| 20:37 | <Philip`> | Possible temporary fix: Remove all <p> tags from your pages ;-) |
| 20:37 | <shepazu> | Philip`: done, I only use SVG semantics on my sites ;P |
| 20:40 | <gsnedders> | shepazu: I don't like them in general :P |
| 20:47 | <hsivonen> | does JS intern new strings by default? |
| 20:47 | hsivonen | is surprised by http://code.google.com/p/google-web-toolkit/source/browse/releases/1.5/user/super/com/google/gwt/emul/java/lang/String.java?r=2940#488 |
| 20:49 | <Philip`> | hsivonen: As far as I'm aware, there's no way to distinguish the interned vs non-interned cases in JS |
| 20:49 | <Dashiva> | "CHECKSTYLE_OFF: This class has special needs." |
| 20:50 | <hsivonen> | Philip`: why is the impl. I referenced return new String(this); instead of return this, then? |
| 20:51 | <Philip`> | hsivonen: No idea |
| 20:52 | <annevk> | hsivonen, it seems to invoke intern() and not return new String(this) |
| 20:53 | <annevk> | euh, wait |
| 20:53 | <annevk> | it seems to not do anything |
| 20:53 | <hsivonen> | annevk: /*-{ ... }-*/ is JSNI syntax--not a comment |
| 20:53 | <Philip`> | Is "String" JS's String, or is it a magic GWT class variable? |
| 20:54 | <annevk> | hsivonen, interesting |
| 20:54 | <hsivonen> | Philip`: String in Java scope is java.lang.String and String inside /*-{ ... }-*/ is JavaScript String as far as I can tell |
| 20:55 | <Philip`> | Hmm, too much magic for me to understand :-p |
| 20:55 | <Philip`> | (I was wondering what type 'this' was, but then I decided it's probably safer to not know) |
| 20:56 | <annevk> | Dean Edwards about GWT / JSNI: "This is basically an admission that in the browser environment, JavaScript is king. If you ultimately need to fall back to JavaScript to fix your abstraction layer then your abstraction layer is leaking. JavaScript is king in the browser and GWT is for cowards." |
| 20:56 | <hsivonen> | Philip`: this seems to be the JS String |
| 20:57 | <hsivonen> | annevk: that's like saying that JNI is an admission that C is king |
| 20:57 | <hsivonen> | which it is |
| 20:57 | <hsivonen> | but HLLs are still useful |
| 20:57 | <annevk> | I thought the last bit was funny |
| 21:21 | <gsnedders> | smedero: Where is the guy you got to test it? |
| 21:22 | <smedero> | San Fran |
| 21:22 | <smedero> | Flickr offices |
| 21:23 | <gsnedders> | smedero: thx. |
| 21:23 | Philip` | tries to update the spec splitter to use lxml, since that's a bit faster than minidom |
| 21:24 | <gsnedders> | Philip`: Don't forget .tail! |
| 21:24 | <gsnedders> | Philip`: That is the most horrible thing ever. |
| 21:25 | <Philip`> | Oh, that looks yucky |
| 21:29 | <Philip`> | Hmm, 13s to parse the (postprocessed) spec, when using lxml and psyco - that doesn't seem too awful |
| 21:30 | <Philip`> | (~16.5s without psyco) |
| 21:31 | <hsivonen> | Google should write a Python to JavaScript compiler, too |
| 21:32 | <Philip`> | That sounds fun with e.g. Python's yield statement |
| 21:34 | <othermaciej> | you can rewrite generators to make the state explicit |
| 21:34 | <othermaciej> | or rewrite to continuation passing style |
| 21:35 | <Philip`> | Can you do that without having to rewrite the caller too? |
| 21:38 | <gsnedders> | peh. get me a decent lexer in Python first :P |
| 21:47 | <gsnedders> | (I need recursiveness and repetition at the parser generator level) |
| 21:47 | <hsivonen> | http://lists.w3.org/Archives/Public/www-tag/2008May/0206.html |
| 21:50 | <hsivonen> | http://lists.w3.org/Archives/Public/www-tag/2008May/0204.html |
| 21:52 | <annevk> | www-tag is funny these days |
| 21:54 | <Philip`> | Argh, lxml's child iteration seems to be skipping over the one element I want to find :-/ |
| 22:27 | <hsivonen> | I had an idea: 1) doc.write <math></math>. 2) Sniff what namespace it was put in. 3) Remove it from the DOM. 4) If it wasn't in the MathML namespace, doc.write a script tag for an HTML5 parser in JS. |
| 22:27 | <hsivonen> | 5) doc.write <plaintext style='display:none'> |
| 22:28 | <hsivonen> | 6) attach the tree builder of the JS parser to the html, head and body elements of the current doc |
| 22:28 | <hsivonen> | 7) take the textContent of the plaintext element |
| 22:28 | <hsivonen> | 8) Remove plaintext |
| 22:28 | <hsivonen> | 9) feed its textContent to the parser |
| 22:28 | <hsivonen> | feasible? |
| 22:29 | <hsivonen> | even subsequent doc.write could work if browsers allow document.write to be replaced with a JavaScript function |
| 22:30 | <Hixie> | sounds pretty sick to me :-) |
| 22:31 | <Hixie> | what license do you waht the live dom viewer in? |
| 22:31 | <hsivonen> | Hixie: MIT license would be the easiest |
| 22:33 | <Hixie> | ok, it's available under the MIT license |
| 22:34 | <hsivonen> | Hixie: thank you |
| 22:34 | <Hixie> | np |
| 22:35 | <Philip`> | hsivonen: I did a (very primitive) parser (for parsing a non-HTML language) like that a while ago, and I think IE had some difficulties (which I didn't bother debugging) but otherwise it seemed to work |
| 22:36 | <hsivonen> | Philip`: cool |
| 22:36 | <hsivonen> | Philip`: IE doesn't have createElementNS anyway |
| 22:36 | <Philip`> | http://canvex.lazyilluminati.com/misc/sexp.html |
| 22:37 | <hsivonen> | so I ignored IE in my GWT tree builder |
| 22:41 | <Hixie> | i'm kinda starting to like the colon |
| 22:41 | <Hixie> | for aira |
| 22:42 | <Hixie> | it would actually do more harm to xml namespaces than anything else i've tried to do |
| 22:45 | <annevk> | fortunately internal consistency is more important than harming XML namespaces |
| 22:46 | <Hixie> | i dunno, it would in fact harm xhtml too |
| 22:47 | <annevk> | to continue this little debate, harming other languages in order to make HTML look better doesn't seem the best approach |
| 22:48 | <annevk> | (especially if you're meanwhile compromising the HTML syntax somewhat) |
| 22:48 | <Hixie> | no no, only to make xml fail |
| 22:49 | <Hixie> | we still have e.g. svg and mathml in text/html |
| 22:49 | <annevk> | i meant that using the colon would not be consistent with HTML |
| 22:49 | <Hixie> | true |
| 22:50 | <Hixie> | well except for xlink |
| 22:50 | <Hixie> | in svg |
| 22:51 | <annevk> | :) |
| 22:52 | <annevk> | very unfortunate |
| 22:52 | <annevk> | MathML uses it too it seems, in some way |
| 22:52 | <Hixie> | the colon? how? |
| 22:54 | <annevk> | xlink:href |
| 22:54 | <Hixie> | oh? |
| 22:54 | <Hixie> | not in text/html it doesn't :-) |
| 22:54 | <Hixie> | oh well i guess it does actually |
| 22:54 | <annevk> | yeah it does |
| 22:55 | <annevk> | also in MathML3 |
| 22:55 | <annevk> | and 2 probably |
| 22:55 | <annevk> | it also has xml:space, xml:lang and xml:id (I commented on that one) |
| 22:55 | <Hixie> | oh well |
| 22:55 | <annevk> | I also see my:color and my:background in the attributes list... |
| 22:55 | <annevk> | maybe I should review MathML |
| 22:55 | <Hixie> | xml:space does nothing, as far as i can tell, but text/html supports it |
| 22:55 | <Hixie> | xml:id isn't supported |
| 22:56 | <Hixie> | xlink in general i don't expect to actually be supported |
| 22:56 | <Hixie> | though i guess that's up to the UA |
| 22:56 | <Hixie> | (xlink being incompatible with svg makes xlink less attractive, so i guess most UAs won't do it) |
| 22:56 | <Hixie> | (if they do SVG) |
| 22:56 | <annevk> | yeah, makes more sense to simply ignore XLink |
| 22:57 | <annevk> | i hope they change SVG slightly while still feasible to not require namespaces for references |
| 22:57 | <annevk> | by introducing href= and src= attributes where appropriate |
| 22:59 | <Hixie> | i actually hope they don't |
| 22:59 | <Hixie> | redundant attributes cause all kinds of problems |
| 22:59 | <Hixie> | you have to define priority, you have to test that the priority works even with DOM manipulation, etc |
| 23:00 | <Hixie> | it works differently for new and old UAs, etc |
| 23:00 | <Hixie> | and it adds no functionality |
| 23:00 | <Hixie> | just look at the trouble that charset="" has been |
| 23:01 | <annevk> | that's a different case, though I suppose that if we get SVG in text/html without the namespaces requirement it's sort of acceptable |
| 23:01 | <annevk> | "MathML uses the namespace URI http://www.w3.org/ns/mathml-cd for the XML encoding of MathML content dictionaries." |
| 23:01 | <annevk> | -- http://www.w3.org/TR/MathML3/chapter8.html#contm.mcd.format |
| 23:03 | <Hixie> | ooooh |
| 23:03 | <Hixie> | david singer just proposed something to me by e-mail that may solve the alt text issue |
| 23:03 | <Hixie> | how about instead of a new attribute to say when the alt text is an alternative vs being something just saying what kind of image it is |
| 23:04 | <Hixie> | we say that the alt attribute is an alternative unless it starts and ends with a [ and a ] ? |
| 23:04 | <Hixie> | alt='photo' is the alt text for a png of the word "photo", alt='[photo]' is the alt text for a photo with no description |
| 23:04 | <Hixie> | the UA can still do all the magic with alt, _and_ it's probably quite compatible with legacy usage |
| 23:04 | <Hixie> | and it works great in legacy uas |
| 23:05 | <Philip`> | How does it work in legacy AT? |
| 23:05 | <annevk> | what does <img> mean? |
| 23:05 | <Philip`> | (Pronouncing "left square bracket image right square bracket" would be quite irritating) |
| 23:05 | <Hixie> | <img> is non-conforming and doesn't mean anything, and UAs use heuristics to decide how to render it |
| 23:05 | <Hixie> | Philip`: well let's see |
| 23:06 | Hixie | grabs out his Jaws 8 install from under the bed |
| 23:06 | <annevk> | leaving it out still seems nicer |
| 23:07 | <Hixie> | leaving it out has issues with legacy content and doesn't convey all the information it could |
| 23:07 | <Hixie> | (actually saying what kind of image it is) |
| 23:07 | <Philip`> | 0.6% of my alt values are of the form "[...]" |
| 23:08 | <Hixie> | you can get data faster than i can :-) |
| 23:08 | <Philip`> | That's because I can just use grep and perl locally :-) |
| 23:09 | <Hixie> | :-) |
| 23:09 | <Philip`> | http://philip.html5.org/data/alt-in-square-brackets.txt |
| 23:10 | <annevk> | some of those seem like real replacements |
| 23:10 | <annevk> | "[Reset Text Size]" |
| 23:10 | <Philip`> | (These are total numbers of occurrences, rather than number of pages) |
| 23:10 | <annevk> | or "[E-Mail]" |
| 23:11 | <annevk> | but maybe that's not too problematic |
| 23:11 | <Philip`> | Given legacy content, UAs can't tell that e.g. alt="[Search]" is any different to alt="Search" |
| 23:12 | <Philip`> | so either the difference is important, in which case that syntax won't work; or the difference is not important, in which case we don't need any syntax at all and you can just put the not-really-alternative text in the alt attribute directly |
| 23:13 | <annevk> | otoh, it's only 0.6% of the images, that's a small percentage according to some people, right? :D |
| 23:14 | <Hixie> | Philip`: we can't add _any_ syntax without breaking some images. However, that's moot, since even just keepign alt="" as defined in HTML4 breaks a huge number of images. |
| 23:14 | <Philip`> | That's huge compared to how many people will bother adopting whatever requirements we specify for their alt-less critical images :-p |
| 23:14 | <Hixie> | hm maybe this should only apply to images that aren't in lnks |
| 23:14 | <Hixie> | links |
| 23:14 | <Philip`> | so UAs would be better off ignoring the square brackets entirely, because most of the time they'll only be there for tooltip-prettiness instead of for HTML5-spec-conformance |
| 23:14 | <Hixie> | that might solve 90% of our problems here |
| 23:15 | <Hixie> | Philip`: initially maybe |
| 23:16 | <Philip`> | Hixie: I imagine it would take a very long time for a feature that has pretty much no effect for any users to reach 0.6% adoption |
| 23:16 | <Hixie> | it has a small effect, it shows an icon for hte image when the user has images disabled. :-) |
| 23:17 | <Philip`> | Users don't have images disabled |
| 23:17 | <Philip`> | and if they do then they're crazy and don't count |
| 23:19 | <Philip`> | and you're assuming some UAs will implement that icon-hiding behaviour, which seems a dangerous assumption since if they thought it was a good idea then they'd probably have hidden the image icons already |
| 23:22 | <Philip`> | lxml's text/tail thing is actually quite ugly |
| 23:24 | <Hixie> | firefox does hide the image icons |
| 23:24 | <Hixie> | has for years |
| 23:28 | <Philip`> | Oh, okay |
| 23:33 | <Philip`> | It's slightly odd that alt="[..." (8215 matches) is so much more common than alt="{..." (214) and alt="<..." (388) and alt="_..." (285) |
| 23:35 | <Hixie> | maybe we should use { for this then |
| 23:44 | <Philip`> | http://philip.html5.org/data/alt-first-char.txt |
| 23:45 | Dashiva | wonders what the data will be used for |
| 23:45 | Philip` | shrugs |
| 23:56 | <Philip`> | Hmm, html5lib serialiser is slow :-( |
