Technical Architecture Group
6 Sep 2007


See also: IRC log


Norm, Stuart, Raman, Dan, TimBL, Henry, Noah


Approve minutes of 30 August 2007

-> http://www.w3.org/2001/tag/2007/08/30-minutes

RESOLUTION: Approved as a true record.

<DanC> +1 meet 13 Sep

Agenda review

Accepted as posted.

Next telcon, 13 Sep 2007

Proposed scribe: Dan

Norm gives regrets.

September F2F Preparation

Stuart reminds the TAG to send their MAC addresses to the local organizers

Stuart points to fuller agenda.

-> http://www.w3.org/2001/tag/2007/09/17-agenda

DanC: It looks like we're favoring redirection and semantic web more than versioning and tag soup. But that's probably ok.

Stuart requests agenda proposals by email

Noah: Chances that I have a new draft on self describing web are dropping fast.
... I think I have a clear sense of what people would like it to be, assuming we decide to do anything, but it's going to take more effort.
... I would suggest that we leave self-describing web off the agenda for the f2f.

<DanC> (move "The Self-Describing Web" from recommended reading to "maybe weds")

Noah: Do we know what the real focus of the f2f is going to be?

Stuart: Redirections certainly.
... Some topics around TagSoupIntegration too

<timbl_> Location: http://www.w3.org/2001/tag/doc/httpRange-14/2007-08-31/HttpRange-14

<timbl_> Line Number 15, Column 18: <day>&doc.day;</day>

Stuart: And an attempt to look to be doing some work in semantic web and WebArch/Web 2.0.

<timbl_> ... from http://www.w3.org/2001/tag/doc/httpRange-14/2007-08-31/HttpRange-14

Stuart: Fragment ID stuff of recent discussion seems relevant.

Noah: Semantic web?

Stuart: We've grouped that with ... scribe missed ...

Noah: This plays into the self-describing web too.
... Microformats to URIs to self describing...

<DanC> (well, yes... lots of this comes down to principle #1: use URIs. But that principle has to be taken down to more detailed issues in order to impact developers, in my experience)

DanC: Question for technical discussion: do we want a new MIME type (for (X)HTML)?
... I'd like new authors to learn to put slashes in their BR tags and such

Raman: By induction, I don't think we do. You can say application/xhtml+xml is slowly succeeding or it will never succeed. If you think it'll never succeed, you'll get beaten up for trying again. If you think it'll eventually succeed, you'll get beaten up by the other group.
... Adding a new MIME type won't fix anything.

DanC: I wonder if conneg is really worthwhile, globally. (Of Henry's conneg story earlier, about shipping application/xhtml+xml to everyone except IE)
... Why is text/html bad

Henry: Because the specs say don't do it.

DanC: But we can fix those specs.

Raman: Right, we can.
... Can we fix the specs so that XHTML 1.0 can ride over text/html.

<Stuart> Take a look at the warning that the validator generates at: http://validator.w3.org/check?uri=http%3A%2F%2Fwww.w3.org%2F2001%2Ftag%2Ftag-weekly

Raman: If Google gets application/xhtml+xml, we aren't sure how to serve it. We don't want an IE download box.

DanC: Meanwhile, MS has this very strong position about versioning that says however IE6 treates web content, that's how IE7, IE8, etc., will treat it the same way.
... So a new MIME type will give them an opportunity to do the right thing.

<timbl_> ; xml=1.0

<Zakim> timbl_, you wanted to suggest the w3.org could ship a lot and see whetehr anyone hurts; b ht, where did you find it in the sepc not to stdn tex/html?

Some arguement about why IE behaves the way it does, missed by scribe.

TimBL: I notice that Henry/Dan didn't seem to agree on whether the specs do or don't say you can't ship xhtml under text/html. It would be nice to track that down.
... We can change specs, but we can the W3C can also lead by example. We could change large chunks of it so that they are XHTML and are shipped as XHTML.

DanC: We already do that, but one of the events coming is an attempt to make the home page more mobile-happy and the mobile web guidelines say you should use the new MIME type.

s/shipped as XHTML/shipped as text/html/

Henry: The validator also gives a warning.

TimBL: *That* we can definitely fix.

<DanC> (we're discussing an action about the validator, after all)

TimBL: The thing that's hard to change is IE.

<Stuart> from http://www.w3.org/TR/xhtml1/:

<Stuart> 5.1. Internet Media Type

<Stuart> XHTML Documents which follow the guidelines set forth in Appendix C, "HTML Compatibility Guidelines" may be labeled with the Internet Media Type "text/html" [RFC2854], as they are compatible with most HTML browsers. Those documents, and any other document conforming to this specification, may also be labeled with the Internet Media Type "application/xhtml+xml" as defined in [RFC3236]. For further information on using media types with XHTML, see the informative note

Raman: What I've seen is that serving cleanly done XHTML content as text/html never does any harm.
... Maybe it does bizarre harm somewhere, maybe if you take some HTML page where you've used some browser specific DOM hackery in your script.
... If you take such a page, bring it into emacs, and balance your tags and then present it as XHTML 1.0, you'll be surprised.

<DanC> (some tricky bits: implicit tbody, <div />, etc.)

<ht> TimBL -- here's the place where W3C REC tells you to use application/xml+xhtml for XHTML: http://www.w3.org/TR/2002/REC-xhtml1-20020801/#media

Raman: That's because your script relied on some bizarre browser fixup that's different from what you will achieve by hand.
... That's not an artifact of versioning or anything. It's a case of debugging your app in one environment and then deploying it in another. YMMV.

Noah: Part of the follow your nose store on XML stuff takes you through RFC 3023. If you see something with +xml, you can still treat it as XML. So there's a chance that something you would other have served as text/html runs into that file and you lose that.

<DanC> (the GRDDL story for text/html is not potential; it's actual, with running code and published test cases)

Noah: Whether there are any tools that rely on that is an open question.

Raman: Since none of the other things have happened, probably not.

Noah: You could put it in an XML database.

Raman: Yes, but you'll be surprised.
... Putting angle-bracket soup into an atom container just results in binary blobs in your database.

Norm tried to prevent that escaped markup crud in Atom and lost

Noah: That's what you're getting today.

Raman: What I'm saying is that the status quo will still arise if you invent a new MIME type.
... The only MIME type that's relevant on the web is text/html.
... The content that turns up *as* text/html *has to be cleaned up*
... We'll continue to create tag soup in text/thml and we have to fix that inorder to do anything useful downstream.

<Zakim> ht, you wanted to elaborate on that pointer

Henry: Coming back to TimBL; someone who was not aware of the controversy would certainly be confused about how to serve their XHTML.
... The most commonly quoted locus is what I put into IRC above.
... That text says that you can serve it as text/html if you follow the compatibility guidelines.
... It doesn't say, but most folks assume, that this means they should use for application/xhtml.

Some arguments about whether or not XSLT's output method has a bearing

Stuart: There are a bunch of action items in this section.

Scribe realizes he's not sure when we moved off of f2f prep.

<ht> s/for application.xhtml/application/xhtml+xml when the XHMTL does _not_ follow the compativility guidelines/

DanC: Dave's message is fine but you can't tell if the HTML WG agrees with it or not because it doesn't ask for any concrete changes.

Noah: DanC, do you think that there's something more effective that we could do?

DanC: Yes. It helps if you give real serious evidence that you read the spec.
... Dave's message doesn't do that.

Noah: Fair enough, but you haven't said what you think we could do specifically that would help.

DanC: What I'm saying is, whether we do the larger thing or not, it doesn't hurt to do the smaller thing.

Stuart: Is there anyone amongst the TAG that has or intends to read the HTML 5 spec.

Raman: I have read it.

Noah/Henry say they've read parts of it.

DanC: I want ACTION 7 continued. There's been a fairly good thread on the www-tag list.
... called XML Namespaces and extensibility or something.

<DanC> (crap; database failure at http://www.w3.org/2001/tag/group/track/actions/7 )

TimBL: What's the next step? Maybe a table of things that people do wrong and what the validator should say about them.

DanC: Right. The "we" in this case is you, me, and Olivier. But it looks like the thread is live again.

<DanC> From: Mark Nottingham <mnot@yahoo-inc.com>

<DanC> To: olivier Thereaux <ot@w3.org>

<DanC> Cc: www-tag@w3.org

<DanC> Subject: Re: XML, namespaces, extensibility and validation

<DanC> Date: Wed, 5 Sep 2007 11:15:56 +1000 (Tue, 20:15 CDT)

DanC: There's some good stuff in that thread, but some of it's more than three months ago.
... I'd like to do something about this in the coming week. Wonders if TimBL is already booked.
... in prep for the 13 Sep TAG meeting.

TimBL: I have to give regrets for 13 Sep.

Some additional discussion of scheduling administrivia

DanC: Ok, maybe we'll have something for the 13th but the odds are low.

Stuart: Moving on to other actions. Action 19 is done.

Raman: Somewhere we should add in the MIME Type somewhere in that document, but I'm not sure how.
... If you know how, please do!

Stuart: Action-42 we've just spoken about. Resolved.
... We plan to go ahead, but we haven't done sufficient reading of the HTML 5 spec yet.

Dave: We don't want to go to far in the way of technical review because it's not our job.
... It's still not on the list of things the WG is going to talk about. What we're saying is, we'd really like you to make sure you talk about this, without too much regard to the technical solutions.

<scribe> ACTION: David to post his message to the HTML WG list [recorded in http://www.w3.org/2001/tag/2007/09/06-minutes#action02]

<trackbot-ng> Created ACTION-44 - Post his message to the HTML WG list [on David Orchard - due 2007-09-13].

Henry: I gave a talk at Extreme. It got a lot of interest at the conference, but Extreme is so far from the heartland of the HTML WG that that's not meaningful.

<ht> http://www.idealliance.org/papers/extreme/proceedings/html/2007/Thompson01/EML2007Thompson01.html

DanC: How about a short presentation at Southampton?

Henry: Sure.

<DanC> (yeah... tech plenary is another possibility)

Raman: Another venue for that could be at the plenary if there's an HTML/XHTML session.


Stuart: I'm not clear on the status of the actions.

DanC: I don't want to stop talking about this altogether because of bug #1974

Henry: I drafted something a year ago. The Schema WG has been trying to get a document to Last Call for 18 months.
... Approving my draft may now get on the Schema WG's agenda.

DanC: The bug is "make a namespace document".

Henry: No, the bug is #int doesn't resolve.

<ht> http://www.w3.org/2001/XMLSchema#int

DanC: The 2001 XML Schema namespace includes the term "int". If you take that full URI and do the webarch thing, you lose.
... What should they put there? If the TAG doesn't care, I guess we can drop it. But lots of folks are going to copy them, so I think we should make sure it's "right".
... It should be both machine and human readable.
... Norm took an action to review the RDDL design and see what's actually deployed.

Norm: It's generally valuable if retrieving a resource returns a human readable document. Sometimes, particularly it seems in the case of XML namespace documents, there's also value if there is a machine readable representation available. RDDL offers a way forward at least for those cases where you can accept a level of indirection to get to the machine readable representations. But the RDDL 1.0 format wasn't seen as perfect, so some effort was invested in produci

ng a slightly nicer format. That effort took so long that RDDL 1.0 became widely deployed. A fallback position was to invoke GRDDL and say that there was a common model behind both formats. As long as the model was simple enough, we figured that'd be good enough. Then some folks objected the way in which the nature and purpose URIs were drafted and to their domain and/or range. I made a few attempts to address those concerns, but largely failed to do so. I don't s

ee a way forward.

<timbl_> 1+ to propose asking DanC to (a) write some RDF which would be useful and (b) fidn a way to use GRDDL to put that in the document and so we can look at the result.

DanC: I think I made some progress, but test cases would help.

<DanC> (a few at oasis and a few at microsoft would be helpful)

Henry: Yes, but they didn't use enough of the RDDL vocabulary to make a useful test case.

Stuart: It'd be nice to find tangible output from your earlier action.

<Zakim> timbl_, you wanted to propose asking DanC to (a) write some RDF which would be useful and (b) fidn a way to use GRDDL to put that in the document and so we can look at the

TimBL: With respect to #int, would a way forward be to ask DanC to write some RDF that would be appropriate and use GRDDL and make that RDF come out when you go to that page.

<DanC> [[

<DanC> fn:base-uri rdf:type owl:FunctionalProperty.

<DanC> ]]

<DanC> -- me, 21 Aug to www-tag

<timbl_> Where do I find bug #1964?

<timbl_> Not in http://www.w3.org/2001/tag/group/track/issues/open

Norm: I think the base-uri example comes from the XPath 2.0 Query Functions and Operators document.

<DanC> http://www.w3.org/2001/tag/group/track/actions/23

<DanC> http://www.w3.org/Bugs/Public/show_bug.cgi?id=1974

<DanC> [[

<DanC> Produced new version of http://www.w3.org/XML/Group/2005/12/XMLSchema.html

<DanC> for editors' consideration.

<DanC> ]]

<DanC> yup, http://www.w3.org/XML/Group/2005/12/XMLSchema.html#int is hypertext-happy

<DanC> (I recommend RDF, RDF Schema, and OWL)

<scribe> ACTION: ht to add GRDDL to that document so that RDF can be retreived. [recorded in http://www.w3.org/2001/tag/2007/09/06-minutes#action04]

<trackbot-ng> Created ACTION-45 - Add GRDDL to that document so that RDF can be retreived. [on Henry S. Thompson - due 2007-09-13].

Stuart: I don't think there'll be consensus to abandon namespaceDocument-8

<DanC> xsd:int rdf:type rdfs:DataType . # I'm pretty sure;

<ht> OK, I can do that much

<DanC> yes, rdfs:Datatype , documented in http://www.w3.org/TR/rdf-schema/#ch_datatype full uri: http://www.w3.org/2000/01/rdf-schema#Datatype

<timbl_> Two methods: one is to have XHTML document with GRDDL f th rDF; the other to have a RDF with some XSLT for the RDF.

Stuart: Part of the motivation for scheduling it today was to see if it was valuable to schedule it for the f2f.
... It seems like it would be.

DanC: I start to turn colors at specific examples of RDDL nature and RDDL purpose.

TimBL: I'm suggesting we just skip that bit and say that you just make the actual data you care about available.

<DanC> HT's draft uses this markup: <div class="resource"><rddl:resource xlink:arcrole="http://www.rddl.org/purposes#normative-reference"

TimBL: We right the actual OWL shemas for the datatypes we care about.

DanC: But Henry does have some RDDL properties that he claims to know about.

Henry: Yes, there was a genuine problem here that everytime Tim, Dan, and Norm have brianstormed on the whiteboard, Norm has been unable to take away anything he could write up.

DanC: Let's work with Henry for a moment. Henry, what MIME type for your namespace document?

Henry: I would expect application/xhtml+xml to most browsers and text/html to IE.

DanC: And in the text/html version, will you be able to get the data out?

Henry: No. How would I?

DanC: Will I see the characters "rddl:resource" in the text/html?

Henry: Yes.
... I only serve one representation.

<DanC> text/html <!DOCTYPE html>

DanC: Another answer is to just use text/html and stick a doctype at the top to get into standards mode.

Henry: But I care about the actual <!DOCTYPE


Dave: The TAG has gone on record saying that RDDL documents are a good thing.
... Ok, maybe we haven't said it formally.
... But there's been a certain intimation about approving it
... I think if we want to move away from RDDL, we need a strong case for doing so.

<ht> The relevant section of WebArch mentions RDDL favourably: http://www.w3.org/TR/webarch/#namespace-document

<DanC> nope. I don't accept any obligation to argue against RDDL. Until we're done, we're not done. I'm OK to review Henry's draft namespace document that uses RDDL in detail, but if the details don't work out, they don't work out.

Dave: I find it passing strange that RDDL works pretty well in practice, and the problem Norm has had is the mapping to RDF. So maybe the solution is to not use RDDL but to use RDF? That seems the wrong way around.

<Noah> It's also a bit disturbing that we're having such trouble connection the RDDL-> GRDDL -> RDF dots should anyone want to do that with existing RDDL. Even if it doesn't emerge as our first choice, it seems troubling that it's so hard to get straight.

Scribe struggles to keep up

Noah: Do some of these problems have to do with the legacy URIs that RDDL use?

<DanC> it's not troubling to me that we can't get it straight; we spend 95% of our time remembering what we forgot since last time. :-/

Norm: yes.

Noah: So we could throw all the old URIs out and start over.

<Noah> To be clear, I wanted to understand where the architetural barriers were coming from, and I think the answer is: from particular legacy RDDL nature or purpose URIs. I was NOT necessarily claiming that users would be happy if we threw them all out in the interest of architectural purity: on that latter point, I'd need to do some more thinking.

Henry: The sticking point was for many of the original role/arcroles

DanC: I think I've made suggestions for the ones I care about and I don't care about the others. Just chuck them.

Noah summarises his point entered above.

Stuart: There are a couple of other things on this agenda. I'll schedule time for this at the f2f.

<Noah> Right. What I just said on the phone was: I don't think the problem seems to be with RDDL as architected; I don't think the problem is the inability to invent a suitable OWL ontology; I think the problem is that some of the particular nature/purpose URI's already deployed are problematic, becuase they do not directly identify the nature or purpose, but rather some namespace (for example) used in documents of the nature.

<Noah> So, as we look for solutions, it's interesting to know that that's where the problem is.

<dorchard> +1 to Noah's point.

<dorchard> I find it interesting

Stuart: I'd like to stop there
... I'd like to skip items 6 and 7.

<dorchard> that speaks to a thesis that maybe RDDL is still good to use

Stuart: What about the plenary

Dave: It looks like distributed extensibility is on the short list for the plenary

<DanC> oh crud... I did a bunch of prep for item 6 XMLVersioning-41 (ISSUE-41); will DaveO and Henry be here in a week to talk about it?

<timbl_> And I wanted to say that we cou;d evne speaialcase existing URIs used with RDDL as ahcak, but require that people use sometjinhg consistemnt in the future.

Dave: Maybe we could get some TAG members and some folks with other opinions up on the stage to make it lively.
... I need to draft the proposal and I'll be working on that.

Stuart: There's also a question of scheduling the f2f for work and recreation.

<DanC> meeting is scheduled to end 15:00 on Wed

Summary of Action Items

[NEW] ACTION: David to post his message to the HTML WG list [recorded in http://www.w3.org/2001/tag/2007/09/06-minutes#action02]
[NEW] ACTION: ht to add GRDDL to that document so that RDF can be retreived. [recorded in http://www.w3.org/2001/tag/2007/09/06-minutes#action04]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2007/09/06 18:55:56 $