18 Oct 2007


See also: IRC log


Stuart, Dan, Norm, Henry, Noah
Dave (partial), Rhys, Raman, TimBL


Accept this agenda?

DanC: It was prepared with the expectation of Tim's presence, so we'll see.

Accept minutes of 11 Oct?

<DanC> +1 OK http://www.w3.org/2007/10/11-tagmem-minutes


Next telcon: 25 Oct 2007

Stuart will chair; Rhys is confirmed to scribe.

We have regrets for Raman and Norm (Norm also for 1 Nov)

Joint IETF/W3C Meeting

<DanC> "DSKPP review brings up HTTPSubstrate-16 ISSUE-16"

DanC: I sent mail just yesterday. We have laison telcons about three times a year.
... This DSKPP review message is one of the things on the agenda.
... It brings up httpSubstrate-16 and maybe SchemeProtocols more than URN-Registries.

<Stuart> http://lists.w3.org/Archives/Public/www-tag/2007Oct/0067

DanC: The IPP protocol is http with a different scheme name; some consider this a feature, I consider it a bug.
... Our URNsAndRegistries is more about persistence than protocols.

Noah: When I started working in this space, I got several different "http is useful stories" but I never got it clear enough to write it down.

<DanC> (hi tracker; I sorta missed; it's schemeProtocols-49 / EditISSUE-49 , not so much URNsAndRegistries-50 ISSUE-50, that relates here]

Noah: Larry Masinter (at least) expressed that this is wrong, you should dispatch on the scheme name so you shouldn't use http instead

Stuart: A lot of the early URI scheme RFCs speak of URIs designating things and then go into an operational, protocol specific descritpion of what's designated.

<Noah> Just to be clear, I'm vaguely recalling discussions with Larry from about 3 years ago. It's my perception that his position is, or was at that time, that if you're using a different protocol, you should use a different scheme.

Stuart: As I understand Roy's position, there's a separation between what the URI designates and the actual protocol that interacts with it.
... Does that make sense?

Noah: Yes, but now we're getting into the space where things get hard.

<DanC> (should we change the topic to schemeProtocols? or should I get back to liaison logistics?)

Noah: At least some of the time, the scheme is bound to the protocol.
... My impression is that the http scheme in particular works along the lines of, if I get back a 200 then that is what the name designates.
... In a way that's deeper than in a mailto: scheme.

DanC: When it comes to these things, I get really blurry until we have test cases.

Stuart: Is that a topic we should come back to later?

DanC: The question is, do we ahve any input to the laison teleconference?
... Where the rubber-meets-the-road is when the IETF does other protocols. Like SIP and IPP which just use http under teh covers but we don't notice.

Noah: Do the SIP/IPP scheme names have the same flavor as HTTP URIs wrt DNS names, etc.
... I could imagine designing a protocol where I didn't use a DNS name but instead a local cache.
... I would argue that in that case, it's not really HTTP under the covers

<DanC> http://esw.w3.org/topic/UriSchemes/sip

<Stuart> http://www.ietf.org/internet-drafts/draft-ietf-keyprov-dskpp-00.txt

Henry: So why not HTTP?

DanC: Chris Newman says at least because its on a different port.

<DanC> (I can't find the sip URI scheme spec.)

DanC: "At this point its inevitable that we'll wind up with firewalls on port 80." So a cert revocation protocol, for example, shouldn't run on port 80 because proxies will be busy inserting ads

Noah: I can see where there might be good reasons to use a different scheme for a diffrent port.

DanC: I'd have to look at detailed cases, I don't think I have an overall philosophy.

Noah: To what extent would it violate webarch to run two servers on different ports which return conflicting representations.

DanC: That's ok because the port appears in the URI so there are two URIs.

Noah: So one way is to put the port numbers in the URIs.

DanC: Yes, but I wouldn't want to recommend that right now. I'd need more background.

<Stuart> b. Initial request from DSKPP client:

<Stuart> POST http://example.com/cgi-bin/DSKPP-server HTTP/1.1

<Stuart> Cache-Control: no-store

<Stuart> Pragma: no-cache

<Stuart> Host: example.com

<Stuart> Content-Type: application/vnd.ietf.keyprov.dskpp+xml

<Stuart> Content-Length: <some value>

<Stuart> DSKPP data in XML form (supported version, supported

Stuart: There are URIs in (scribe wonders what document) that don't use port numbers.

<Stuart> algorithms...)

DanC wonders if Noah would like to attend the meeting next week.

Noah agrees, pending any conflicts on his schedule.

<scribe> ACTION: DanC to wrangle an invitation for Noah to the W3C/IETF liason call [recorded in http://www.w3.org/2007/10/18-tagmem-minutes.html#action01]

<trackbot-ng> Created ACTION-67 - Wrangle an invitation for Noah to the W3C/IETF liason call [on Dan Connolly - due 2007-10-25].

Some discssion of logistics

Stuart: Any more discussion of this topic?

DanC: There are some patent policy and http issues.
... http is an IETF spec. Their patent policy is what it is. Yahoo is happy with that. Some of our members are less so.
... We're now engaged in the business of figuring out how much time we can contribute to the 1.2 effort.
... I'm afraid that the lawyers are going to jump in here, but I can't get out of the way.

<Zakim> ht, you wanted to mention headers

<DanC> (1.2? I don't think anybody's talking about a new HTTP version.)

<DanC> (pointer to the issues list would be handy)

Henry: This is all in aid of a new http: edition; if you look through the issue list you'll find Larry Masinter's "deprecate content negotiation"
... The thing I was looking for was, at some point in the last three years, we talked about a solution to aproblem that involved using a request header which never made it out of internet draft into the final spec.
... I can't remember what it was.

<DanC> Link:

Henry: Ok. Is it appropriate to get that on the list of things?

DanC: Yes, that's already on the agenda.


Stuart: What's the status on action 7?

DanC: I did a bunch of prep work to brief TimBL, but since he's not here...

<DanC> Subject: HTML validation and extensibility, update [tagSoupIntegration-54 ISSUE-54] [ISSUE-33 mixedUIXMLNamespace-33]

<DanC> Date: Wed, 17 Oct 2007 17:13:22 -0500

<Stuart> http://www.w3.org/mid/1192659202.25511.206.camel@pav

<DanC> Message-Id: <1192659202.25511.206.camel@pav>

DanC: I have this action to get Tim and Olivier together in the same room. No success there yet.
... Olivier has asked if the validator can be spruced up to do namespaces.
... Doug Schepers is trying to work out how to validate compound namespace documents with the role attribute.
... He has a black and white test case, which makes me happy
... Chris Lily says the SVG approach to this is both formal and in prose, using NVDL ignore everything except SVG and then use the RELAX NG validator.
... You can take this validator at Doug Schepers test case and it gives you a thumbs up.
... The relaxed user interface (NVDL) allows you to choose which namespaces you want to validate with.
... In our discussion of substitution groups and stuff, my goal has been to start with the document and follow your nose.
... It seems feasible that the CDF WG could develop an algorithm for computing NVDL scripts in "follow-your nose" bottom-up style.

Henry: I think the implied architecture there is awfully baroque
... NVDL is already a validation pipeline, now you're saying you have to start by computing that pipeline.

DanC: What I want to do is walk up to the validation service, give it a URL, push a button, and get a result. I don't want to have to do anything else.

Henry: There are two problems: One is that NVDL/RELAX NG, etc. only work with XML.

DanC: Can't we just pretend everything is XML for a while?

Henry: Sure, but then the scope of options is much larger. For instance, if we're prepared to say we're only using XML, then I think a W3C XML Schema for XHTML with the necessary wildcards in all the right places is a possibility.

DanC: I pushed in that direction and Chris Lily pushed back pretty hard.

Henry: Why?

DanC: I don't know. I was pushing on using wildcards/substutition groups and he came back with this NVDL/RELAX thing that did just what I needed. I don't care about the technology as long as it works.

Noah: How are we using these schemas?

<DanC> (replay, for dorchard : DanC: What I want to do is walk up to the validation service, give it a document, push a button, and get a result. I don't want to have to do anything else. )

Noah: There's a sense in which a grammar sets down what the rules are. So in some cases, you make sure the rules are ones that you can express in the schema.

<DanC> (oops... URL, not document)

Noah: In other cases, you know that you can only realistically capture 80% of the rules.
... I thought we were working on the story about what rules we could express and how easily.

DanC: I want to put some software in the validator. I don't care about which schema language.

Noah: We should be working on the rules we want, aware that only some of them can be expressed in the schema.

Dave: Right.

DanC: I think we're largely agreed on the general direction, but we have rubber-meets-the-road cases coming up soon.
... MathML 3, for example, mixes MathML and Forms (X + <input> = 3)
... The part of the spec that talks about the schema is "TBD"
... I pointed them at the SVG approache of NVDL+RELAX NG because I'd had a positive expreince with it.
... There's the aria specific case. That's really hairy.

Noah: I'm still a little nervous. I think we need to decide on the rules independent of the schema language.
... By all means if we can get a uniform answer, that's valuable.
... Then the burden is on the schema langauages to make it easy to support the cases that users want.
... I worry that if we say that all solutions should use an NVDL-like mechanism because that was a mechanism that was convient to use

Scribe missed a bit of that

<Zakim> DanC, you wanted to talk about validator and users, and note that NVDL is friendly with lots of schema languages

DanC: Let's focus on user expectations
... What's good about validators is that they setup a feedback loop between standardization and the authoring community.

<ht> HST reminds us all of the wise words of Dan Connolly "Validate at trust boundaries"

DanC: One simplistic view is that a guy puts together some web content and he wants to know if the W3C thinks this is good.
... One of the thing the TAG has talked about is languages and texts.
... One simplistic answer is to give the single-bit answer: is this in the set of texts or not?
... The other extreme is all the best practices..
... The current state of the art is to just give the single-bit answer based on the schema provided.

Noah: I'm completely supportive of that. It just might be good to do more than the schema validation.

DanC: I know almost nothing about NVDL except that the script for the SVG case was less than a page and was friendly with lots of schema languages.
... So we do have this CDF WG (charter renewal aside), and there is this talk of compound document owrk.
... It would be nice if the Aria, the SVG, and CDF folks could all do what's best for their community and the validator could just get it right.

Stuart: So your action continues, yes?

DanC: Yes.
... XHTML Modularization is in last call so there's this weird chain of events where the Aria design is constrained by DTDs by way of XHTML modularization.

Henry: What!?
... What is Aria?

DanC: The design sort of walks this gray area between being a module of XHTML and a set of HTML kludges.

<Stuart> this may be place to start: http://www.w3.org/TR/aria-roadmap/

DanC: The HTML 5 design methodology is "ok, there's clearly a problem that needs to be solved, let's talk it through and pick the 27 short names we need"
... At the other end of the spectrum, Aria is designed with full semantic-web buzzword compliance.
... It's totally framework complaint.
... I can appreciate both approaches.
... The reason that this is challenging is because accessibility folks are asking "what code should we commit tomorrow". They'll wait a little but, but...

Some discussion of how DTDs are involved

<ht> HST sent email to www-tag pointing out the XHTML role attribute WD and its relevance to the TAG. . .

DanC: The XHTML 2 WG has a weekly call. The Aria folks know how to do that style.

<Stuart> vis abbreviatedURIs-nn ??

<ht> Stuart, yes

DanC: Those guys are doing DTD-based modularization so that's what the Aria folks are working on.

<Zakim> ht, you wanted to ask DO if he thinks HTML4's 'must understand' can be captured w. XSDL1.1

Henry: How close do the new "wildcards everwhere" shorthands in Schema 1.1 come to allowing you to write a schema for XHTML that implements the must ignore rule.

<Stuart> http://www.w3.org/TR/xhtml-role/

<DanC> (HST's question sounds like the sort of black-and-white test case sort of thing that would make me happy)

Noah: I think the answer is probably close enough to say yes, but there might be a lot in the last 10-20%
... Because you ignore the tags and not the content, you don't want a skip wildcard.
... That feels like a lax wildcard and it might work, but I can't decide if it might undercut any other constraints that you might have.

Dave: yes, this comes up from the phtml example with the bananas in it
... When you add the banana content, what are you actually constraining the banana content to have wrt to the already defined elements.
... So the issue about adding something into V2 and figure out what its restrictions are is tricky.

<ht> HST notes that of the 85 elements in XHTML11, only _five_ have substantive content models: frameset, head, html, ruby and table.

Noah: The other thing that I think might be an issue is, so my language consists of all this stuff (images, paragraphs, etc.) plus...pretty much anything.
... And I'm a little worried that the anything might suddenly allow a head inside a paragraph.
... I'm not saying it doesn't work, but I think it would be real work to prove the edge cases.

Henry: I wasn't asking about XHTML 2. I'm looking for something that Olivier could do tomorrow to make the validator do what people want.
... If it can't be done, then we don't need to talk about it. But if we really could do it tomorrow, then we need to talk about whether or not we should.
... I think it might be possible because of the on-the-fly design discussion that Noah and Dave just outlined.
... There's a new version of my XHTML modularization review that we need to discuss.

<ht> http://www.w3.org/2001/tag/2007/09/xhtml-modularisation-thoughts.html

Henry: but probably not today.
... the headline review is that I implemented the substitution groups and it works.
... The other thing is that the element side is clean and easy to understand, but the overall result is pretty opaque because you need to use redefine for the attributes.

<DanC> "M12N has a CR transition call soon" --

<DanC> http://lists.w3.org/Archives/Public/www-validator/2007Oct/0047.html

Henry: I need substitution groups for attributes and I'm trying to get that feedback to the Schema WG.

DanC: Not only is the Schema WG doing some design but the modularization folks are also working.
... I've heard rumors of a CR transition call
... I'd like the results of Henry's investigation to be fed back to the XHTML WG

Norm would like to review it, but won't have time.

<DanC> (for my money, http://www.w3.org/2001/tag/group/track/actions/48 is record enough of HST's action)

Scribe missed something about CURIES and a Rec-track document

<DanC> (HT sent mail about it, Norm; don't worry)


DanC: Related mail takes us to the message from Martin writing to the Core WG.

<ht> CURIEs have surfaced for the first time in a REC-track document, namely http://www.w3.org/TR/xhtml-role/, so we need to look at that document wrt abbreviatedURIs-??

DanC: I think the takeaway is, we'll update the IRI spec.

(As opposed to a separate “HRRI” spec.)

DanC wonders about test cases

Norm: I think the XLink test suite has test cases for that.

DanC: What's the brand name?

Henry: It's going to be an "Extended legacy IRI"

<ht> LEIRI -- Legacy Extended IRI

<DanC> I'm happy with that long name as a deterrent to doing href="Some document"

Henry: There are too many specs out there that say things like system identifiers are strings. So we need a way to say how you turn one of those into IRIs.

DanC: And the flip side is that code is also copied and pasted around.
... If I continue to get yes answers about whether the Core WG has done the tests, I might declare victory.

<ht> http://www.w3.org/XML/2006/03/xlink11-tests

<DanC> http://esw.w3.org/topic/UriTesting

DanC: I haven't made any progress since we last talked about it.

Stuart/DanC agree that it's still important though.

DanC: So our action continues, ok. And so does Tim's.

<ht> http://example.org/xlink target


Henry: I've not made any progress on this yet.

DanC: That's ok, the point is to not let Chimezie's comment get too old.
... Can you reply just saying we'll get to it.

Stuart: I think you agreed to turn the document around a bit in response to that comment.

<DanC> http://lists.w3.org/Archives/Public/www-tag/2007Aug/0030 Chime's comment

Henry: Yes, I think you're right. I'll reply along those lines.

<DanC> "Comment on URNs, namespaces, and registries "finding": Some unfair characterizations"


Stuart: I've started working on that item. There's an email thread...

<Stuart> http://lists.w3.org/Archives/Member/tag/2007Oct/0035.html

Stuart summarizes

Henry: I didn't think their reply was exactly positive.

Stuart: Yes, I think there was some of that flavor.

Stuart describes some tension about where the document actually lives

Stuart asks about the possibility of adding an editor.

DanC: My preference is that we get quick turnaround, not that we add an editor.
... They're waiting for input from another group which is useful information.

Stuart: Would we like to invite them to one of our meetings?

DanC: That's a complicated question.

Stuart: I meant a telcon.

DanC: Oh, yes. That's straightforward. I think we should try to get both the chair and the editor on the call.

Henry: I think a meeting is a good idea.

TAG Nov F2F items

Stuart: I've started a list, if you want more or different items, please speak up.

Henry: I think it might be good to talk about what the namespace document 8 document actually *is*
... Supposing we like the story, what should we then do?
... The interrelated questions of the schema document, schema, namespace document, namespace, slash, hash, etc. are all candidates.

Stuart: I'd like to make some progress with Dave on versioning.

Dave: We got some feedback from Noah.

Stuart: Should we talk about the comments first?

Noah: I think it probably does make sense to see what others think.

Stuart: They resonated with me.

Noah: I think anything we do on drafting is a gamble on Dave's time.

<scribe> ACTION: dorchard to draft an update for 25 Nov 2007 [recorded in http://www.w3.org/2007/10/18-tagmem-minutes.html#action04]

<trackbot-ng> Created ACTION-68 - to draft an update for 25 Nov 2007 [on David Orchard - due 2007-10-25].

Noah and Dave discuss how best to address the drafting question.

Any other business



Summary of Action Items

[NEW] ACTION: DanC to wrangle an invitation for Noah to the W3C/IETF liason call [recorded in http://www.w3.org/2007/10/18-tagmem-minutes.html#action01]
[NEW] ACTION: dorchard to draft an update for 25 Nov 2007 [recorded in http://www.w3.org/2007/10/18-tagmem-minutes.html#action04]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2007/10/18 20:40:13 $