HTML WG face-to-face meeting — 23 October 2008

See also: IRC log

MikeSmith reviews agenda

Cynthia: unconference topic: implicit accessibility roles/topics

Lachlan: authoring guide

MikeSmith: exec4 is reserved for unconference sessions this PM


<mjs> ooh, implicit accessibility rules

<mjs> *roles

MikeSmith: re testing, I suggest we update the annotations on the spec about which parts are stable, tested, etc.

<anne> they are inside the WHATWG version of the spec DanC_lap

<anne> in the sideline

there's some php interface to update them, yes?

<Hixie> perl, but yes

<Hixie> alt-double-click a section to update that section

<Hixie> you have to log in first (link at the top right) -- if you don't have an account (most people who have sent feedback automatically have one set up) let me know

hmm... one day I'm pretty sure I figured out how to just GET the annotations

<anne> there's an API somewhere

<Hixie> DanC_lap: http://www.whatwg.org/specs/web-apps/current-work/status.cgi?action=get-all-annotations

<Hixie> DanC_lap: i think there's documentation but i can't find it at the moment. drop me a mail if you want me to hustle some up

the docs were sufficient last time I needed it

Agenda planning + scheduling time for discussion of open issues

ISSUE-13 handling-http-401-status

<gsnedders> http://www.whatwg.org/issues/#WF3-httpauth

<pimpbot> Title: WHATWG Issues List (at www.whatwg.org)

Julian_Reschke: I saw discussion on the whatwg mailing list about authentication...

<Hixie> http://www.whatwg.org/issues/#WF2-http-auth-login-logout

<pimpbot> Title: WHATWG Issues List (at www.whatwg.org)

Julian_Reschke: web designers rarely used http authentication...
... if we could [missed], it might help

Hixie: I have a pile of mail on this authentication issue

<Julian_Reschke> http://www.w3.org/html/wg/tracker/issues/13

<pimpbot> Title: ISSUE-13 - HTML Issue Tracking Tracker (at www.w3.org)

Anne: it's probably important to handle logout at the same time

Hixie: this would make a good unconference session topic... I'd like to brainstorm... I haven't seen a good transition strategy

AlGilman: want to bookmark some connection... [missed]

Hixie: while we're at it, we'd like people to use digest rather than basic

Julian_Reschke: yes, there's a bit of a chicken-and-egg deadlock between http protocol design and browser development
... co-chair WAI PF WG

ISSUE-20 table-headers

Al: issue summary: data presented in tables depends on context...
... in a visual scan, it's usually easy to scan to the top of the column or start of row to get context
... we [WAI?] have been asking that the context be machine readable...
... assistive technology provides a gesture for asking for this context for a cell
... we've seen some proposals... we're interested in deployment in browsers of some algorithms

MikeSmith: discussion supports the utility of the functionality in general...
... the disagreement in the discussion is about whether the @headers attribute may refer to <td> elements or is constrained to <th>

[refinements from LH/HS... too fast for scribe... can you guys write them down?]

<anne> (From what I gathered during a PF meeting having headers for headers would be enough.)

<Laura> Proposals: http://tinyurl.com/6phdwg

<pimpbot> Title: HTML/IssueTableHeaders - ESW Wiki (at tinyurl.com)

<anne> (Example why that is needed anyway: http://lists.w3.org/Archives/Public/www-archive/2007Aug/att-0003/offset-mess.htm )

<pimpbot> Title: layout height attributes on body and html elements (at lists.w3.org)

AlG: some [hallway? PF?] discussion made progress... [something about table header chaining]

MS: I'd like to have Josh before we get too much further in

Hixie: I'm about 86 messages behind on discussion of use cases for this design issue

AlG: see "function and impacts" thread

<smedero> Al's email: http://lists.w3.org/Archives/Public/public-html/2008Sep/0362.html

<pimpbot> Title: function and impacts (was: scope and headers reform) from Al Gilman on 2008-09-14 (public-html@w3.org from September 2008) (at lists.w3.org)

AlG: see "function and impacts" thread for a re-cap and high-level framing

MichaelCooper: the discussion around use cases seem more productive than discussion of tags/attributes/conformance

<Zakim> MichaelC, you wanted to suggest the use cases need to be further understood before getting into conformance rules

Hixie: quite. I always consider use cases before making markup design decisions

MurrayMaloney: there's a lot of existing practice since 1994... why not use that?

Hixie: we're re-considering design of many[all?] HTML details in the light of another 10 years of experience

AlG: to recap this week's discussion briefly, there's room for improvement... what's in the field is arduous for authors

MS: so I hear relevant parties are more likely to be available tomorrow PM

AlG: but I'm not sure I have Josh tomorrow

ISSUE-37 html-svg-mathml

MS: SVG WG asked for more discussion before releasing a draft including an earlier proposal

MM: and MathML?

MS: MathML advocates seem satisfied with current draft on MathML integration

<smedero> SVG WG's counter proposal to the HTML WG is on their wiki: http://www.w3.org/Graphics/SVG/WG/wiki/SVG_in_text-html

<pimpbot> Title: SVG in text-html - SVG (at www.w3.org)

ISSUE-41 Decentralized-extensibility

<gsnedders> The current draft is http://dev.w3.org/SVG/proposals/svg-html/svg-html-proposal.html AFAIK

<pimpbot> Title: SVG and HTML (at dev.w3.org)

MS: the tech plenary discussion yesterday touched on this.

AlG: is the TAG session intended to cover this?

MS: given the time, the TAG expressed a preference to discuss modularization. Some TAG members particularly interested in distributed extensibility aren't here

<anne> http://lists.w3.org/Archives/Member/tag/2008Oct/0031.html (Member-only)

<anne> (in response to gsnedders)

<smedero> Julian_Reschke: didn't you say you found a new info on ISSUE-54? That there was a similar problem with ASP.NET?

<Julian_Reschke> on issue 52: do we have a separate issue for other issues for current HTML generators, such as wrt to new empty tags?

<Julian_Reschke> smedero: yes, the "xslt-compat" name is misleading; it's needed for more content producers

no, I haven't seen a separate issue on new empty tags, though I'd want to see a more concrete problem/issue before adding it to the list

<smedero> Julian_Reschke: have you sent an email on that anywhere? I was going to link that up to the issue... if not, nevermind.

<Hixie> we could change it to "legacy-compat" or some such

eek... "legacy-compat" sounds like a huge swap

<Hixie> that's the idea :-)

<Julian_Reschke> smedero: can't recall; maybe it was mentioned on IRC somewhere

swamp, I meant

<Hixie> we need something that sounds bad so that people don't think it's the more cool thing

MS: tomorrow at 2pm for SVG/HTML?

<anne> Hixie, maybe just "compat"

MS: how many ppl? 12-ish

"compat" with what? more specific, please

<anne> tools that can't generate <!DOCTYPE html>


ah. that suggest overlap with issue-4 "HTML Versioning and DOCTYPEs"


<fantasai> ScribeNick: fantasai

Mike: Next major issue is modularization of the spec
... The TAG has concerns about this

Anne: Shouldn't we discuss other topics?

Mike: About the authentication discussion

Mike asks about scheduling

Mike: So from 4:45 until ... 5:30?
... For the authentication discussion

Spec-splitting and finding/recruiting editors

Mike: So from 4:45 until ... 5:30?
... For the authentication discussion
... So just before we take a break and before we have the TAG members show up
... Does anybody have any thoughts on modularization?
... So the issue is .. we've had this discussion a lot ourselves
... I think there's general consensus to split out certain parts of the spec
... The issue is do we have editors that are willing to work on these separate parts. That's been the biggest blocking factor
... ... discussion with the TAG. That's where we're at as a WG with the issue
... Any other thoughts on that before we talk with TAG?

?: Do you want to take a shot at explaining that?

MM: Whether there's one editor or three editors, there's still one document or multiple documents

Hixie explains that the overhead of editing multiple specs is high

MM: That's not my experience

marcos: It is mine

<Julian_Reschke> nor mine

<DanC_lap> +marcos

Mike: We have had some discussions about modularization, but the resolution -- or non-resolution -- was that we haven't had people volunteer to take on other parts of the spec
... A specific example of this is ...

<DanC_lap> Marcos Caceres

Mike: A large part of the spec is the spec for the window object
... It used to be a separate spec
... And we agree it should be a separate spec
... But we didn't have an editor, so we merged it into the HTML5 spec
... There are lots of things in HTML5 that rely on it
... If we had someone to take over ...

Kai Scheppe: ...

Hixie: The window part is a big part of the spec

Mike: what about ..

Hixie: there's too much stuff that relies on it
... The remote event target would be a better choice. That's a reasonably self-contained thing. i'd estimate 5 hours a week for a few months and then 1 hour a week for a year

Cynthia: ... might get more people to volunteer if you have chunks like 5 hours / week

Hixie: The more trivial sections are done
... Stuff like canvas etc. that are 40 hours a week for a year, those are where we really need editors
... Particularly rendering view
... that really is a separate document
... The section that defines legacy attributes

Mike: If you look at the current spec, that section says "to come"

Hixie: It didn't really make sense to define it until about now.

Mike: What we expect from having a discussion with the TAG, one of the tangible things we can talk about
... ... we do have time set aside tomorrow to go through and look at the spec section by section and decide which parts of it are mature and stable with an eye towards what we're ready to write test cases for.
... But also look at what sections we can split out and look for editors for
... Then we could make proposals about parts of the spec that could be taken on by separate editors
... And if we have a concrete list, an assessment about which parts and what level of effort would be needed to maintain that part of the spec

Mike: Then we could maybe get more people interested in parts of the spec

Hixie: I've been privately approaching people. Also someone from Opera recently asked about working on the timer starts
... Some sections are marked "i'm looking for an editor"
... Even then it's taken me a year to find someone for timer
... Timer is a good example of how hard it is to estimate time needed
... between the time when I first started looking for an editor and now, the work tripled
... because the webapps group became looking at a next-generation timing ...
... the problem we had with the window object was that we thought it was very small, 2-line api
... And now it's a third of the spec, and the editor couldn't cope with what became the scope of the work.

Hsivonen: Another thing is that the editor to do a good job needs to have extended exposure to the bug database of a browser engine

Hixie: or preferably more than one

Hsivonen: And there aren't very many people with that kind of exposure.

<DanC_lap> (seems like we could separate editing and authoring/design more.)

Mike: So let's put together a list of what parts of the spec we could split out and how much work we think they'll be
... but let's take a break and come back at 11.


<mjs> DanC_lap, it's not clear to me that would improve matters; few people are qualified to do the authoring/design, while the mechanical edits are at the same time only a small fraction of the work and also opportunities for introducing errors into the design

<Yves> 2

<dino> what is the whole camera thing?

TAG joint meeting

Mike: Next hour and a half discussion with TAG

Mike: A number of issues potential topics

Mike: Modularization of spec is top one
... Other topics are on the list as well.

<anne> celebrities at the table man

<anne> (re dino)

Mike: Since we have time and attention of TAG, we should try to discuss those issue too
... But we start with modularization
... First I'd like to do a quick self-intro

<DanC_lap> Henry Thompson

<dbaron> Henry Thompson, U of Edinburgh

<DanC_lap> Norm Walsh

<dbaron> Norm Walsh, Mark Logic

<Norm> Norman Walsh, Mark Logic Corporation

<dbaron> Tim Berners-Lee

<dbaron> T. V. Raman, Google

<Norm> Ashok Malhotra

<dbaron> Ashok Malhotra, Oracle, Oracle

<dbaron> Noah Mendelssohn, IBM

<ht> ssohn/sohn/

<noah> Noah Mendelsohn, IBM

<dbaron> (and Dan Connolly)

<Norm> Missing from the TAG: Stuart Williams, Dave Orchard, Jonathan Reese

Mike: So first topic is Modularization of the HTML5 spec
... I think best way to start discussion is for members of the TAG want to discuss the problem they see here.

<dino> fantasai, there is a Castorama 10mins walk up the road. It will have duct tape.

Henry: I care a lot about distinguishing about the definition of the language as a formal artifact and the discussion of the behavior of what browsers do with something that purports to be of that language.
... There are two issues, one is whether the spec for the formal language is a separate spec, the other is whether that is defined.
... Is there such an authoring spec?

Hixie: There has always been an intent to have that.
... There are two: one is a stylesheet applied to the spec, the other is a non-normative guide.

Tim: ... the parts that define the valid document and that that define error-recovery are separate and flagged differently.

<anne> (Simon being Simon Pieters)

Tim: ... synchronization with the spec

Hixie: Because the spec has a lot of churn, it would be hard to ...

Tim: Once it's been done once could one assume it would be simple to add tags to the new content?

Hixie: I'd like to emphasize that the spec is not just for browsers, but for all user agents

Noah: I'm coming from the same place that Henry is.
... I'd also like to say that the TAG as a whole has no opinion on this. We're here to learn
... I'm speaking for me personally, not for the TAG.
... One lurking issue is to what extent you believe that over a period of 5-10 years, most user agents will do pretty much what the spec you're writing says.

Hixie: Over a period of 5-10 years either the browsers will change to match the spec, or the spec will change to match the browser.

Noah: Browsers do crazy stuff. They will keep parsing through tag soup, everything but the kitchen sink.
... What about parsers for other applications. Will they do the same quirky stuff?

Hixie: Yes.

Noah: I think it's good to distinguish over time a clean HTML format.
... And try to get people to write that.
... Over time that that becomes the language spec.
... It becomes a contract between authors and consumers
... The spec I would ideally like to see would be an HTML5 spec that isn't mentally for authors, but is the language spec, that says this is a table, this is ..

Hixie: That is the current spec minus the implementation stuff.

Noah: ... good, readable, well-ordered introduction to the language.

Hixie: we want the spec to be understandable and coherent.
... We have various options. We can use a CSS style sheet to hid stuff. If that doesn't work, we could maintain a separate document (which would be a maintenance nightmare), or do some kind of transformation.

Noah: I want it to be good quality

Hixie: of course

Karl: The content.. of HTML5 is well-defined and stricter than that in HTML4.01

<DanC_lap> (stricter... for example?)

Karl: The content inside the spec, is not parse-able, not readable, for most people.
... An automatic translation with CSS or XSLT will not be enough
... So there is a need for a separate document.

Lachlan: Just wrt tools that parse HTML that aren't browsers and wrt clean HTML spec

<marcos> +q

Lachlan: The spec needs to define an algorithm that parses all the crap on the web.

<karl> DanC_lap, for example no align attribute

<DanC_lap> thanks.

<marcos> -q

Lachlan: The parsing section that handles the non-clean stuff is very tightly integrated into the part that handles the clean stuff
... people who want to just parse an HTML table can grab an off-the shelf library

<Hixie> DanC_lap: also much stricter rules for nesting interactive elements, iirc

Lachlan: Having a tool that only accepts clean input does not seem particularly useful.

<pimpbot> Title: HTML 5 And The Hear-Write Web - W3C Q Weblog (at www.w3.org)

Anne: The spec does allow such an implementation.

<Zakim> timbl, you wanted to point out to Noah that in this case he does NOT want to see the proposed contract as in that model it has the full error recovery. In fact authors write who

Tim: When you asked for the contract document, I suspect in this case you don't want it. Because in this case the contract is that the reader will understand the clean stuff as well as the garbage
... You'll get the whole mess.

<MikeSmith> http://www.w3.org/html/wg/html5/#syntax

<pimpbot> Title: HTML 5 (at www.w3.org)

<MikeSmith> http://www.w3.org/html/wg/html5/#conformance-requirements

<pimpbot> Title: HTML 5 (at www.w3.org)

Tim: ... generate a clean document without any quirks.
... But the browser language , which is what this group is working on, ...
... but there's a separate language that is described as conforming HTML5
... that's described in the spec
... but that's not the same as the contract for the browsers

<marcos> +q

Tim: ... ideally everything will be clean and won't have any quirks
... You appeal to the contract between the reader and the writer, and the ...

Noah: What you call conforming HTML5 is what I call HTML5.

Tim: That has nothing to do with what the browser accepts. It's not part of the contact.

Hsivonen: Two issues. One, for whom is this language spec intended. Karl mentioned authors.
... Then there's the suggestion that there be a normative description of the conforming language.
... These are two different things.

<Hixie> http://www.whatwg.org/specs/web-apps/current-work/multipage/#writing-html-documents

<pimpbot> Title: HTML 5 (at www.whatwg.org)

<Hixie> er, http://www.whatwg.org/specs/web-apps/current-work/multipage/syntax.html#writing-html-documents

<pimpbot> Title: HTML 5 (at www.whatwg.org)

Hsivonen: The Writing HTML section is a language-lawyer description of the character streams that will go through the parsing algorithm without hitting any error conditions
... ....
... But if you are a typical author, you don't want to read the stuff in that section.
... Part of it is because it is written very strict. It lists all the unicode characters, e.g.

<marcos> -q

Hsivonen: If we have this normative language spec, we don't get the document Lachy was talking about.
... Using this language spec for parsing, because there are no off-the shelf parsers that don't do error recovery

<Hixie> -q

Hsivonen: If you wanted something that only parsed clean conforming HTML, then you'd have to write one. But to do error recovery you have libraries.

Marcos: ... fixation on having clean markup.
... maybe because you want a tree that is serializable as XHTML
... You get a DOM tree anyway. You can convert to XML anyway.

Tim: Because your authors want to know how to write HTML, what element to put for what.

Marcos: But that's what an authoring guide would be for.
... ...

Noah: Are you saying you don't feel any need for people to fix their non-nested tags?

Marcos: Yes. Who cares?

Tim: There are two problems. In the short term it's confusing, teaching someone they won't get it.
... In the book about how to build a web page you say inside the body there are paragraphs.
... If you're somebody who has written all kinds of other documents and has been asked to produce HTML5 output, then you want something to point this programmer to

Marcos: There needs to be a balance between teaching people the markup and how the DOM is created.
... If I write a paragraph and I write a doctype and the <p> element.
... The <html> and <body> are auto-generated by the parsing algorithm

Tim: ... if they're just going to script it, and not ...
... And it looks really cool .. to outputting documents they want it visible and usable by people .. in 200 years time. they want to produce something that is valid.
... There are a class of users that need to know what exactly .. being able to roll up some documents .. DOM obviously it's much easier spec to read

Marcos: The parsing already algorithm doesn't do what you say, it inserts elements at random


Marcos: well not exactly at random
... The parsing algorithm of HTML5 already defines this.

Marcos tries to give an example

Mike: I'd like to interrupt, we're straying off-topic and other people are on the queue
... we want to focus on the idea of having a separate spec for the language.

Murray Maloney: We were drilling down into why we need such a spec

Tim: It might be useful to have two different documents. One describes what you should send down the wire to get this result in the DOM. The other is what you should send down the wire to make a web page

<karl> is it just a matter of finding an editor for this document, and then discuss later on about the requirements and conformance options of this document

Henry: I'm perfectly happy for as many ppl out there as want to never to use a strict parser as long as they don't mind I want to use a strict parser.

<karl> I'm willing to write this document after November

Henry: I think the case today that the students in CS are told they must submit HTML that passes the w3c validator. That's part of the education process.
... There's a substantial history of curriculum development that led us to want to do that.

<Hixie> MikeSmith, can i jump in here?

Henry: As long as you're fine with us doing that, then I'm happy for you to use whatever parser you want to use.
... If the spec is going to discourage that, then I think we have a problem here.

Hixie: The parser defined in HTML5 today allows any implementation to abort on the first error.

Henry: That's good enough for me.

<Zakim> DanC_lap, you wanted to think out loud about the character encoding detection algorithm

DanC: The spec modularization problems that I run into are things like the character detection algorithm, which is e.g. used in some other webapp spec

Tim: Sounds like a separate spec to me.

DanC: I think it's been copied into the webapps spec?

Marcos: we reference it?

DanC: Is the scheduling ok so that HTML5 will be done before you need to advance through REC track?

Hixie: That section is very specific to determining character encoding for HTML
... It's not like the URI spec which is independent
... this is literally part of HTML

<MikeSmith> http://www.w3.org/html/wg/html5/#determining-the-character-encoding

<pimpbot> Title: HTML 5 (at www.w3.org)

Hixie: It makes sense to refer to HTML to talk about HTML

DanC: Suppose they want to finish their spec before HTML5?

Tim: 2 groups use the same algorithm, then typical way of doing this is to rip out that part and put it in a separate module

<MikeSmith> fantasai, I'll try to summarize after Tim

Tim: Because that piece is small, it gets reviewed by a bunch of people who wouldn't look at HTML5
... i18n will pore over it
... and it'll go to REC fast and become a useful tool for the community

MikeSmith: We already have a lot of other specs besides HTML5 that already normatively reference or will need to normatively reference parts of HTML5

<anne> I note that i18n and people from Unicode have in fact reviewed those parts of HTML5

MikeSmith: HTML5 will block them

Hixie notes that the character detection depends on other parts of HTML

<anne> Character encoding detection was e.g. discussed last year with the i18n WG

MikeSmith: I don't want to split hairs on exact status details, but if our spec is not mature enough and people need to depend on parts of it, it would be easier to facilitate those references if those pieces of the spec were separate specifications and were moved along on a faster track.

Hixie: it depends on the rest of the language.
... So the character encoding determination section does it looks for a <meta> element with a charset attribute
... While doing that it tries to skip comments and other syntactic things.

Anne: there's hookbacks from the parser

Hixie: once you.. whole thing about scripts
... the idea that we have a charset attribute in HTML5. That's new. We didn't have that on HTML4.
... we don't have agreement on that. It might change.
... until that gets accepted, then we can't move it forward.

Anne: Those other people are reviewing the draft. We had comments from unicode and i18n last year.

MikeSmith: But we aren't communicating the changes.
... They aren't aware of our changes to that.

Anne: I don't entirely agree. When we publish a new working draft, we list the changes.

Mike: I've been trying to do that, but the number of changes is huge
... It's good that we are improving the spec, but it's a large number of changes. But expecting that the ppl outside the group will be able to understand all those changes ...

Hixie: experience with CSSWG is that splitting the spec doesn't work.

Hsivonen: I'd like to .. marcos about .disagree with parsing algorithm
... I don't think author should have to figure out what kind of crazy stuff can be done.
... I think we should say if you do this, it will work. Don't need to explain all the other crazy stuff that could be done that would also work
... If you want a strict parser the right way tot do it is to take an existing one and register an error handler that the only thing it does is throw an exception
... that gives you a strict parser

<DanC_lap> (bummer... the pointer to encoding tests from http://lists.w3.org/Archives/Public/public-html/2007JanMar/0054.html has gone 404...)

<pimpbot> Title: Re: brainstorming: test cases, issues, goals, etc. from James Graham on 2007-03-14 (public-html@w3.org from January to March 2007) (at lists.w3.org)

Hsivonen: I think the discussion about stability and referencing, might point out a problem in the Process.
... It's not only HTML5 that has this problem.

<DanC_lap> CURIEs

<karl> http://www.w3.org/QA/2008/10/normative-references-conformance

<pimpbot> Title: Normative References to Moving Targets are Dangerous - W3C Q Weblog (at www.w3.org)

Hsivonen: E.g. RDFa copy-pasted CURIEs
... So there's an instance of this problem in a case where the specs are much more closer to each other.
... Then SVG 1.2 Tiny can't reference CSS2.1 and instead are referencing CSS2.0 even though every implementor knows nobody should be looking at CSS2.0 and should look at CSS2.1 instead
... Does it help anyone for the process to force these things?

<smedero> DanC_lap: I think you're looking for this: http://html5lib.googlecode.com/svn/trunk/testdata/encoding/

<pimpbot> Title: Revision 1229: /trunk/testdata/encoding (at html5lib.googlecode.com)

Hsivonen: In HTML5, if it's going to change in HTML5, we want webapps to match that.
... The point is for the two to match, so you use the same code path

<shepazu> hsivonen, the parts of the CSS spec SVG references hasn't changed between 2.0 and 2.1

<smedero> DanC_lap, (well and then the scripts that deal with that... which have moved here: http://html5lib.googlecode.com/svn/trunk/python/tests/)

<pimpbot> Title: Revision 1229: /trunk/python/tests (at html5lib.googlecode.com)

Hsivonen: Even if webapps goes to REC, if the HTML5 encoding detection changes then webapps wants to match that.

<timeless> /ignore #tpchat join

<DanC_lap> (smedero, I only see 4 tests there. is that how many should be there?)

Hsivonen: If we're working around a problem in the process that other WGs are facing to and the solutions aren't really helping with implementing the specs, perhaps instead of working around the problem we should solve the problem

Tim: The process is a tool, it's ours to use. The fact that there's a two-step difference in level between something that you can reference and something in your own level is unusual in the standards world and
... ISO and IETF you could only reference a standard
... Or maybe something at the same level

<karl> That it is not the W3C Process which gives this requirement

Tim: Being able to reference something less mature is regarded as a bug.

<karl> This is the transition document

Tim: There's a good reason. If you write code for a technology and it meets the standard
... And then the less mature spec is changed, then your software doesn't match the spec any more and stuff breaks.

<karl> Evidence that dependencies with other groups met (or not)

<karl> # Does this specification have any normative references to W3C specifications that are not yet Proposed Recommendations? Note: In general, documents do not advance to Recommendation with normative references to W3C specifications that are not yet Recommendations.

<karl> # Is there evidence that additional dependencies related to implementation have been satisfied?

<karl> -- http://www.w3.org/2005/08/online_xslt/xslt?xmlfile=http://www.w3.org/2005/08/01-transitions.html&xslfile=http://www.w3.org/2005/08/transitions.xsl&docstatus=pr-tr

<pimpbot> Title: How to Organize a Recommendation Track Transition (at www.w3.org)

<smedero> DanC_lap: comparing against the 1.0 branch, yes.

Tim: We could change the process, but that will only help us create broken software.
... With something like this, where we have another solution -- which is to pull out this bit of technology and make it separate spec
... It's got the ability to be stabilized well in advance, then you can have an appropriate ordering between your specs
... And it'll all work. You don't need a cycle.
... It's when you have a cycle that you need to have this slack between the two specs.

Noah: Picking up in part on what Tim said.

<Zakim> noah, you wanted to wonder if we're just disagreeing on what should be normative

<DanC_lap> (interesting point, karl; ht, I don't find any constraints on dependencies in the process document. As I recall from discussions with Ian, we rely on reviewers to complain about pointers to stuff that's not sufficiently mature.)

Tim: about language tutorials

Noah: I wanted to add up all the parts that everybody agreed on and not
... part I heard agreement on is that the spec should go on the say you're writing it. Maybe you should re-title it, but otherwise no issue
... I think I heard everybody agree that there should be documents that help novices learn to write HTML

<pimpbot> planet: Chris Wilson on Internet Explorer 8 and the W3C HTML Working Group <11http://standardssuck.org/chris>

Noah: I think I heard agreement that it should discourage improperly nested tags

<Al> .. if there could be reference to a function call, we could freeze the interface to that function call and there is not breakage if the body of the function is definitized later

Noah: The part I didn't hear agreement on is .. defining a clean language with no errors

Hixie: That already exists. I sent a link to it in IRC.

<ht> DanC, the (in)famous '2 steps back' rule comes from the XML Plenary in San Jose in 1998 (?). Consistent with what you said, it's a guideline -- other things being equal, the presumption is that 2 or less is OK, more than that is not, but it's only a presumption

Noah: Question is should it be a normative document

Hixie: It is. It's section 8.?

Noah: It seems to me there's a question whether you advertise to the community conforming HTML5 that is a big deal
... The way XML is a big deal
... If you decide you don't want to do that the world becomes very simple. You're just writing non-normative guides. Can write lots of them. They may not be very consistent.
... If you do write this section then the question is how can you write such that it is understandable for mere mortals

<Al> ht, what Chaals learned on researching this recently is that any down-level reference will bring close scrutiny and demand for "three good reasons". I don't presently think a two-level presumption is safe.

Noah: My personal preference is that you do create such a spec. I don't have a reason, just abstract intuition
... I'm curious if people ...
... The optional things are,
... Do you write one or more non-normative informative guides to help authors write stuff.
... I heard that in general they should encourage the creation of clean content
... The more controversial option is should you write a normative and precise document that specifies only the clean language and its semantics.

<pimpbot> Title: HTML 5 (at www.whatwg.org)

<Hixie> Philip: ^ see timbl's comment

Noah: I'm not using the term authoring guide, because it's not how I think of it, but it seems that's what you're thinking of

Mike: One issue is the authoring language spec for HTML5

<Hixie> Philip: if you could set that up I can re-gen the spec straight away

<Philip> Hixie, what would the page title be?

Mike: The other issue is the concern Dan brought up , of other parts of the spec that don't relate to authoring conformance but refer to browser implementation details

<Hixie> Philip, first h2, i guess

Mike: for which there is rationale to have separate specs

<anne> Philip, section title of 8.1 &mdash; HTML5

Mike: we do want to talk about both those things
... I want to go through the queue.

<Zakim> ht, you wanted to point out that there are two solutions to Henri's point within the process

Henry passes

<Philip> Hixie, some pages don't start on an <h2>, but I suppose I could just use the first heading

<Zakim> karl, you wanted to mention that it is not part of W3C process

<Zakim> Al, you wanted to say that Henri's point about managing dependencies of concurrently progressing modules shows that Henry's emphasis on declarative spec introduces problems.

<Philip> Hixie, (though it'll be a bit misleading on pages that have multiple significant sections)

Karl: The discussion about normative references was not about process document requirements, but transition document

<zcorpan> Philip: "8 The HTML syntax - HTML 5"

Al: I'd like to +1 what Tim was saying earlier, that there's the contract involves two levels of strictness
... The contract includes the ideas that the consumer has the support of a browser which processes strings of several varieties

<Hixie> Philip: yeah, but it'll be better than nothing

Al: While the author gets instructed in how to be strict in what they emit

<gsnedders> Philip: Why can't we just have one major section per doc?

Al: This is what makes the Web interoperate today. Is that we have both statements.
... This is what the HTML5 spec tends to do. I want to say that's valuable.

<anne> Philip, alternatively you give us an API so we can make up a title per page and such :)

Al: That's how you make large systems interoperate. You have some space between these two.

Hixie: I'd like to encourage people that want a document that defines a "clean language" to read section 8.1
... and see if that's what they mean.

<MikeSmith> http://www.w3.org/html/wg/html5/#semantics

<pimpbot> Title: HTML 5 (at www.w3.org)

Hixie: I'm not sure that document is useful to authors. Because it wouldn't be something they'd understand.
... And that's where a non-normative authoring guide comes in.
... I don't think that splitting sections out of the spec to keep them more stable will work.

<MikeSmith> http://www.w3.org/html/wg/html5/#the-a-element

<pimpbot> Title: HTML 5 (at www.w3.org)

<noah> What I have in mind is a document would be a document that would include the syntax, as well as the normative definitions of what a table is, a paragraph, etc. Ideally, I would then NOT repeat those semantics in the "larger" user agent spec.; I would have the user agent spec refer to the language spec for that.

Hixie: The problem isn't that HTML5 isn't stable,the problem is that the part that changes is the part that we want to split out
... I don't have a solution to that.

Mike: I'd like to talk about specifics.
... We seem to have consensus that we should have an authoring spec.
... Several parts of the spec talk about authoring conformance criteria

<marcos> +q

<DanC_lap> "8.1 Writing HTML documents"

Mike: There's 8.1
... But there are also parts on semantics

<Zakim> ht, you wanted to suggest we look at the URI issue

Mike: We should go to that part of the spec and see if it has what we think needs to be there

Henry: I'd like to back you up and move on to another issue.. modularization
... IA different spin on it. It really is a matter of "spheres of influence"
... That's the URI/URL parsing section
... It did feel to the TAG at least at first blush that this looked like
... A misjudgement wrt serving the web community
... That people need to have a consistent picture of identifiers for web resources
... It's not up to the HTMLWG to decide what that string looks like
... At the very least the IETF has a stake in this. They own the relevant specs
... I'd like to see if there's a willingness to look at refactoring that discussion at least.

Lachlan: Wrt splitting the spec, I'm a bit curious about who exactly we're targetting this other spec for.
... there are a whole range of .. that use HTML5.
... Each of those need different overlapping sections of the spec.

<anne> XMLHttpRequest currently refers to the HTML5 URL concept...

<Lachy> Authors, markup generators, authoring tools, validators, generic consumer tools, browsers.

<Lachy> Each needs a different set of overlapping sections of the spec.

<Lachy> Authors: semantics, conforming syntax

<Lachy> authoring tools: semantics, conforming syntax, parsing, sometimes rendering

<Lachy> validators: parsing

<Lachy> consumer tools: parsing, DOM/tree

<Lachy> search engines: semantics, parsing

<Lachy> browsers: semantics, parsing, rendering

Lachlan summarizes his notes above

<hsivonen> validators need a lot more than parsing

Lachlan: The problem is who exactly are we trying to target with this split spec, given that there are so many overlapping needs?

<marcos> -q

Marcos: Lachlan said my point

<CynthiaShelly> @MikeSmith: are the "semantics sections" sections 3 and 4?

<pimpbot> cshelly: Huh?

Hsivonen: On URIs what the spec does it takes the IRI RFC and defines the delta of what you need on top of the RFC.

?: It states what rules you need to change in the RFC to parse

<karl> http://www.w3.org/TR/qaframe-spec/#implement-principle

<pimpbot> Title: QA Framework: Specification Guidelines (at www.w3.org)

Henry: Why do you need a delta at all JR:

Hsivonen: Suppose I'm writing a browser. There's existing content out there that was written before the IRI spec
... If you have a form and you input characters, and the you submit that form using GET.

<MikeSmith> cshelly, basically just section 3

<Philip> http://www.whatwg.org/specs/web-apps/current-work/multipage/syntax.html

<pimpbot> Title: 8 The HTML syntax HTML 5 (at www.whatwg.org)

<MikeSmith> cshelly, plus the first part of section 8

Hsivonen: You encode the unicode characters in the form fields to bytes using the character encoding that the document itself was labelled as when it was parsed.

<Philip> MikeSmith, you still need to fix pimpbot entity handling :-p

<Philip> s//'s/

Hsivonen: And this is required for backwards compatibility, otherwise servers will receive form submission that they didn't expect and they will break.

<Hixie> woohoo, nice work Philip

Hsivonen: there's another eq that flows from this one. If you have on an html page a URI that has a query string

<karl> http://www.w3.org/TR/spec-variability/#spec-cat-cop

<Hixie> timbl_, multipage spec version is updated to have better <title>s now

hsivonen: You are linking to a form submission such as the one I describe

<pimpbot> Title: Variability in Specifications (at www.w3.org)

<MikeSmith> Philip, patches welcome

<timbl_> If the form submission has non-ascii then you can't encode it a la IRI encoding.

Hsivonen: The kind of string path's in href, if it has non-ascii in it to get an ascii-only URI you need to use the encoding of the document to encode those characters into bytes and then do the % encoding on those bytes and then give the result to the HTTP library

<timbl_> You have to use the encoding which the document itself was parsed with.

<anne> timbl_, it would be URI percent encoded afaict

Hsivonen: Sites are depending on this behavior. So browsers are.

<gsnedders> http://www.whatwg.org/specs/web-apps/current-work/multipage/infrastructure.html#resolve-a-url

<pimpbot> Title: 2 Common infrastructure HTML 5 (at www.whatwg.org)

Julian: : We all know that and we are in total agreement that the spec needs to do that.

<gsnedders> That defines what hsivonen is summarizing

Hsivonen: So we've established that browsers need to do that.
... I'm writing something else. The image .. of validator.nu
... It's not a browser, but it deals with content browser deal with
... I can't implement the IRI and then find out later that it won't work
... The RFC doesn't give me a spec for that
... The .. library that implements 6 different profiles for IRIs,but doesn't implement the most common profile -- href links in HTML

<MikeSmith> ?

Hsivonen: if IETF isn't providing then, then I want to have a spec that I can write to and not have to reverse-engineer anything.
... I want to use the reverse-engineering that Hixie did
... If that's not in the IETF, then it should be somewhere. HTML5

Mike: ... whether we should do that in a normative W3C REC
... we need to come back to what specifically we need to get to

<Julian_Reschke> all the stuff that Henri said is correct, except the conclusion that it needs to be done the was it is done right now

<ht> Henri, Stipulate that the RFC is broken, why not fix the RFC, instead of starting a turf war?

TV: The meta issue is that there's an RFC for IRIs and browsers violate that IRI. Meta-issue is do we need to codify that violation fo the IRI spec?

Mike: So maybe we need to clarify that this is a big public issue that is open

<hsivonen> ht, the IETF wasn't cooperative when doing that was suggested

TV: The meta-issue here isn't about bits and bytes, but about whether this WG should be codifying existing violations of existing RFCs.

Tim: ... go into another RFC.

<ht> Henri, did you raise it through the W3C-IETF liaison call?

Murray: modularization, goes to IETF

LarryMasinter: It's not clear to me that forms submission properly with in the domain of HTML to define how HTML define form submission
... the IRI spec doesn't say how HTML does form submission
... ... URI specification...

Larry: It may be that if you see a URI in a form and you have query string that you take that URI and then you transform that URI and then you do the submit on that URI

Larry: It's not changing the IRI spec. It's doing some weird processing before.
... It might solve the issue to rewrite the spec so to define this behavior as pre-processing the IRI string.
... ... and maybe submitting this back to IETF would also be a good idea.
... back to authoring spec.
... I think rather than thinking about browsers and authors, since most content today is created by other software.
... You should use terms producers and consumers

<ht> HST likes Larry's suggestion, because it's consistent with the fact that IRIs are generic, but the query-string format we're talking about is 'http:' specific, I believe

Larry: I think that will help with the discussion.

<DanC_lap> (I disagree. authors are humans that need nice successive-elaboration documents.)

<karl> what Larry just said is part of the list of class of products - http://www.w3.org/TR/spec-variability/#spec-cat-cop

<pimpbot> Title: Variability in Specifications (at www.w3.org)

Hixie: Before IRI section was drafted, I joined the relevant IETF group and asked if I could create a spec there
... they told me they weren't interested in doing this.
... So then I put this section into HTML5.

<anne> interested parties weren't interested, nice

Hixie: If anyone wants to edit that section, I'm happy to pull it out into a separate spec.

Tim: ...
... My understanding was that if you want what henri said.
... A string with non-ascii characters and then encodes it ...

<DanC_lap> (the algorithm takes not just a URI reference, but also the encoding fo of the document it came from.)

Tim: That URI string will have % in it, but if you interpret it as IRI ..
... So it would be inappropriate to hand it over to IRI

Anne: Isn't IRI a superset of URI?

Henry: Suppose your doc is encoded in 1252
... Larry's suggestion is that the HTML spec says take that string, and re-encode it in unicode and the it is a URI and you can apply the escaping algorithm to it

Hsivonen: You get the wrong URI if you do that.

Mike talks about lunch and AC meetings

Mike: we have a time constraint

<DanC_lap> (Julian, we were close? close to what? is it something you could write on a line or two?)

Mike: Let's move beyond the specifics and talk about how we can move forward , what can we do after today to continue the discussion.
... If everybody can agree to hang around for an extra 15 min that gives us time to get through the queue and try to talk about moving forward

<Zakim> DanC_lap, you wanted to note that after review of the URL stuff, I found it acceptable, but I haven't "sold" others. the IETF has right of review, and I think we have an

<Julian_Reschke> for the record: Larry's proposal does work; we just need to use ASCII to encode the characters that we don't want the IRI processing to map the wrong way

<MikeSmith> q/

DanC: So the URL stuff, I figured out what it was saying and harumphed over it a lot and made my peace with it
... But as go around to other people, the trick is.. since this deals with URI and IRI stuff
... It's good to get consensus with them
... They say it would be ok if you rewrite it
... With the technical design, I see no way out of this local minimum
... People that are interested to take a whack at rewriting

<Zakim> noah, you wanted to ask whether the issue is the use of the term URI or IRI for something that doesn't conform to the RFCs

Noah: This may actually relate to paper consensus / durable consensus
... I may be misinformed, but I believe an issue is in the draft spec it refers to things as URIs that are not URIs, that are the input to this process or the output
... If there are such things, and it contributes to stable discussion, to change that. It would be nice to have convenient terms

<Al> content negotiation for href; can we do server sniffing as a way to serve the

Noah: but whatever. Call it an HTML URI. Write your spec that way

<Al> .. serve the 'right' URI transcription as per IRI RFC?

Noah: Then in the informal guides you can call it whatever you want, URIs, URLs
... I think being strict with the terminology will help

Hixie: the term that's used in the spec is URL, but I agree it's a conflict. The reason we used that
... even when we used URI or IRI, a lot of people said what the heck's a URI, or what's a IRI
... It was a compromise I came to by balancing the people who would be confused about this term vs that term
... The actual section, if you give it a valid URI or IRI it gives you a valid URI or IRI.
... the only thing it defines is how you handle invalid URIs and IRIs

Hsivonen: What he said isn't exactly correct. It's correct if your document is encoded in UTF-8 or doesn't have query strings
... I don't think it's a good idea to give an ugly name.

Noah: no, come up with a nice name

Tim: I'm happy saying at the top of the document we redefine URL as used in this specification.

Philippe: We do this in a lot of specs that use URI where they really reference IRI.

Mike: So at a high level we're all here because we want to produce a W3C REC for HTML5

<Julian_Reschke> problem: there's also an important distinction between "URL" and "valid URL".

argument about whether TAG can prevent this, or whether it's only the Director

Mike: Anyway, W3C can block this if there are parts of the spec that conflict with Web Architecture and they're not ok with that.

<anne> hsivonen, that's not exactly correct, UTF-16 is also fine

Mike: There are parts of the spec that it wouldn't be possible to move forward unless we work through these issues.

<DanC_lap> (this seems like a negative way of saying: a goal of W3C is to get consensus on specs)

<hsivonen> anne, true

Mike: We could wait until CR, or we could try to resolve these issues now.

<noah> I don't think anyone prefers to defer the process of working through issues.

Mike: I've gotten a list of things Roy Fielding plans to formally object to
... It is to the benefit of all of us to try to get resolution sooner rather than later.

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

Tim: Whether Roy or others are on the TAG has got nothing to do with how much we should listen to their opinions

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

<smedero> ahh

<smedero> olivier: thanks

Tim: .. nor should you look at this as what should we do to do the least procedurally to get this to rec. You want to look at what's going to make this the best spec.

Larry: Let's move away from thinking about personalities
... And consider whether there are communities whose needs are not met by this specification.
... maybe the HTML5 spec doesn't meet the need of the community that builds web servers
... I'm more representative of the parts of my company that make authoring tools
... ...

<DanC_lap> (I think "needs" is a strong terms; there are communities whose _wants_ are not satisfied; some of them get dissuaded by discussion tactics before they finish learning whether their actual needs are met.)

<Hixie> s/more/a poor/, i think

Larry: I look at it from the pov that there are some communities that have a lot less time to spend on this, but have requirements that might not be addressed as well

<Hixie> but i could be wrong

Mike apologizes for wording

<gsnedders> Hixie: You're right

Mike: .. I'm really glad Roy is paying attention
... This work has been going on for 4 years
... For a lot of that time people weren't paying attention
... Now they are

<DanC_lap> TimBL notes development of HTML goes back 18 years and 6 months

Mike: I think it's good that we have more people paying attention to this work now and we ant to make it possible for them to facilitate the discussion going forward.
... So again, where are we going to continue this discussion? We have public-html
... but only members can post to it

<DanC_lap> (only members of the wg can post to public-html? I don't think that's so

<DanC_lap> )

TV proposes copying public-html and www-tag

<gsnedders> DanC_lap: anyone can, but it's covered by the patent policy, so it's kinda non-obvious

<Hixie> i just wanted to mention that i work with people from the apache foundation, and that google also writes web servers, and i try my best to address their needs and desires

<Hixie> but no need to say that out loud i guess

Tim: modularization or URI stuff?

Henry: There is a w3c IETF liaison call
... Query strings in their particular formulation as they're used in HTML ...

Danc: Henry please, this is a technical question, right?

Mike: I would like we don't leave today with hanging on where to continue discussion?
... How can we move forward with discussion with TAG with these issues?
... Specific issues are authoring spec, and modularization -- splitting pieces out, and whether we should have normative definitions for some stuff like URI

<anne> email message to uri@w3.org from Hixie: http://lists.w3.org/Archives/Public/uri/2008Jun/0002.html

<pimpbot> Title: Error handling in URIs from Ian Hickson on 2008-06-24 (uri@w3.org from June 2008) (at lists.w3.org)

<ht> HST was going to ask if there was agreement that the core of Henri Sivonen's summary only applies to http: URIs

MM: I would've thought it'd be up to the WG to discuss what they've just heard from TAG
... And sort out what they think they've heard, and what they think they should do about it
... and then tell the tag

<DanC_lap> fantasai, meet Murray Maloney. Murray, fantasai.

MM: and then the TAG can say that's great

TV wants to CC www-tag

MM: Do you want to follow the whole discussion or just the conclusion?

Mike: So action is on me to bring this back to HTMLWG and to communicate back to TAG what we want to do about these issues.

MM: What are the other three issues we wanted to talk about?

<gsnedders> What email?

TV: They were listed in the email, we should talk on email

Mike lists the issues, someone please paste the email

Mike: I'll just repeat what I just said, action item is on Chris and me to bring back to the group
... come to resolution on what we plan to do and communicate back to the tag

Tim: I'm wondering from TAG pov whether we should also ...

<gsnedders> The email anne mentioned earlier that's member only and therefore people like me can't actually see it?

Tim: from discussion last couple days, one thing that comes up often when people see HTML5 spec for first time

<DanC_lap> ACTION: Mike lead HTML WG to response to TAG discussion and report back to TAG [recorded in http://www.w3.org/2008/10/23-html-wg-minutes.html#action01]

<trackbot> Created ACTION-77 - Lead HTML WG to response to TAG discussion and report back to TAG [on Michael(tm) Smith - due 2008-10-30].

Tim: they look at the spec and they say that's not what I mean by a spec.
... I wonder whether the TAG could write something to point out there's two ways of doing this
... E.g. TAG's done work on versioning.
... ....
... If the HTML5 folks could say what we're making is one of these and point to the TAG's writeup
... Would working at the meta-level be helpful. Maybe useful to IETF and understanding HTML5 as opposed to language specs as traditionally written

Mike: Maye a summary from the TAG perspective on what we discussed today, that would be very helpful

TV: It looks like Tim signed up to write such a thing. :)

Tim: I'll nominally take an action item.

<scribe> ACTION: TimBL write up a description of the kind of spec HTMLWG is writing for TAG [recorded in http://www.w3.org/2008/10/23-html-wg-minutes.html#action02]

<trackbot> Sorry, couldn't find user - TimBL



bye everyone

RRSAgent: make logs public

RRSAgent: make minutes

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

<gsnedders> hiho

<timbl> ericp?

<karl> ac

<timbl> phenny, tell ericP Does your query pipeline system support Sparql DESCRIBE ?

<gsnedders> Do we have a scribe?

<Lachy> I can

<Lachy> try

<timeless> ScribeNick: Lachy

Forms in HTML5

Steven: I can find someone to say something about Forms

MikeSmith, The most significant change is the integration of web forms 2 into current draft

<Steeeven> Forms=architectural consistency part

<Steeeven> there is a separate task force for that

<oedipus> no longer - expired in July 2008

Hixie: I'm still responding to feedback. Around start of sept, I started merging HTML4 forms and WF2 changes into HTML5

going through old feedback since 2004

<myakura> http://www.whatwg.org/specs/web-apps/current-work/multipage/forms.html

<pimpbot> Title: 4.10 Forms HTML 5 (at www.whatwg.org)

<oedipus> http://esw.w3.org/topic/PF/XTech/HTML5/Forms

<pimpbot> Title: PF/XTech/HTML5/Forms - ESW Wiki (at esw.w3.org)

Hixie: Not everything made it through, e.g. repetition model

MikeSmith: so that's kind of the last missing semantics that was missing?

Hixie: Aria is the other

MikeSmith: We need to prepare a summary of what's changed

<gsnedders> Lachy: colons and ...

MikeSmith: Doesn't need to be as detailed as the last time

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

MikeSmith: We need something a little less detailed.
... anne put together a bulleted list of the changes, but we need something a bit between that and the detailed change list
... The forms stuff was integrated based on the WG survey
... We had forms TF for a year. The TF did not do anything

<oedipus> meeting: HTML face2face TPAC 2008 Day 1

Anybody else have any questions or concerns?

<karl> Meeting: HTML WG - TPAC - October 2008

<DanC_lap> (anybody got a pointer to that msg?)

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

Nick_Van_den_Bleeken: Someone sent an email with some question regarding the integration of forms into html5 was sent to the forms tf list, but no-one responded

MikeSmith: Please find a pointer to that message
... We also asked for feedback on the WF2 proposal from the Forms WG, but they didn't provide any

<anne> might be http://lists.w3.org/Archives/Public/public-forms-tf/2008Jun/0000.html

<pimpbot> Title: Re: Status update from Keith Wells on 2008-06-02 (public-forms-tf@w3.org from June 2008) (at lists.w3.org)

Charlie Wiecha: [talking about Ubiquity XForms on Google code]

<oedipus> my comments to the HTML WG on the forms survey & forms in general: http://esw.w3.org/topic/GregoryRosmaita/FormsFeedback2008-07

<anne> http://code.google.com/p/ubiquity-xforms/

<pimpbot> Title: GregoryRosmaita/FormsFeedback2008-07 - ESW Wiki (at esw.w3.org)

<pimpbot> Title: ubiquity-xforms - Google Code (at code.google.com)

Charlie: The main thing is to take some of the attributes on the data model and project those up to the controls.

Kai Scheppe: Is the entire WF2 spec taken into HTML5?

Hixie: About half made it in. Some sections dropped: Repetition model, forms pre-seeding, XML serialisation mode,
... a few other minor issues as well
... mostly in response to implementer feedback

<Hixie> (or implementor apathy)

<anne> http://www.whatwg.org/specs/web-apps/current-work/notes has an informative list at the end of features from WF2 not included in the spec

MikeSmith: we have public-html-comments@w3.org
... anyone can post to that
... public-html@w3.org is intended for the WG, but anyone can still post to it

<DanC_lap> comments archive: http://lists.w3.org/Archives/Public/public-html-comments/

<pimpbot> Title: public-html-comments@w3.org Mail Archives (at lists.w3.org)

MikeSmith: The other thing is that the group is open to anyone
... Anyone can be an Invited Expert

<CharlieWiecha> Link to Forms-A internal working document on attribute-oriented forms notation, i.e. non-MVC authoring: http://www.w3.org/MarkUp/Forms/specs/XForms1.2/modules/streamlined/index-all.html

MikeSmith: Need to agree to the patent policy

<pimpbot> Title: Forms-A: Streamlined Expression of Data-Rich Web Applications (at www.w3.org)

<DanC_lap> (for reference, this is a patent policy FAQ: http://www.w3.org/2003/12/22-pp-faq.html#lists Can non-participants subscribe to a mailing list of a Working Group under the Patent Policy? )

<pimpbot> Title: Frequently Asked Questions (FAQ) about the W3C Patent Policy (at www.w3.org)

<PIon> + MathML would be happy to know what happened in this morning's discussion?

MikeSmith: Any other issues?

<nick> Last e-mail from Keith http://lists.w3.org/Archives/Public/public-forms-tf/2008Jun/0000.html

<pimpbot> Title: Re: Status update from Keith Wells on 2008-06-02 (public-forms-tf@w3.org from June 2008) (at lists.w3.org)

Neil Soiffer: We had a discussion about questions we had

<nick> e-mail I sent on 22th of May http://lists.w3.org/Archives/Public/public-forms-tf/2008May/0003.html

<pimpbot> Title: Re: Status update from Nick_Van_den_Bleeken on 2008-05-22 (public-forms-tf@w3.org from May 2008) (at lists.w3.org)

MathML in text/html

Neil: The first issue is namespaces
... If you have a namespace attribute, what happens to it in the DOM?

Hixie: This may change, but the current proposal is that if you have MathML in the document, the xmlns attribute will be in the DOM, but must have MathML namespace
... the DOM nodes will be in MathML namespace

<CharlieWiecha> Link to Forms-A document with controls in the XForms namespace: http://ubiquity-xforms.googlecode.com/svn/branches/webforms-a/_samples/Loan/xforms-webforms-a-loan.html

Neil: Often people put in private attributes

<CharlieWiecha> The above form should load (FF3 at least) directly from SVN

Hixie: The non-mathml stuff will appear in no namespace, non-mathml xmlns declarations will be ignored
... There are constraints with what we can do with namespaces.

<CharlieWiecha> Form in HTML namespace (though this one seems to have trouble loading, pls use as syntax example for now): http://ubiquity-xforms.googlecode.com/svn/branches/webforms-a/_samples/Loan/webforms-a-loan.html

<DanC_lap> <math> <special:stuff /> </math> <-- special:stuff goes in the mathml namespace, if I'm following

Hixie: main aim is to support MathML, not other proprietary things

<gsnedders> DanC_lap: correct

Neil: There is a problem related to Open Math

<gsnedders> DanC_lap: with a localName of special:stuff

hsivonen: open math is not supported

<DanC_lap> <math> <mx special:attr="abc"> </math> <-- special:attr goes in no namespace, if I'm following

<gsnedders> DanC_lap: Again, correct

Neil: The issue is not the browser support, but the ability to copy-paste and maintain the semantics

hsivonen: non-mathml content will end up in no namespace

<anne> http://www.whatwg.org/specs/web-apps/current-work/multipage/tree-construction.html#parsing-main-inforeign has details

<pimpbot> Title: 8.2.5 Tree construction HTML 5 (at www.whatwg.org)

Hixie: Because of the constraints, if we wanted to support open math, we would have to explicitly include that in the spec

<DanC_lap> hmm... that last example with presentation/semantics/content ... I'd like to see an example. I don't think I can cook one up

Hixie: The line was drawn at HTML, MathML and SVG

Neil: People want to style the text they have inside <mtext>

<anne> hsivonen, it seems "semantics" is not supported

Neil: What are the rules inside of <mtext> and you encounter an HTML element, when does it revert to MathML ns?

<anne> hsivonen, that is, what you're saying about semantics is not what happens

<anne> hsivonen, it's true for <mtext> though

Hixie: <mglyph> will be in the MathML ns, other things will be assumed to be HTML
... [other technical details]

<hsivonen> anne, sorry. it's <semantics-xml>

Neil: there is legacy MathML designed to work in IE with plugin
... I sent an email about a year ago, what's the status of dealing with this legacy?

<anne> hsivonen, doesn't change anything unless I'm missing something in the parsing algorithm

Hixie: if there are no prefixes, it should just work

Neil: IE requires the prefix

hsivonen: That doesn't seem to match my testing with MathPlayer

Neil: The way IE binds it is when the namespace is seen, invoke the activex control

<oedipus> http://www.w3.org/Math/DOM/

<pimpbot> Title: The MathML DOM Bindings (at www.w3.org)

<anne> timeless, http://www.dessci.com/getmp

<pimpbot> Title: MathPlayer: Download and Installation (at www.dessci.com)

Neil: [saying the prefix is needed for IE]
... There are real websites that have used prefixes

Hixie: We could explicitly support it, but it'd be weird

Neil: We need to warn content sites that prefixes will go away in the future

MikeSmith: We have time for SVG WG tomorrow
... you're welcome to be there for that

Neil: Everything else looks great

GRDDL, RDFa, @rel value registry, extensibility in text/html

HarryHalpin: GRDDL WG
... People are happy with the ability to get the browsers to implement a singular DOM across real world HTML
... From the perspective of GRDDL, the issues are related
... The fundamental problem is that when the WGs were chartered, HTML5 wasn't on our radar
... We designed with compat with XHTML

<myakura> http://www.w3.org/TR/grddl

<pimpbot> Title: Gleaning Resource Descriptions from Dialects of Languages (GRDDL) (at www.w3.org)

<pimpbot> Title: Google (at google.com)

<oedipus> http://www.w3.org/2001/sw/grddl-wg/

Harry: We want to figure out how our technology remain compatible with HTML?

<pimpbot> Title: W3C GRDDL Working Group (at www.w3.org)

<oedipus> http://www.w3.org/TR/grddl-primer/

<pimpbot> Title: GRDDL Primer (at www.w3.org)

Harry: Seeing mild uptake with Drupal

<oedipus> http://www.w3.org/TR/grddl-tests/

<pimpbot> Title: GRDDL Test Cases (at www.w3.org)

Harry: The first thing to talk about is <link rel="">

<oedipus> http://www.w3.org/2007/08/grddl/

<pimpbot> Title: W3C GRDDL service (at www.w3.org)

Harry: GRDDL uses it. Takes a DOM, gives you a graph
... Also a concern of POWDER WG
... If you're a new WG, and want to have something like a stylesheet, we want to know how
... the problems are with the values that are allowed in the rel attr

<DanC_lap> (since we're in france, we shouldn't treat "bis" as some secret code. it's french for 2nd, yes?)

Harry: We need a rel value registry

<DanC_lap> http://www.w3.org/html/wg/tracker/issues/27

<pimpbot> Title: ISSUE-27 - HTML Issue Tracking Tracker (at www.w3.org)

<oedipus> http://esw.w3.org/topic/HTML/AbbrAndInitialisms#head-2df84342c22a995e35980b4adb3b3573040f8c2d

<pimpbot> Title: HTML/AbbrAndInitialisms - ESW Wiki (at esw.w3.org)

Harry: Could try to combine the list for HTML and Atom and other specs

<DanC_lap> bummer... tracker doesn't use the issue name for the title. it's rel-ownership @rel value ownership, registry consideration

<oedipus> proposed use of RDF-type resource for reuse site-wide for abbreviated forms

<Hixie> DanC_lap: "bis" is a fancy way of saying "again", i believe

<Hixie> (in french)

hsivonen: the Atom, XHTML2 and HTML5 all have ways of using full URIs as rel values, but they do it differently

<DanC_lap> ok

hsivonen: GRDDL can go the WHATWG wiki and document their rel values

<gsnedders> DanC_lap: second is deuxieme

hsivonen: You can use a full URI, but there are limitations that tokens are compared case sensitively

PhilArcher: [Chair for POWDER] I talked with Mark Nottingham

<DanC_lap> (Bis, a musical term and a little-used Interlingua word meaning "encore", "again", or "twice" -- http://en.wikipedia.org/wiki/BIS )

<pimpbot> Title: BIS - Wikipedia, the free encyclopedia (at en.wikipedia.org)

Phil: The WHATWG wiki is one possible way. The IANA solution is another
... Microformats has another
... Everyone does it differently. Can we have one

I don't think it should be one groups wiki, I don't think it should be difficult to use

scribe: Need some sort of control over it

<Steeeven> RDFa allows a CURIE

<anne> From Murray Maloney: http://xml.coverpages.org/maloneyRelationships.html

<pimpbot> Title: Hypertext links in HTML (at xml.coverpages.org)

<oedipus> why not in w3.org space?

scribe: I believe that W3C would be willing to operate such a registry
... If it is possible to come to consensus, then we could take it to the appropriate people and move forward

<fantasai> ... it could be a wiki at w3.org

<fantasai> ... it would have to be in agreement with IANA / IETF

MikeSmith: We're not chartered as a WG to make binding decisions at face to face meetings
... But we could discuss what would be the best way to proceed

Julian: The goal from Mark's draft is to unify the meanings of relationship names
... Needs to be more formal than a wiki, but I'm not going to defend IANA's registry

hsivonen: I'd be ok with sharing the short values with Atom
... with the caveat that I haven't reviewed the Atom values to check for conflicts
... If you use a short name, then you compare it ASCII-case-insensitively and try to get them on the same list in Atom and HTML

<pimpbot> Title: XPointer Registry (at www.w3.org)

<smedero> DanC_lap: :-/

hsivonen: I can understand why w3.org is preferred over whatwg.org

<hhalpin> +1 on Henri's algorithm for comparison

<DanC_lap> xpointer registry. boo. hiss. I'm _still_ against registries at W3C.

Philippe: Traditionally W3 have been against registrys, but we have one

<Steeeven> +1 to DanC

Philippe: there's a process involving mailing some list and waiting a couple of weeks

<Julian_Reschke> I don't care a lot where the registry lives, at long as it is one, not just a Wiki.

Philippe: if no-one objects, it's yours

<Zakim> plh, you wanted to mention http://www.w3.org/2005/04/xpointer-schemes/

<pimpbot> Title: XPointer Registry (at www.w3.org)

<oedipus> there should be one central place to which people can turn, and w3.org is the most logical and shortest

<DanC_lap> issue-27: note also http://www.w3.org/2005/04/xpointer-schemes/

<trackbot> ISSUE-27 @rel value ownership, registry consideration notes added

<pimpbot> Title: XPointer Registry (at www.w3.org)

Harry: There's one registry to rule them all approach, or different registries. We don't care too much as long as standards people know where to go and who to ask to update it
... IANA might make it harder. If it's W3C, then I'm ok with a wiki, or a human edited web page, with requests sent and responded to by email

<gsnedders> http://www.ietf.org/internet-drafts/draft-nottingham-http-link-header-02.txt — Mark Nottingham's Link header draft (I don't think there's been a link yet)

Julian_Reschke: One things is that there's already an IANA registry, then if it can be used, we should

<DanC_lap> (ah. I wasn't sure whether it existed yet. http://www.iana.org/assignments/link-relations/ )

<pimpbot> Title: IANA | Atom Link Relations (at www.iana.org)

Julian_Reschke: XHTML2 puts CURIEs into the rel attribute, which concerns me a lot
... We need to be careful with the question about whether a link relation can be an IRI
... It gets tricky when you want to use it in the HTTP Link: header
... Can map IRIs to URIs, but transformation doesn't round trip

Murray: There's rel values as might be recognised by browsers. Next, prev, TOC, etc.

<pimpbot> planet: Misdirection <11http://intertwingly.net/blog/2008/10/23/Misdirection>

Murray: It seems to me that it would be useful for there to be a list of all these that a commonly supported by browsers
... The second issue is rel values that are used by other types of agents

<oedipus> murray, do you mean something like: http://www.w3.org/1999/xhtml/vocab/

<pimpbot> Title: XHTML Vocabulary (at www.w3.org)

Murray: The way that HTML was designed early on is that the <head profile=""> was meant to contain a list of URIs, which could be used as the place to look for the rel value
... Used for GRDDL
... Use it to link to a profile document, look it up and find the ones you want to operate on

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

Murray: It's suboptimal because there could be conflicts among multiple URIs
... Would be helpful to use a URI as rel value
... Maybe we should have a new attribute like rel-uri or something to declare teh URI onthe element

<oedipus> http://www.w3.org/TR/html401/struct/global.html#adef-profile

<pimpbot> Title: The global structure of an HTML document (at www.w3.org)

<myakura> xhtml2 defines rel=profile if i remember correctly

Richard Ishida: I wanted to clarify that you can round trip between IRIs and URIs, but not IDNs and URIs

<hhalpin> I think that adding new attributes unless absolutely needed is a bad idea.

<oedipus> "This attribute specifies the location of one or more meta data profiles, separated by white space. For future extensions, user agents should consider the value to be a list even though this specification only considers the first URI to be significant." (HTML 4.01)

Julian_Reschke: [example of a IRI that won't round trip]
... We would need to specify the comparison rules
... In general, the rule has been to do a plain string comparison

<hsivonen> anne, sorry, my MathML is rusty: <semantics><annotation-xml>...

<anne> hsivonen, neither seems to be considered by HTML5 at this point though, am I missing something?

Steven: We need to plan for a future where everything uses IRIs

<oedipus> myakura, @profile is currently not in the XHTML2 draft: http://www.w3.org/MarkUp/2007/ED-xhtml2-20071024/attributes.html

<pimpbot> Title: XHTML 2.0 - List of Attributes (at www.w3.org)

Phil Archer: I understand that HTML5 is happy with an absolute URI being a rel value

Hixie: Yes

<myakura> oedipus, i meant the rel value "profile," not the attribute

anne: I don't want to be able to write everything as a URI. e.g. rel="stylesheet" shouldn't have an equivalent using a full URI

<gsnedders> This is how Atom works too, FWIW

Phil Archer: If you see a string that isn't a URI, it is taken as a relative URI in the HTTP Link header draft

<DanC_lap> ("stylesheet" isn't registered at http://www.iana.org/assignments/link-relations/ )

<pimpbot> Title: IANA | Atom Link Relations (at www.iana.org)

<gsnedders> http://www.iana.org/assignments/relation/alternate and alternate are identical @rel values in Atom

<Julian_Reschke> DanC: because currently it's atom only

scribe: We need a way to have a token that can be normalised, but be short. URIs can be long

<anne> having a token registry is fine with me

<anne> e.g. the WHATWG wiki :)

<DanC_lap> (I was just clarifying, since I heard phil say "... stylesheet, which is registered")

DanC_lap: I prefer one registry, but don't like the W3C for that

gsnedders: The only problem I can see with absolute URIs is using current relations like "stylesheet" is that you can't express them as absolute URIs without breaking backwards compatibility

<Julian_Reschke> See http://tools.ietf.org/html/draft-nottingham-http-link-header-02#section-4.1 for registrations for existing HTML link relations

<pimpbot> Title: draft-nottingham-http-link-header-02 - HTTP Header Linking (at tools.ietf.org)

Harry: The RDFa discussion was unclear to me, not sure what the technical issues were. Could someone please explain what the issues are?

<Hixie> MikeSmith, i believe you can just type "ack" to get the next speaker

hsivonen: I was vocal in that discussion against taking RDFa as-is
... A lot of it revolves around the way RDFa uses XML namespaces

<gsnedders> (my comment was meaning using URIs in the same way Atom does)

<DanC_lap> (Anne made the same comment, gsnedders . or pretty close. but maybe the scribe missed it the first time)

hsivonen: You have to know the namespace mapping context from the lower layer, that you then use as an RDF property
... It's confusing for authors to have to deal with these namespace mappings
... there's a problem for implementers ...
... [something about parsing and namespace mapping issues]

<DanC_lap> (hmm... I wonder if the TAG finding on qnames in content brings up that dynamic/scripting concern.)

hsivonen: We have a case where the DOM consistency principle is violated
... xml:lang in the XML namespace, the lang attribute in no namespace. This has caused a lot of bugs
... That's why I object to using XML syntax
... The other objection was related to people who aren't in the RDF community having to pay the "RDF tax"

<pimpbot> Title: HTML WG -- 23 Oct 2008 (at www.w3.org)

hsivonen: [example of having to compare local name and namespace for an element]
... every time I have to do that, it causes me to do something costly that doesn't really solve a problem
... RDF has similar complications

<anne> From Murray Maloney: Murray is the only person in the world who voted against namespaces in XML

hsivonen: I'd like a solution similar to GRDDL. People who want it can, but doesn't cause problems for people that don't

<Hixie> ack

<Hixie> aw, i was wrong

Steven: CURIEs aren't namespaces, they're just shortened URIs
... Similar things appear in e.g. Wikipedia
... The other thing is that RDFa works in HTML now.
... People use JavaScript to extract it
... I don't see what the problem is
... If you don't want it, don't use it

Harry: We generally have consensus that maybe the way XML implemented namespaces is kind of crazy
... But it's interesting that CURIEs as just one string, unlike prefixes/localnames. It's simpler
... 2 issues
... How can we use these URIs in attribute values? Is there a short version of the URI?
... This is a problem for people who want to use RDF
... Maybe all that needs to happen is that RDF technologies need to be clear about how they fit within HTML5
... as long as there's a way that RDFa can signal the use of CURIEs

hsivonen: It would be objectionable if the mechanism for associating the base URI with the prefixes were different for HTML5 and XHTML5

Harry: XHTML5 just uses xmlns?

hsivonen: Currently there is no defintion of RDFa in XHTML5
... You don't have a property attribute in XHTML5 today, it's not specified how to do RDFa now

<Steeeven> <meta name="prefixes" content="xhtml http://www.w3.org/1999/xhtml foo http://example.net/foo" />

hsivonen: The RDFa community assumes you would do it in XHTML5 the same way as XHTML 1.0

<pimpbot> Title: XHTML namespace (at www.w3.org)

Harry: Would HTML5 be willing to sort out some way to use short CURIE like values

hsivonen: I think the indirection of prefixes is a bug
... If the URIs are so long that you don't want to deal with them, then that suggests a problem with the RDF naming scheme

I suggestion you use a registry, and concatenate the short values with a base URI

<MikeSmith> ack

<hhalpin> I'd just like to note that instead of a religious war, it appears the delta between *some* usable form of RDFa and HTML5 is relatively small.

Julian_Reschke: What hsivonen just proposed seems like a harder indirection mechanism

<hhalpin> +q

hsivonen: yes, but the "RDF tax" is only paid by the RDF community

<MikeSmith> ack

<anne> Steeeven, note that technically the xmlns attribute from an HTML parser is different from an XML parser (the latter has a namespace)

<anne> Steeeven, same for xmlns:* of course

<hhalpin> anne - so, unknown attributes though, such as a unknown non-namespaced xmlns, would still be in the DOM and thus serialized by XHTML5.

<MikeSmith> http://intertwingly.net/blog/2008/10/23/Misdirection

<pimpbot> Title: Sam Ruby: Misdirection (at intertwingly.net)

<anne> hhalpin, actually, xmlns couldn't be serialized back

<anne> hhalpin, though maybe it indeed depends on the serializer


<anne> so is table delayed?

+Silvia Pfeiffer

<MikeSmith> anne, in 5 minutes

Fragment identifiers for audio and video content

Sylvia: [About <video>]
... We're talking about fragment identifiers applying to media resources, like video

e.g. video.ogg#time=5,12

scribe: also query strings like &track=1,2,3
... For video, audio and images

<Hixie> RB: Please don't use an ampersand!

scribe: The effect it might have on the <video> and <audio> element in HTML5

Hixie: start and end attributes will be removed. They overlap with this stuff
... We're simplifying the looping attributes

Raphael: Establishing the communication between the UA and server
... through HTTP headers
... The UA will still know that they only got a fragment of the resource

<anne> Hixie: we'll replace looping attributes with a simple loop boolean attribute

Raphael: Please join the Media Fragments working group if interested. It's open, public mailing list

<raphael> Media Fragment Home: http://www.w3.org/2008/WebVideo/Fragments/ and public mailing list: http://lists.w3.org/Archives/Public/public-media-fragment/

<pimpbot> Title: W3C Media Fragments Working Group (at www.w3.org)

<mjs> Hixie, fragments in the media URI aren't really replacements for start/end/etc timing control attributes, because you can't readily change them through API

<mjs> Hixie, you have to parse and reassemble the URI to do it

<mjs> Hixie, and I think doing so would per the spec require the video to be reloaded as well

<Hixie> mjs, join the media fragments wg and let them know :)

<Hixie> (or send feedback)

<mjs> Hixie, I don't see how the media fragments WG can address HTML5-defined DOM APIs....

<mjs> Hixie, are you saying you want them to take over the HTMLMediaElement interface?

<mjs> Hixie, I'm giving you feedback on your minuted statement that "start and end attributes will be removed. They overlap with this stuff"

<Hixie> mjs, oh they'll be removed regardless of this

<Hixie> mjs, (probably)

<mjs> Hixie, why?

<Hixie> see recent feedback to the list

<Hixie> there aren't good use cases

<mjs> for audio there are

<mjs> playing a fragment of an audio clip is easier than editing the actual audio file

<sicking> mjs, but do you want to change that fragment?

<darobin> mjs, as we say in French, absent people are always wrong

<mjs> sicking, let's say you are writing a web app that lets you select a range from an audio or video clip and then play that

<nessy> mjs, media fragment URIs allow you to do just that: play a fragment

<Hixie> mjs, for that use case, you'd use javascript anyway, so no need for anything but the existing js api

<mjs> nessy, but they don't let you change what fragment is referred to without having to construct a new URI

<nessy> mjs, what's the difference between changing a URI in a href attribute and changing the start and end attributes?

<mjs> Hixie, so you don't mean to remove the equivalents of this stuff in the JS API?

<mjs> nessy, creating a specially formatted string is a sloppy API design

<Hixie> mjs, seek to the start, cue a pause at the end, and play

<mjs> Hixie, that doesn't let you actually change the range that the play controls reflect

<Hixie> mjs, nor do the existing attributes

<mjs> Hixie, what do you mean? I would expect if start and end are specified, the little slider thing goes between start and end, not the full time range of the underlying media item

<Hixie> mjs, what if loopend is after end?

<Hixie> anne: i'm not scribe, it doesn't matter if i have a colon

<anne> s/mjs:/mjs,/g

<mjs> Hixie, the loop attributes make things more complicated, presumably you need the union of {start, end} and {loopstart, loopend} to be presented

<mjs> Hixie, I tentatively agree that fine-grained loop control is probably more complex than justified

<Hixie> mjs, i don't really understand the use case you are presenting, but i really should pay more attention to the room

<mjs> Hixie, although "just do it in script" is not a good solution for looping

<nessy> mjs, we will have a boolean loop attribute

<nessy> mjs, that's an independent issue

<mjs> nessy, a string syntax is not a substitute for a proper API (though an ok way to specify things declaratively)

<nessy> mjs, there already is an API to control media resources

<mjs> nessy, you don't talk to real media APIs by formatting special strings to tell them what time range to play and how many times

<nessy> mjs, the start and end attributes don't provide you with an API

<mjs> nessy, I am talking primarily about the HTMLMediaElement interface

<nessy> mjs, so am I

<mjs> nessy, not the content attributes

<mjs> (unfortunately the word "content" is ambiguous)

<mjs> nessy, the start and end attributes in the HTMLMediaElement interface certainly do provide an API

<nessy> mjs, we should take this offline from this room

<nessy> mjs, there will be more discussions on the mailing lists

<mjs> anne, do you recall the title of the thread?

<anne> mjs, i'll look for you

<anne> mjs, "video tag : loop for ever "

<anne> mjs, people from Apple have been involved in the thread, e.g. http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2008-October/016692.html

<pimpbot> Title: [whatwg] video tag : loop for ever (at lists.whatwg.org)

<mjs> that's a long thread

table headers

JoshueOConnor: The issue is how to mark up complex data tables, header associations

<Joshue> this is scary url http://esw.w3.org/topic/HTML/IssueTableHeaders#head-4e755761c9194f726c62cf815b251a464e9c4635

<pimpbot> Title: HTML/IssueTableHeaders - ESW Wiki (at esw.w3.org)

<Joshue> These are also relevant http://www.w3.org/Bugs/Public/show_bug.cgi?id=5822#c14

<pimpbot> 115822: gez.lemon, P2, NEW, 13The headers attribute should be able to reference a td

Al Gilman: The operational requirement from the consumer is associating an arbitrary cell with any other cell that provide context for interpreting the data

scribe: Chained headers
... Multiple tiers of context information within tables

<Joshue> Gez's complex table example http://juicystudio.com/wcag/tables/altcomplex.html

<pimpbot> Title: Child investment portfolios (at juicystudio.com)

scribe: It doesn't work in HTML4 because you can't have a TH that is a header for a TH

AlGilman: There has been dispute on the mailing list, people using many different terms.

AlGilman: There are 2 ways to do it: 1. Have a TD cell be a target of a headers attribute.

<Hixie> mjs, maybe their use cases aren't relevant, or won't apply to html5, and that would be good for them to know

AlGilman: 2. Allow TH to reference another TH with a headers attribute

<Joshue> some more background on a proposed solution using existing header/id combinations which is well supported in current AT http://esw.w3.org/topic/HTML/Action72Headers

<pimpbot> Title: HTML/Action72Headers - ESW Wiki (at esw.w3.org)

AlGilman: We would certainly like to make scope more effective

<Joshue> Smart span algorithm http://lists.w3.org/Archives/Public/public-html/2008Mar/0075.html

<pimpbot> Title: Smart span algorithm for table cells from James Graham on 2008-03-10 (public-html@w3.org from March 2008) (at lists.w3.org)

AlGilman: But we need to come together on a proposal for what all of the markup features are, including the headers attribute, algorithm, etc.

AlGilman: We still have to work out details

<MikeSmith> ?

<Joshue> +q

Joshue: There are a couple of solutions

Joshue: about half way down the wiki page

Joshue: Smart headers algorithm, TH referencing TH, etc.

Joshue: Smart Headers is a good algorithm

Joshue: Allowing headers="" to reference a TD element

Joshue: is also an elegant solution
... headers="" is widely supported by screen readers
... lets see if we can come to some kind of concensus

<Joshue> http://esw.w3.org/topic/HTML/IssueTableHeaders#head-fc23268c19f6b7ad3dbde901743900ab1053b433

<pimpbot> Title: HTML/IssueTableHeaders - ESW Wiki (at esw.w3.org)

BenMillard: At the moment, Firefox doesn't do anything with scope or headers

<Hixie> mjs, i wouldn't suggest removing curentTime and the cue range API, which is what you'd use for this.

AlGilman: But it's all in the DOM, even though it doesn't do anything special with them

Michael Cooper: A lot of accessibility is done through the accessibility API and would require the UA to do a mapping from the HTML DOM to that API

scribe: But it's also done through the DOM for web content

Joshue; it's important to bear in mind the limited support for scope="" in current UA and AT

MikeSmith: Is there a timeline for when you'll get to the remaining table headers feedback?

Hixie: I'll get to it when I get to it, but I can prioritise it if there are vendors that need me to

AlGilman: This is not an urgent matter, because in the short term people can still use headers="" in HTML4

AlGilman: I'd like us to work with Ben to tune the algorithm, and getting that into HTML5 and implemented in browsers

Hixie: It sounds like sooner rather than later would help, so I can help with that in the next few months

<Joshue> +q

<smedero> karl: ben doesn't have a laptop so I'll jump in: mozilla

AlGilman: If you only have headers="", then it's grotty for authors. But we should move forward with change to improve that

MikeSmith: Does that timeline work for you guys?

Joshue: I can't speak for any vendors, but it would be nice

<karl> smedero, I hope it is enough for him to be doing that correctly without making his "life at risk".

Hixie: The remaining ~2000 currently in the queue should be dealt with within a year or so

hsivonen: There's an open moz bug about exposing table relationships to accessibility APIs
... I encourage you to keep an eye on the mozilla bug, especially if the Smart Headers algorithm will be going in the spec
... It would be a shame to have them implement what's in the current spec, and then have it updated in the spec later

BenMillard: Al mentioned tuning the proposal. I wouldn't want to kind of shortcut the proposal

<anne> "hixified"

BenMillard: I think that the algorithm in James's inspector should be "Hixified" into the spec
... I think an Action item would be to document the diff between the spec and Smart Headers algorithm

Joshue: One of the things that came out in the PF meeting was that we were all talking about the same thing, using different names
... We should know what we're talking about
... get on the same page

<Zakim> MichaelC, you wanted to ask if it's not more efficient to get a round of feedback from implementers before putting a proposal in the spec

MichaelC: It might be more efficient if there is a round of feedback from implementers before it's in the spec

<hsivonen> Hixie, https://bugzilla.mozilla.org/show_bug.cgi?id=441445

<pimpbot> Bug 441445: was not found.

<Joshue> From the PF meeting we found an issue around the use of terminology in this issue. Conceptual headers, chained headers, nested headers etc are all essentially the same. I would like to see the terminology tied up so we are all on the same page. This could help the process all round.

hsivonen: I heard that there was direction of consensus on the validation side that headers would point to th, be valid and point to td might not be valid

<MichaelC> Just to clarify, since it needed verbal clarification, I do not advocate implementation of proposals, but was suggesting implementers might have worthwhile insights on a proposal (but should not implement it until it matures)

AlGilman: The current practice doesn't use HTML5 validation yet, we don't need to trouble you now

AlGilman: What we want to do in face time, we've done

[end of table headers discussion]

MikeSmith: We could now talk about HTTP auth

Julian_Reschke: OK

<anne> scribe: anne

<Lachy> MikeSmith: Lachlan will not be scribing tomorrow because he's done such a great job today

AG: the Web Security Context WG deals with the look and feel
... of authentication dialogs

[scribe is unclear if this was actually said]


[MS describes topics for tomorrow]

[Topic discussed now will be HTTP authentication integration with HTML]

HS: will unconference sessions be integrated into this schedule

MS: we need to talk about the authoring guide

HS: 11:45 AM will be about implied ARIA semantics

MS: that is until 12:30 PM

<Hixie> AM is 00:00..11:59 and PM is 12:00..23:59

HTML integration point for HTTP authentication

Julian: issue is that user agents put up a primitive user interface that authors cannot style
... originally the idea was that the response would actually return HTML with a login form, but that never happened
... user agents only pop up this dialog which is one reason user agents are not using authentication
... there are other issues: e.g. i18n
... these are not being solved because of chicken / egg problem

<MikeSmith> can somebody please find and post the link to the related discussion for this that was on the WHATWG list recently?

<MikeSmith> aaron schwartz e-mail, I mean

JR: AvK mentioned that another missing piece is the logout button
... I don't have a personal agenda. Just noticed there was an old proposal 1999? where the same issue was described

<Hixie> http://www.whatwg.org/issues/#WF2-http-auth-login-logout

<pimpbot> Title: WHATWG Issues List (at www.whatwg.org)

JR: simple form that the user agent would know that there's a user name and password and use that

Hixie: I was wondering why we want HTTP auth to succeed

JR: it's not necessarily HTTP authentication, but form based login works very badly if the user is not a person

Hixie: there are other ways as well, with tokens and cookies

JR: what specs?

Hixie: cookies, HTTP, etc.

JR: you need out of band information to realize the server requires authentication

<Philip> MikeSmith, http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2008-October/016742.html ?

<pimpbot> Title: [whatwg] fixing the authentication problem (at lists.whatwg.org)

Hixie: you need that anyway

JR: there's a framework for that; not sure we should get rid of it

Hixie: it's not clear we should make it better if it's not needed

JonasSicking: whenever we work on widgets, "we want to style this form widget"

JR: that's the issue I want to get fixed

<Philip> (in which the problem is that people don't want to use the proper solution (HTTPS), and instead use the worst possible solution (sending passwords in the clear) because the compromise (HTTP Digest) is ugly)

Hixie: not sure that's the right answer, but not immediately obvious why it isn't

<Philip> (or at least that's how I might interpret it)

JS (sicking): HTTP auth is broken in many many ways

scribe: I have 50 different passwords to remember; if one of these sites start failing, I'm hosed
... the form based login is really bad. Teaches phishing and such.
... we need to rid of passwords entirely
... OpenID, infocards, LiveID
... should we go directly for that, or should we try to make incremental changes on the mechanisms that exist today
... e.g. making HTTP auth secure (digest does clear text)
... and fix the styling issue so people will use it
... should we go for ID management or should we try to do some fixes

CWilso, that's not what he said :p

Hixie: The way to solve the problem is not to assume HTTP auth is the solution

<adrianba> anne, he said Microsoft Passport

JR: the HTTP auth framework does in theory support other authentication schemes
... e.g. OAuth
... one example, portal project where document management was through the browser and webdav clients
... the automated clients need HTTP authentication to work
... servers start sniffing user agent string or even method name

<hhalpin> Although I am technically outside, and would like things like LiveID/OpenID and OAuth pushed, I am not sure if they should be put in HTML5.

adrianba, right, so the correction was wrong, doesn't matter though

scribe: etc. need nasty workarounds with the current situation

HS: one of the important things for adding something to the platform requires a rollout strategy that cannot be blocked by a browser vendor
... e.g. ARIA, the way it was defined it could not be ignored; though it did end up being implemented by all four in record time
... how do you see the rollout strategy where you use OpenID for browser authentication and OAuth for servers

JR: I don't know; I thought brainstorming would be good

SP: I had a chat with TBL last night; he didn't like OpenID but he had some ideas for a new authentication scheme
... maybe someone should have a chat with TBL about this

JS (sicking): I'm not sure what he didn't like about OpenID

SP: complexity; going to another website, extra steps, etc.

JS (sicking): I think with having OpenID in the browser will solve that problem

JS (sicking): the redirect is needed because they innovate within the browser box

JS (sicking): as vendors we can change the browser

scribe: we had issues, not sure what the issues were
... people have voiced security concerns with OpenID
... my ultimate point was having ID management rather than password management

AG: ... working with captchas(sp?)
... we want to encourage single sign on
... the PFWG is reopening our note regarding captchas
... the state of the art is that these tests that the commercial botnot businesses can't crack, people can't do either
... it's more rationale to use a single sign on service
... for more of these transactions and we'll be pushing people in that direction

HH: nobody disagrees with single sign on and that it should be put in the browser
... there are several solutions in this space
... should HTML5 endorse one solution or provide a hook
... or will the true solution arrive in a year or two, maybe wait it out a bit
... really important, should be done

AG: does this sound like a requirement for security API between the browser and the protocols

HH: not sure how to design such a hook
... I would like to see some statistics on various ID mechanisms

AG: we should live with the redirects for longer is what you're saying

HH: get some stats, see what's out there;

sicking: I think HTML5 should be completely silent on this
... I think that should be a separate work item

<hhalpin> this is really important, someone should eventually sort this whole identity thing out.

sicking: we're sure to get something better eventually
... there might be some API that is needed but in general you'd like to have the security features separately designed

HS: I have a question about what candidates are out there other than OpenID, SAML

<hsivonen> SAML WebSSO

<Adam> http://en.wikipedia.org/wiki/SAML

<pimpbot> Title: Security Assertion Markup Language - Wikipedia, the free encyclopedia (at en.wikipedia.org)

<hhalpin> also, note that I recommended some tight liasoning with people from id and security fields.

AB: MS has introduced infocards

<hhalpin> I always thought it was OpenID/Liberty Alliance and LiveID/Passport were the two big options in this space.

<hhalpin> There may be other experimental ones.

<hhalpin> +1 that HTML5 is interested in these issues.

ThomasRoessler: no particular input right now

MS: where do you think we are?

<hhalpin> Has Microsoft adopted OpenID anywhere? I thought it was possible, would HTML5 endorsement make a difference?

JR: I wished we would have made some progress on that, but it seems we need more research

MS: maybe discussion needs to continue in Web Apps?

Hixie: depends on the solution, not clear what the requirements are

sicking: is it a W3C matter?

JR: IETF is waiting for W3C; it's a user agent issue

<CWilso> Actually, my understanding is infocards interoperates w/OpenID if desired

sicking: depends on the technical solution probably
... I think it should be an effort solely concentrated on security

<CWilso> (And it's called CardSpace now, not infocards)

<adrianba> infocard/cardspace = http://msdn.microsoft.com/en-us/netframework/aa663320.aspx

<pimpbot> Title: Windows CardSpace (at msdn.microsoft.com)

ThomasRoessler: in principle that makes a lot of sense. (TR says he missed bits of the discussion) The one thing I'm wondering what problem you are looking at?

Jonas: I don't see how headers solve the problem

JR: the issue is as follows. you do a GET request. server wants you to authenticate. can do a 200 with HTML form
... works extremely badly with clients that are not HTML clients
... or you can use 401 and then you've got a problem in HTML agents that they pop up a dialog most designers and people running servers don't like
... both are bad
... you really don't want to reply with 200 OK if you require login

Hixie: there's also sites where you can either login or not login

ThomasRoessler: the current HTTP auth mechanism are no longer state of the art (to be polite)
... there is a question of what HTTP auth should look like
... there is also something like client certificates e.g. SSL

<hhalpin> CWilson - that was my understanding as well, but I was not sure if that had ever happened yet. Thus my point that if HTML5 gets in this space, then maybe asking some advice from experts from LiveID, Passport, Liberty, OpenID, etc. and some data is the way to go. Also, could this activity, which sounds large, be done at HTML5 or somewhere else that integrates with HTML5?

ThomasRoessler: [..] this is a rather large ocean. in the CABForum they are trying to find out what the obstacles are with client side PKI
... is assuming a working PKI; problems with e.g. business and social interaction
... there is a member submission from 1999 along these lines; might be useful
... assuming this will solve auth problems on large scale would be naive
... assuming it would solve phishing is massively naive
... this is a HARD problem
... the 1999 document is not solving that problem, I suggest to not rathole in this topic in this WG
... don't think this is the right community (no disrespect)
... one solution to look at is OAuth
... using HTTP and HTTP auth to pass authorization tokens around
... this you use to e.g. upload your Flickr photos
... will be discussed at IETF meeting in Minn... in the US

<adrianba> hhalpin, cardspace does work with openid - i have my openid in an infocard, it does require the site to do something extra to enable cards though

ThomasRoessler: I won't be there, pay attention at the IETF though
... they're attacking a useful problem; not saying it's attacking the all the problems discussed here

<Julian_Reschke> http://www.w3.org/TR/1999/NOTE-authentform-19990203

<pimpbot> Title: User Agent Authentication Form Elements (at www.w3.org)

AG: the HTML processor in the browser is going to see an API
... going to be disconnect point so you can't change schemes
... you can annotate on that
... would be easier if we had an XForms model
... what's defined is what you can style
... scheme by scheme you get controls that the web author then can style
... so first IETF, then HTML so you get the style

ThomasRoessler: one problem with usability of passwords; password managers can log what passwords are entered when you login
... what can be useful is to distinguish the password fields for registration
... the user agent can take control and help out

<Hixie> you can set pattern="" on type=password already in html5

<Hixie> there's no way to say it's for a new password though

HS: if you have passwords you can use these from any browser
... in countries were banks want some weird auth scheme, e.g. Java
... if you have a full browser but no binary plugins on a phone, you cannot get authenticated
... it's important that new mechanisms work on mobiles as well
... e.g. native OpenID support in the UA; but fallback for mobiles

ThomasRoessler: in order to compat phishing we need to [...]
... one way out of that would be to help people out in the interaction
... just taking the UI out will not solve that problem
... e.g. Google wants to style their logins forms to the pixel

AB: at boeing it would be really good if we got single sign on
... lots of different APIs that could make use of that

ThomasRoessler: you should point people to boeing presentations given at various concordia workshops

<karl> http://projectconcordia.org/index.php/Main_Page

MS: we are adjourned for the day

Summary of Action Items

[NEW] ACTION: Mike lead HTML WG to response to TAG discussion and report back to TAG [recorded in http://www.w3.org/2008/10/23-html-wg-minutes.html#action01]
[NEW] ACTION: TimBL write up a description of the kind of spec HTMLWG is writing for TAG [recorded in http://www.w3.org/2008/10/23-html-wg-minutes.html#action02]
[End of minutes]