| 00:00 | <annevk> | I almost replied to html4all's latest e-mail |
| 00:00 | <annevk> | meh |
| 00:40 | <Hixie> | http://lists.w3.org/Archives/Public/www-archive/2008May/0105.html |
| 00:46 | <Dashiva> | "Again, we need an editor who uses his judgment in editing the draft. However, we need that judgment to be informed by the opinions of others in the WG." |
| 00:46 | <Dashiva> | He needs to make up his mind |
| 00:50 | <Hixie> | http://www.secinfo.com/d11MXs.2WQ9.d.htm is probably a good test case for html5 browsers |
| 00:51 | <Hixie> | (my own html5 parser runs out of element id bits parsing it) |
| 00:51 | <Hixie> | (so that means it has more than 64k elements) |
| 00:51 | <Hixie> | hsivonen's parser has a fatal error on line 236 |
| 00:51 | <Hixie> | validator.w3.org returns a 500 |
| 00:51 | <Hixie> | the html5 dumper just returns no data |
| 00:52 | <Hixie> | html5lib dumper, i mean |
| 00:52 | <Dashiva> | What is that thing... |
| 00:55 | <Dashiva> | 4.5 MB. gEBTN says 103430 elements, among them 58 tables, 39093 cells (and one <th>). |
| 00:59 | <Philip`> | html5lib seems happy to parse it, and it only takes a minute |
| 01:37 | <Hixie> | http://gareth53.blogspot.com/2008/05/atmedia2008-day-1-notes.html |
| 01:37 | <Hixie> | very, very rough notes |
| 01:37 | <Hixie> | "HTML5 by Jame Hunt & Lachlan Young" |
| 01:37 | <Hixie> | "html and xhtml - smae syntax, dif vocab" |
| 01:37 | <Hixie> | some amusing typos there :-) |
| 01:37 | <Hixie> | but the rest of the notes are pretty good |
| 01:37 | <Hixie> | bbl |
| 01:43 | <Philip`> | Hmm, the spec splitter runs in 20s on my laptop but 105s on my web server :-( |
| 01:53 | <Dashiva> | Double capitals are a plague on modern society |
| 01:53 | <Dashiva> | "move from IE^ to IE&" |
| 02:29 | takkaria | sighs at RB |
| 05:51 | <roc> | annevk, shepazu: you around? |
| 05:52 | <shepazu> | a little |
| 05:52 | <shepazu> | hitting the sack soon, kinda tired |
| 07:18 | annevk | is now |
| 07:32 | <othermaciej> | so did Google announce anything interesting at Google I/O? |
| 07:32 | <othermaciej> | I heard something about Gears announcements |
| 07:32 | <othermaciej> | (#whatwg is my lazyweb) |
| 07:35 | <hdh> | Hixie said something about his VP publicly endorse HTML5, there was a transcript link too |
| 07:35 | <annevk> | Gears and Android I saw |
| 07:35 | <hdh> | http://www.techcrunch.com/2008/05/28/live-from-google-io/ |
| 07:36 | <othermaciej> | I heard they announced various sites will use Gears, I wonder if those will work with browsers that implement the relevant HTML5 features |
| 07:37 | <annevk> | I saw somewhere that they believed (maybe hoped) Gears would eventually become obsolete, so I hope so, but I'm not sure |
| 07:44 | <hdh> | the implement status box is covering part of the spec text, how can I make it go away? |
| 07:44 | <annevk> | by using the w3.org version |
| 07:47 | <hdh> | user stylesheet FTW |
| 07:55 | <hsivonen> | Hixie: wouldn't [] in alt degrade ungracefully? |
| 07:56 | <hsivonen> | Hixie: also, if a validator complained about missing alt, how would you make the author insert [] instead of e.g. file name? |
| 08:05 | <hsivonen> | http://parsetree.validator.nu/?doc=http%3A%2F%2Fwww.secinfo.com%2Fd11MXs.2WQ9.d.htm WFM without error 500 |
| 08:05 | <hsivonen> | looks like the HTTP connection times out, though |
| 08:06 | <hsivonen> | (Validator.nu has deliberate short timeouts) |
| 08:09 | <annevk> | it prints IOException at the end |
| 08:11 | <hsivonen> | annevk: I'm guessing it's a timeout |
| 08:14 | <annevk> | "In the meaning of technical in the W3C policies it is certainly a technical question whether the document is published or not." |
| 08:14 | <annevk> | -- RB |
| 08:14 | <annevk> | -- http://www.w3.org/2002/09/wbs/40318/wd-html5-may/results |
| 08:14 | annevk | laughs |
| 08:18 | <gavin_> | haha |
| 09:25 | <annevk> | hsivonen, http://ajaxian.com/archives/gwt-15-release-candidate-announced |
| 09:25 | <hsivonen> | annevk: that's what I'm using |
| 09:27 | <othermaciej> | so if you combine GWT and Rhino on Rails, it seems that Google's emerging strategy is to run Java on the client and JavaScript on the server |
| 09:28 | <othermaciej> | (someone should tell them they've got it backwards) |
| 09:31 | <Hixie> | GWT compiles the Java language code to JS for the client |
| 09:32 | <hsivonen> | GWT's <super-source/> needs *much* better docs |
| 09:32 | <Hixie> | so really it'd just be using JS as a bytecode :-) |
| 09:35 | <othermaciej> | I wonder if it would be worth making a restricted-JS-to-JS optimizer |
| 09:35 | <othermaciej> | I mean, I know Google has one internally |
| 09:35 | <othermaciej> | but available to the general public |
| 09:35 | <othermaciej> | for example JS implementations cannot soundly do inlining in the general case, but you could pre-inline simple functions if the author promises not to use arguments or caller properties on functions |
| 09:37 | <Hixie> | i believe the Caja project has something like that |
| 09:37 | <othermaciej> | (though I am not sure what else if anything is better done as a pre-pass; maybe if the author makes some promises about not doing weirdshit with getters and setters you can do some common subexpression elimination) |
| 09:37 | <othermaciej> | Caja is a restricted dialect, is it not? |
| 09:37 | <othermaciej> | oh |
| 09:37 | <Hixie> | yeah but the optimiser might be separate, i'm not sure |
| 09:37 | <othermaciej> | it's restricted in a more severe way than what I have in mind |
| 09:38 | <Hixie> | do you have any opinions on whether it is worth adding a feature to make tabindex scoped in html5? |
| 09:38 | <hsivonen> | aaronlev: ^ |
| 09:38 | <othermaciej> | it's an interesting idea |
| 09:38 | <othermaciej> | though I tend to think 0 and -1 are the only tabindex values anyone should ever use |
| 09:39 | <othermaciej> | I guess scoping makes it potentially possible to use other values in a non-broken way |
| 09:39 | <aaronlev> | scoped in what way? |
| 09:40 | <othermaciej> | aaronlev: I assume what Hixie means is that you can have a container element and set up the tab order inside it, but without affecting tab order of things before the container or after the container |
| 09:41 | <othermaciej> | which would solve the problem that tabindex is currently all-or-nothing per page |
| 09:42 | <aaronlev> | yeah that's a problem, but i'm not sure we'd want to use tabindex with a new meaning |
| 09:42 | <aaronlev> | otherwise what about older browsers |
| 09:42 | <aaronlev> | another problem is reading order for a screen reader |
| 09:42 | <aaronlev> | it's like tabindex but it can link containers that aren't necessarily focusable |
| 09:42 | <aaronlev> | i should say it = aria-flowto |
| 09:43 | <othermaciej> | in an ideal world, document order would be set up right for reading/tabbing order and you use CSS to visually reorder things if needed |
| 09:43 | <aaronlev> | yes |
| 09:43 | <othermaciej> | but that does not work well with some table setups |
| 09:43 | <annevk> | what are the use cases? |
| 09:44 | <othermaciej> | annevk: making a reusable control that drops into any page, which has internal focusable items but their logical tab order is not document order |
| 09:44 | <aaronlev> | for flowto or modified tabindex? |
| 09:44 | <othermaciej> | perhaps a grid control where you want tab to go vertically first, then to the next column, etc |
| 09:44 | <aaronlev> | for flowto, you might have a sidebar for an article, and want to decide where the best point in the article to branch off to the sidebar is |
| 09:45 | <aaronlev> | because navigation is not just abotu focusable objects |
| 09:45 | <othermaciej> | sidebars are an inherently nonlinear form of document organization |
| 09:45 | <aaronlev> | yes things like PDF have dealt with this for years |
| 09:45 | <aaronlev> | you have to decide a reading order |
| 09:45 | <aaronlev> | also there is printing |
| 09:45 | <annevk> | sounds like XBL / datagrid |
| 09:46 | <annevk> | and menu |
| 09:46 | <aaronlev> | i remember similar problems turning page layout files into something for braille embossers |
| 09:46 | <aaronlev> | in that case the doc is a set of pages with each page having a set of objects |
| 09:46 | <aaronlev> | kind of a mess |
| 09:46 | <othermaciej> | annevk: XBL doesn't give you anything to control tab order |
| 09:46 | <othermaciej> | (afaik) |
| 09:46 | <aaronlev> | i don't know, does HTML have any goal to go into more of a word processing / publishing space? |
| 09:46 | <othermaciej> | annevk: the problem is that tabindex as specified currently basically can't be used |
| 09:47 | <othermaciej> | because everything with an explicit tabindex comes before all other things |
| 09:47 | <othermaciej> | HTML doesn't have goals, it's just a markup language |
| 09:48 | <aaronlev> | othermaciej: don't you know what i mean, the most active leaders in html-wg |
| 09:48 | <othermaciej> | I imagine some people have goals to use HTML for more word-processy things |
| 09:48 | <aaronlev> | you guys are some of them |
| 09:48 | <aaronlev> | thanks |
| 09:48 | <othermaciej> | I think a lot of the action in the HTML space right now is about using it as an application format |
| 09:48 | <othermaciej> | I would say Web applications are a more interesting target than Web documents right now |
| 09:48 | <aaronlev> | another interesting use case is key navigable diagrams |
| 09:49 | <aaronlev> | if i'm in an org chart what's the navigation order |
| 09:49 | <othermaciej> | if the org chart is done in SVG, you can make document order anything you want and use that to control the tab order |
| 09:49 | <aaronlev> | changing tabindex in a non-compatible way isn't as good as implementing something new which takes care of the needs completely |
| 09:50 | <othermaciej> | I'm not sure that when Hixie says "scoped tabindex" he necessarily literally means to reuse the tabindex attribute |
| 09:50 | <othermaciej> | of course, we just had a long conversation about it without him clarifying what he meant |
| 09:50 | <annevk> | othermaciej, XBL + nav-index from CSS prolly does |
| 09:50 | <aaronlev> | the truth is, within many containers the last active element is in the tab order, but as a whole the container is just in the tab order once |
| 09:50 | <othermaciej> | annevk: nav-index? |
| 09:50 | <aaronlev> | within the container directional keys may be used to navigate |
| 09:50 | <othermaciej> | annevk: I don't think CSS is the right place to specify tab order |
| 09:50 | <aaronlev> | tabbing is a limited way to think about navigation really |
| 09:51 | <annevk> | othermaciej, why not, if you reorder items, you might want to change tab order too |
| 09:51 | <hsivonen> | I'm inclined to think tabindex shouldn't change for compat reasons |
| 09:51 | <othermaciej> | tab ordering things is tied to the content |
| 09:51 | <hsivonen> | and that authors should always use 0 or -1 |
| 09:52 | <othermaciej> | and is behavioral, not presentational |
| 09:52 | <othermaciej> | and if nav-index is not itself scoped, it does not solve the problem |
| 09:52 | <annevk> | othermaciej, if I position one link before the order in one layout and the other way around in another, I might want to change tab order as well |
| 09:52 | <hsivonen> | it seems to me that in foo-index, "index" is the problem |
| 09:52 | <othermaciej> | I can imagine you could implement scoped tabindex in a way that older browsers would see tabindex="0" |
| 09:52 | <hsivonen> | and ARIA's IDREF-based flow makes more sense |
| 09:53 | <othermaciej> | that is true |
| 09:53 | <othermaciej> | a navigation order should be about next/previous items |
| 09:53 | <othermaciej> | not absolute index |
| 09:53 | <annevk> | hsivonen, there's directional navigation properties as well |
| 09:54 | <annevk> | http://www.w3.org/TR/css3-ui/#keyboard |
| 09:54 | <othermaciej> | nav-index is broken in the same way as tabindex |
| 09:55 | <othermaciej> | up/down/left/right do not substitute for previous or next |
| 09:55 | <othermaciej> | and I still think defining keyboard navigation in CSS does not make sense |
| 09:55 | <othermaciej> | even though CSS can change the visual order |
| 09:55 | <hsivonen> | does C++ have something that roughly matches the semantics of Java's static { ... } block? |
| 09:56 | <hsivonen> | i.e. code that runs once when a .dll/.so/.dylib is loaded? |
| 09:56 | <othermaciej> | if you tell me what Java's static { } block does I can probably tell you |
| 09:56 | <othermaciej> | oh |
| 09:56 | <othermaciej> | nothing that you should actually use |
| 09:56 | <othermaciej> | if you have an extern or static global variable of class type, the constructor runs on startup |
| 09:57 | <othermaciej> | but the order is ill-defined |
| 09:57 | <othermaciej> | and it may or may not work right in shared libraries on various systems |
| 09:57 | <hsivonen> | :-( |
| 09:57 | <othermaciej> | it is better to have all entry points call an initIfNeeded() type function |
| 09:57 | <hsivonen> | ok |
| 09:58 | <othermaciej> | (I think order of global constructor calls is defined within a translation unit but not between translation units) |
| 09:58 | <hsivonen> | or perhaps if the V.nu parser was ported to C++, the static blocks could be replaced with some tormented preprocessor/template code |
| 09:59 | <hsivonen> | I use static {} blocks to initialize a couple of sorted arrays with magic data in them |
| 09:59 | <hsivonen> | I *could* generate the arrays in the right order ahead of compile time |
| 10:00 | <hsivonen> | but then I'd need an additional script to keep in sync |
| 10:00 | <othermaciej> | for cases of magic data arrays |
| 10:00 | <othermaciej> | what I like to do is generate the source code for the array constant from a template file by script |
| 10:00 | <othermaciej> | (we do that for CSS properties, tag names, attribute names, CSS property keywords, JS language keywords, etc) |
| 10:01 | <hsivonen> | ok. thanks. I'll do that if the need arises |
| 10:10 | <Hixie> | sorry, got distracted |
| 10:10 | <Hixie> | this would be a new attribute |
| 10:10 | <Hixie> | which modifies the behaviour of tabindex |
| 10:11 | <Hixie> | so that numbers greater than 0 only consider other tabindex attributes within that same scope |
| 10:11 | <Hixie> | i tend to think that maybe it's not worth it |
| 10:11 | <hsivonen> | Hixie: what's the compat story? |
| 10:12 | <Hixie> | none really, this likely wouldn't be used until this generation was obsolete |
| 10:12 | <zcorpan> | hsivonen: in my experience, generally, pages that use tabindex today have bad usability in today's browsers anyway |
| 10:12 | <Hixie> | aha, zcorpan! i was looking for you earlier |
| 10:12 | <Hixie> | this was originally your diea |
| 10:12 | <Hixie> | idea |
| 10:13 | <zcorpan> | hey Hixie |
| 10:13 | <othermaciej> | Hixie: I think a way to specify next element in the tab order by ID would be more useful, if any such thing is to be added |
| 10:14 | <othermaciej> | (and then you could also specify tabindex="0" if needed or not give a tabindex, for compat) |
| 10:14 | <othermaciej> | but I am not sure it is critical in any case |
| 10:15 | <Hixie> | yeah |
| 10:16 | <zcorpan> | Hixie: i think source order is a powerful enough tool to specify tab order, and browsers would probably be more usable on legacy content if all positive values were treated as 0 |
| 10:17 | <othermaciej> | zcorpan: tables are about the only case where you may want a tab order that is not natural source order, and it's hard to fix the source order w/o breaking the layout |
| 10:17 | <Hixie> | yeah. the spec allows that, iirc |
| 10:17 | <othermaciej> | (at least so far as I can think of) |
| 10:17 | <Hixie> | it seems like this might be a CSS thing anyway |
| 10:17 | <Hixie> | css3 ui does try to take this responsibility |
| 10:17 | <zcorpan> | othermaciej: true |
| 10:17 | <aaronlev> | Should be able to point to id of a non-focusable container |
| 10:18 | <aaronlev> | as next thing to navigate to |
| 10:18 | <othermaciej> | it should be able to point to anything, and tabbing will find the next item in document order from there |
| 10:18 | <zcorpan> | othermaciej: wasn't there a proposal to swap the direction of tables so that rows become columns? |
| 10:19 | <othermaciej> | zcorpan: I do not recall such a thing but I have not followed every proposal |
| 10:19 | <othermaciej> | that particular proposal sounds like it would have a lousy compat story |
| 10:20 | <aaronlev> | othermaciej: after tabbing past the end of the container pointed to, should tabbing continue after the container? Or should it continue on from the original point that branched to the container |
| 10:20 | <aaronlev> | iow |
| 10:20 | <aaronlev> | i might want to say, tab through all the stuff over there before continuing on here |
| 10:20 | <aaronlev> | more like a subroutine than giving up control |
| 10:21 | <othermaciej> | sounds dubious |
| 10:21 | <othermaciej> | if you want that, the target container can point back to after you |
| 10:21 | <othermaciej> | tabbing into the same container from two different points does not sound like good UI |
| 10:22 | <othermaciej> | Cocoa just lets you specify the previous item, the next item, and whether any given item is keyboard focusable |
| 10:22 | <othermaciej> | it works pretty well |
| 10:24 | <aaronlev> | othermaciej: but that's not sufficient for documents |
| 10:24 | <aaronlev> | because for screen readers the reading order of doc text needs to match the nav order or it gets confusing |
| 10:24 | <aaronlev> | so it's better if each container points to the next |
| 10:24 | <aaronlev> | plus it's fewer things to change |
| 10:24 | <aaronlev> | for the author |
| 10:25 | <othermaciej> | I just suggested each thing pointing to the next |
| 10:25 | <othermaciej> | I don't think I understand what is different about your proposal |
| 10:25 | <othermaciej> | (other than being about more than just tab order) |
| 10:26 | <aaronlev> | it might not be different, just IRC communication ambiguity |
| 10:26 | <othermaciej> | that is possible |
| 10:26 | <othermaciej> | back in a few |
| 10:54 | <hsivonen> | boohoo. I get unending unresponsive script dialogs when I try to run the script GWT compiled |
| 10:56 | <hsivonen> | hah. The script crashes Drosera. |
| 11:03 | <zcorpan> | Error: = in an unquoted attribute value. |
| 11:03 | <zcorpan> | At line 234, column 38 |
| 11:03 | <zcorpan> | istrant.asp?CIK=30908 |
| 11:03 | <zcorpan> | http://validator.nu/?doc=http%3A%2F%2Fwww.secinfo.com%2Fd11MXs.2WQ9.d.htm |
| 11:03 | <zcorpan> | hsivonen: looks like a false positive |
| 11:05 | <hsivonen> | zcorpan: per spec http://www.whatwg.org/specs/web-apps/current-work/#attribute4 |
| 11:05 | <zcorpan> | hsivonen: yes i know |
| 11:05 | <zcorpan> | hsivonen: still, false positive in the sense that it wasn't a mistake |
| 11:05 | <Hixie> | hsivonen: drosera crashes a lot for me |
| 11:06 | <Hixie> | anyway bed time |
| 11:06 | <Hixie> | nn |
| 11:08 | <annevk> | zcorpan, hmm yeah, escaping = in URIs would be annoying |
| 11:09 | <zcorpan> | annevk: the straightforward "fix" is to use quotes |
| 11:10 | <annevk> | meh |
| 11:34 | <Philip`> | zcorpan: Forbidding = also detects errors like <img src=whatever.gifalt=something>, which wouldn't be caught by forbidding " and ' |
| 11:34 | <zcorpan> | Philip`: true |
| 11:34 | <annevk> | you'd catch that one by looking at the page |
| 11:35 | <annevk> | (though not for <a rel=xtitle=foo> I suppose) |
| 11:35 | <othermaciej> | drosera is obsolete |
| 11:36 | <hsivonen> | othermaciej: interesting. has the replacement been shipped? |
| 11:36 | <othermaciej> | hsivonen: yeah, debugging is built into the Web Inspector now |
| 11:36 | <othermaciej> | the latest WebKit nightlies don't even include Drosera |
| 11:36 | <hsivonen> | oh. I was unaware |
| 11:36 | <hsivonen> | thanks |
| 11:37 | <othermaciej> | we haven't widely announced it yet, but we should soon |
| 11:37 | <othermaciej> | the new debugger is much more stable and performance |
| 11:37 | <othermaciej> | *performant |
| 11:38 | <zcorpan> | Philip`: perhaps we should pages that have = in unquoted attributes to get an idea whether it helps or hurts |
| 11:38 | <zcorpan> | s//look at/ |
| 11:41 | <zcorpan> | yay.. http://lists.w3.org/Archives/Public/w3c-wai-ig/2008AprJun/0087.html |
| 11:44 | <zcorpan> | (aria editor's drafts are now public) |
| 12:18 | <Philip`> | zcorpan: http://www.rottentomatoes.com/m/changing_lanes/ -- <iframe valign=top WIDTH= HEIGHT= MARGINWIDTH=0 ...> |
| 12:19 | <Philip`> | http://www.ahaus-online.de/ -- <table width=100%cellpadding="4" cellspacing="1"> |
| 12:19 | <zcorpan> | Philip`: ooh interesting |
| 12:20 | <Philip`> | Most of the cases look like <a href=foo?bar=baz> |
| 12:22 | <annevk> | <table width> is not conforming anyway :) |
| 12:22 | <Philip`> | zcorpan: If you want some extremely rough raw data, see http://philip.html5.org/data/some-unquoted-attributes-with-equals.txt |
| 12:23 | <Philip`> | (That's all lines (for the first n pages before I got bored and stopped it) matching /<[a-zA-Z0-9:]+[^"'>]+\s[a-zA-Z0-9:]+\s*=\s*[^\s>"']+=/ which seems to kind of work alright as an approximation) |
| 12:24 | <Philip`> | http://www.specialtybottle.com/: <INPUT type=hidden value=BGCOLOR=#FFFFFF name=nonUPS_body> |
| 12:24 | <zcorpan> | Philip`: don't you think the value was intended to be "BGCOLOR=#FFFFFF"? |
| 12:25 | <zcorpan> | Philip`: the other two cases would get caught by other errors |
| 12:25 | <Philip`> | Oh |
| 12:25 | <Philip`> | That's possible |
| 12:26 | <Philip`> | http://www.gencat.net:8000/osial/owa/p01.dad_ens?cod=1708650006: <TD VALIGN=TOP BGCOLOR=6A92E3><B><FONT FACE=ARIALSIZE=1 color="#FFCC66">DADES GENERALS</FONT></B></TD> |
| 12:26 | <Philip`> | <table width= otherstuff=etc> seems surprisingly common |
| 12:27 | <zcorpan> | seems people think that whitespace starts a new attribute |
| 12:27 | <zcorpan> | perhaps that should be banned? <p class= foo>? |
| 12:28 | <gsnedders> | I should probably blog something this month before this month ends. |
| 12:29 | <annevk> | i don't think so |
| 12:30 | <Philip`> | gsnedders: For any particular reason? |
| 12:30 | <gsnedders> | Philip`: Because I normally blog at least once per month :P |
| 12:32 | <zcorpan> | <TABLE width=240 height=300 border=0<cellspacing=0 cellpadding=0 bgcolor="083a81"> |
| 12:35 | <Philip`> | Far too many people use <table> :-/ |
| 12:35 | <zcorpan> | <FONT face=Arial,color=#ffffcc> |
| 12:38 | <Philip`> | (and <font>) |
| 12:39 | <Philip`> | (and uppercase tag names) |
| 12:39 | <Philip`> | (and unquoted attributes) |
| 12:39 | <Philip`> | (XHTML is much prettier, except for all the closing-tag stuff) |
| 12:43 | <zcorpan> | <a class=chlnk href onclick=this.style.behavior='url(#default#homepage)';this.setHomePage('http://www.elarcadenoe.org';); style=CURSOR:hand> |
| 12:44 | <zcorpan> | i think that's the first time i've seen <a href> in the wild |
| 12:45 | <virtuelv> | hm, new favicon for Google |
| 12:49 | <Philip`> | zcorpan: You haven't been looking at many pages if that's the first <a href> you've seen :-p |
| 12:50 | <zcorpan> | Philip`: i mean as opposed to <a href=''> |
| 12:50 | <zcorpan> | or <a href=""> |
| 12:50 | <Philip`> | I mean that too |
| 12:50 | <zcorpan> | ok |
| 12:51 | <Philip`> | But I can cheat and use grep instead of actually looking at pages :-) |
| 12:51 | <zcorpan> | :P |
| 12:51 | <Philip`> | Hmm, maybe it's not actually extremely common |
| 12:52 | <Philip`> | http://philip.html5.org/data/some-minimised-href.txt is some (limited to those where the href is the first attribute) |
| 12:53 | <zcorpan> | still not many pages |
| 12:53 | <Philip`> | That's because my computer's too slow and hasn't finished checking all the pages I've got :-( |
| 12:54 | <zcorpan> | hmm, <meta name="keywords" content="<meta name=verify-v1 content=B+yjuq6j1MLG1uzPsO3Z2gR2ECWibgtutlpJeA/3g7k= />"> |
| 12:54 | <Philip`> | http://dementia.seesaa.net/: <a href><a href="http://dementia.seesaa.net/article/9909416.html">FmÇÌè`</a><br>; |
| 12:55 | <Philip`> | (http://philip.html5.org/data/some-minimised-href.txt now updated a bit) |
| 13:00 | <Philip`> | http://www.pscelex.com/: <a href HREF="quotes.html" class="psc1">Quotes</a> |
| 13:00 | <Philip`> | The HTML5 parser breaks that |
| 13:00 | <Philip`> | but maybe that's okay since Firefox breaks it already |
| 13:01 | <Philip`> | ...and so does IE |
| 13:01 | <Philip`> | (but not Opera) |
| 13:02 | <Philip`> | (Uh... Not Opera 9.2) |
| 13:02 | <Philip`> | (but 9.5 does break it) |
| 13:04 | <annevk> | 9.5 has bug fixes :) |
| 13:05 | <Philip`> | And bugs, since it says there's two 'href' attributes on the element :-p |
| 13:05 | <Philip`> | (both with value "") |
| 13:06 | <Philip`> | (though innerHTML shows the two attributes with distinct values) |
| 13:07 | <Philip`> | zcorpan: http://philip.html5.org/data/some-minimised-href.txt is now finished updating, showing all lines matching /(?i)<a\s*href\b\s*[>a-z]/ |
| 13:08 | <Philip`> | so I guess you were right that it's not particularly common :-( |
| 13:08 | <zcorpan> | http://tinyurl.com/5e2c7s - trimmed list (removed rows with src=, href=, action=, rows that were just js, and rows that contained URLs with = ) |
| 13:09 | <zcorpan> | uh |
| 13:09 | <zcorpan> | why doesn't it work |
| 13:09 | <Philip`> | Works for me |
| 13:09 | <Philip`> | assuming 'it' == that tinyurl |
| 13:10 | <Philip`> | and not in Firefox |
| 13:10 | <zcorpan> | http://preview.tinyurl.com/5e2c7s works |
| 13:12 | <Philip`> | <META name= robots= content "INDEX, FOLLOW" > is impressively wrong |
| 13:53 | <zcorpan> | Philip`: yeah |
| 13:57 | <zcorpan> | hsivonen: <!DOCTYPE html><title></title><embed x "y src=''> validates in v.nu |
| 14:02 | <zcorpan> | uh, i can't spell to majority |
| 14:05 | <zcorpan> | hsivonen: trying to validate the empty string using the textarea doesn't seem to work (the textarea becomes disabled) |
| 14:06 | <Philip`> | zcorpan: foo=="bar" wouldn't have shown up because my regexp didn't match that |
| 14:08 | <zcorpan> | Philip`: ah |
| 14:09 | <Philip`> | but if I do search specifically for <something something== then I find mostly |
| 14:09 | <Philip`> | if (p<1 || p==(val.length-1)) errors+='- '+nm+' must contain an e-mail address.\n'; |
| 14:09 | <Philip`> | except for: |
| 14:09 | <Philip`> | http://www.epinions.com/game-Titles-Consoles-SonyPlaystation-All-Brave_Fencer_Musashi: <span class=lkr>Enter your ZIP code to see if there is tax & shipping </span><span class="rkr"><input name="zip" type="text" value="Your ZIP" maxlength="9" size="10" style="color:#7D7D7D" onfocus="javascript:clearDefault(this)"> <input name=="zip_submit" align="middle" type="submit" value="Enter"></span><br> |
| 14:09 | <Philip`> | http://www.angelfire.com/movies/bcrudup/: <font color==#9370D8> |
| 14:09 | <Philip`> | http://www.delhieducation.net/: <td colspan==2 height=10> |
| 14:10 | <Philip`> | http://www.watsonvillehs.net/: <TD BACKGROUND=="http://www.watsonvillehs.net/images/departments07/home_bw.jpg"; ALIGN="CENTER" > |
| 14:10 | <zcorpan> | so it does occur, that's good to know |
| 14:10 | <Philip`> | http://www.ottosell.de/jds/jdsworks.htm: <meta name=="revisit-after" content="20 days"> |
| 14:10 | <Philip`> | http://www.caribbean-travel.com/airnegril/: <P align=left><FONT color=red face=Arial size==+2><STRONG>In US Toll |
| 14:10 | <Philip`> | and more |
| 14:10 | <zcorpan> | feel free to reply that to the list :) |
| 14:12 | <mpt> | Philip`, stop it, you're depressing |
| 14:12 | <mpt> | :-) |
| 14:13 | <Philip`> | mpt: Don't blame me, blame the web :-) |
| 14:26 | <gsnedders> | Damned web. |
| 14:26 | <gsnedders> | We need Web 2.0! |
| 14:26 | <gsnedders> | Oh, wait. |
| 14:26 | <gsnedders> | Web π! |
| 14:28 | <zcorpan> | gsnedders: is Web π where authors don't make mistakes? |
| 14:28 | <gsnedders> | zcorpan: Yeah, and where XHTML 2.0 and XML validating processors are used |
| 14:29 | <zcorpan> | gsnedders: then hsivonen would be out of business |
| 14:30 | <gsnedders> | zcorpan: True, but Web π is an idealistic dream, and nothing more. |
| 14:43 | <takkaria> | I think it's a bad ideal, too |
| 15:17 | <Philip`> | gsnedders: http://benjamin.smedbergs.us/blog/2008-05-30/google-cache-only-spam/ |
| 15:21 | <gsnedders> | Philip`: Yeah, he emailed us too |
| 15:23 | <takkaria> | hmm, http://robert.accettura.com/blog/2008/05/29/rebreaking-the-web/ |
| 15:28 | <annevk> | Rome wasn't built in a day |
| 15:28 | <gsnedders> | build('Rome', 86400); |
| 15:28 | <gsnedders> | Done. |
| 15:31 | <takkaria> | I'm just not sure what he proposes as an alternative |
| 15:31 | <annevk> | no |
| 16:23 | <annevk> | the stronger people advocate modularization, the more I start to think it's not that easy |
| 16:40 | <gsnedders> | School now officially sucks |
| 16:40 | <gsnedders> | I am no longer in Secondary 5. |
| 16:40 | <gsnedders> | I am now in Secondary 6 :( |
| 17:33 | <hsivonen> | is there a reason why <!DOCTYPE html><title></title><embed x "y src=''> should not validate per spec? |
| 17:41 | <annevk> | yeah "y is not conforming |
| 17:44 | <hsivonen> | annevk: which part of the spec says that "y is not conforming? |
| 17:45 | <Philip`> | hsivonen: http://www.whatwg.org/specs/web-apps/current-work/multipage/section-writing0.html#attributes1 |
| 17:45 | <Philip`> | "Attribute names must consist of one or more characters other than the space characters, U+0000 NULL, U+0022 QUOTATION MARK ("), U+0027 APOSTROPHE ('), U+003E GREATER-THAN SIGN (>), U+002F SOLIDUS (/), and U+003D EQUALS SIGN (=) characters, the control characters, and any characters that are not defined by Unicode." |
| 17:46 | <hsivonen> | Philip`: but the writing section isn't normative for conformance checking |
| 17:46 | <hsivonen> | Philip`: it's just a restatement of something else |
| 17:46 | <hsivonen> | (potentially a buggy restatement) |
| 17:47 | <Philip`> | If it was just a restatement of a normative requirement, it shouldn't say "must", I would have thought |
| 17:49 | <hsivonen> | time to send email, I guess |
| 17:55 | <annevk> | whoa, why does Mozilla want to make Access Control so damn complicated :( |
| 17:58 | <hsivonen> | whoa! when did Hixie put " and ' back as tokenizer parse errors in the attribute name state? |
| 17:59 | <hsivonen> | hmm. I even have it implemented |
| 18:00 | <hsivonen> | the problem is in the previous state |
| 18:00 | hsivonen | is puzzled |
| 18:02 | <hsivonen> | hmm. it only happens if the previous attribute is a boolean attribute |
| 18:03 | <hsivonen> | email sent |
| 18:05 | <hsivonen> | I need to go through all the intra-tag parse errors and add probable causes to the messages |
| 18:07 | <hsivonen> | annevk: it seems to me that the AC cookie issue could be mitigated by instructing servers to make same-site-XHR duplicate the cookie value in a custom header |
| 18:09 | <hsivonen> | annevk: it might be worthwhile to add a Note about verifying that a request is same-site by checking for a custom header that duplicates the cookie |
| 18:09 | <hsivonen> | annevk: since non-same-origin site could not have read the cookie in order to set the custom header |
| 18:10 | <hsivonen> | (trick from the GWT security guide) |
| 19:09 | <hsivonen> | http://www.oreillynet.com/xml/blog/2008/05/ubiquity_xforms.html |
| 19:14 | <hsivonen> | what's the relationship of Backplane Ltd. and x-port.net Ltd.? |
| 19:15 | <hsivonen> | they seem to share the address and the managing director |
| 19:15 | <annevk> | hsivonen, if it is a real issue suggesting workarounds for servers is probably not a good solution |
| 19:16 | <annevk> | hsivonen, because people won't do it |
| 19:16 | <annevk> | I'm not really convinced it's a real issue though |
| 19:16 | <hsivonen> | annevk: Google seems to be serious about suggesting cookie duplication as an anti-CSRF measure |
| 19:17 | <hsivonen> | it's already needed for protecting same-site XHR endpoints that could be fooled by cross-site form POST or the like |
| 19:43 | <Philip`> | Handy hint: When upgrading GRUB on Gentoo, check whether it says it's going to make your system unbootable unless you do the MBR setup process again |
| 20:32 | <hsivonen> | w00t. I have the V.nu parser runnig as JS |
| 20:32 | <hsivonen> | it needs a setTimeout event pump before anyone can actually use it for anything useful |
| 20:40 | <Philip`> | :-o |
| 21:24 | <hsivonen> | so firebug crashes on me when I try to debug an unresponsive script |
| 21:25 | <hsivonen> | and Web Inspector in WebKit beach balls on unresponsive script and doesn't offer to debug it |
| 21:29 | <hsivonen> | looks like there'd be a use case for Canvas 3D: http://googlesystem.blogspot.com/2008/05/google-earth-browser-plugin.html |
| 21:49 | <kingryan> | hsivonen: how are you running it as javascript? |
| 21:50 | <hsivonen> | kingryan: compiled it with GWT with a JS entry point |
| 21:50 | <hsivonen> | however, the tokenizer eof() method goes into an infinite loop in JS but not in Java bytecode |
| 21:50 | <Philip`> | Does GWT produce readable JS code? |
| 21:50 | <hsivonen> | kingryan: with the browser's native DOM as the tree impl |
| 21:51 | <hsivonen> | Philip`: it can produce semi-readable code which is how I figured what loop it gets stuck into |
| 21:51 | <hsivonen> | Philip`: by default, the output is *totally* unreadable |
| 22:04 | <Hixie> | hsivonen: if you can find a way to reproduce it in a small test file, i can report the bug |
| 22:05 | <hsivonen> | Hixie: do you mean small HTML input or a minimized Java case? |
| 22:05 | <Hixie> | small gwt input; minimised java case |
| 22:05 | <hsivonen> | Hixie: that would be *hard* |
| 22:05 | <hsivonen> | Hixie: I'll try to debug this a bit more in the morning |
| 22:05 | <Hixie> | yeah, i figure it would be |
| 22:06 | <Hixie> | realistically i doubt the gwt team would have the time to track down the bug if they had to learn your code too :-) |
| 22:06 | <hsivonen> | yeah :-( |
| 22:06 | <hsivonen> | I'll see if the method behaves the same way in isolation |
| 22:07 | <Hixie> | it's an open source project though right? there should be a public bug reporting form if you want to just attach the entire project with steps to reproduce |
| 22:07 | <Hixie> | also does it do the same thing in other UAs? |
| 22:07 | <hsivonen> | yeah, there's a public bug tracker |
| 22:07 | <hsivonen> | same problem both in Gecko and WebKit |
| 22:09 | <Hixie> | k |
| 22:09 | <Hixie> | probably a fundamental compiler bug then unless it's actually a bug in your code somehow |
| 22:09 | <hsivonen> | it works in the hosted mode |
| 22:10 | <hsivonen> | which suggests it is a compiler bug |
| 22:10 | <Hixie> | yep |
| 22:12 | <hsivonen> | nn |
| 22:26 | <smedero> | "TBL: I feel it is important that the HTML5 spec be split into smaller chunks." http://www.w3.org/2001/tag/2008/05/21-minutes |
| 22:26 | <Hixie> | i think everyone agrees with that |
| 22:26 | <Philip`> | It's fun trying to install Python 2.3 modules on Gentoo, since I have to make the /usr/bin/python symlink point to python2.3 so they get installed for 2.3, but Gentoo's package manager relies on Python 2.4 features :-( |
| 22:27 | <zcorpan_> | Hixie: i don't think it's particularly important |
| 22:27 | <Philip`> | jgraham / annevk / etc: Is it know that several tests like "32-bit hex numeric entity" fail in Python 2.3? |
| 22:27 | <Philip`> | s/know/known/ |
| 22:27 | <Philip`> | (...in html5lib) |
| 22:27 | <Philip`> | ("OverflowError: string/unicode conversion") |
| 22:32 | <bzed> | jgraham: hey :) any news on a html5lib release? |
| 22:32 | <zcorpan_> | hsivonen: i agree that always allowing = in unquoted attributes seems bad |
| 22:33 | <zcorpan_> | hsivonen: src and href (and action) are not the only attributes that take uris in practice, <option value> in particular was pretty common |
| 22:35 | <zcorpan_> | hsivonen: (or rather, letting the authoring mistakes in go uncatched seems bad) |
| 22:35 | <zcorpan_> | s/in // |
| 22:37 | <zcorpan_> | hsivonen: moreover, uris were not the only cases of deliberate use of = |
| 22:40 | <annevk> | Philip`, I forgot whether we cared about Python 2.3 |
| 22:41 | <annevk> | Philip`, I guess we might |
| 22:41 | <Philip`> | annevk: It's weird since it works fine when I run test_tokenizer.py, but a few of those tokeniser tests fail when I run runtests.py :-/ |
| 22:42 | <Philip`> | Does an earlier test do something funny to the Python environment? |
| 22:47 | <Philip`> | Oh, test_parser.py does warnings.simplefilter("error") |
| 22:54 | <Philip`> | Hmm, the problem is too non-obvious and hard to reproduce :-( |
| 22:54 | Philip` | therefore gives up on it |
| 23:19 | <Philip`> | Why are EBCDIC encoding not a "must not support"? They seem to have the same security problems, if you write a U+0014 and convince the user's browser to decode it as EBCDIC to get a '<' |
| 23:19 | <Philip`> | *encodings |
| 23:20 | <Philip`> | *same security problems as UTF-7 etc |
| 23:31 | <Philip`> | s/U+0014/U+004C/ |
| 23:49 | <Hixie> | the problem with UTF-7 is that IE autodetected it too eagerly, so non-UTF-7 pages let people smuggle in UTF-7 stuff that got treated as such |
| 23:49 | <Hixie> | does it do that with EBCDIC? |
| 23:50 | <Hixie> | the only difference really is that EBCDIC has more than zero users |
| 23:50 | <Hixie> | probably not much more than zero, but probably enough that we don't want to make them non-conforming |
| 23:55 | <Philip`> | I have no idea how much support/usage EBCDIC has |
| 23:57 | <gsnedders> | The fun of EBCDIC is all the different codepages |
| 23:59 | <Philip`> | It just seems like UTF-7 and EBCDIC have the same inherent technical problems, so they should both be handled the same, i.e. either both permitted (but require that they don't get autodetected) or both forbidden |
![http://www.watsonvillehs.net/images/departments07/home_bw.jpg"](http://www.watsonvillehs.net/images/departments07/home_bw.jpg"){kind=link}