W3C Technical Architecture Group

31 Jan 2008


See also: IRC log


Stuart, Norm, Jonathan, Tim, Ashok, Dan, Noah
Dave, Raman


Agenda review

Stuart: Pretty much as published, with a little reordering and a new item from Henry

Agenda accepted

Accept minutes of 17 Jan 2008


Next telcon: 7 February 2008

Proposed to scribe: Dave

Stuart to chair

Possible regrets from Ashok for 7 Feb; Tim for 21 Feb.

Welcome to new members

Stuart: Welcome in a more formal way to Ashok and Jonathan. Also congratulations and welcome back to Henry and Raman.
... Perhaps we could do a bit of a round table.

Dan: I co-chair the HTML WG, occupying about 150% of my brain. Tag soup integration is always on my mind. Also IETF liason so mime-type issues always pique my interest. I'm interested in the Namespace Document 8 and sem-web related issues.

Henry: I have three documents on the critical path: Namespace Document 8, which is close, XML Functions 34, URNsAndRegistries 50. Otherwise known as why all schemes other than http: are evil.

<DanC_lap> (I forgot to say: I'm interested in learning about information theory and economics, since large-scale considerations often dominate semicolon-vs-comma level design decisions, even in HTML)

Henry: I'd like to spend more time on the vocabulary work currently going on in the sem web subgroup.
... we could do better making it clear about what URIs are and what resources are, etc.

Jonathan: I'm at Science Commons and from that PoV we have a strong interest in the semantic web and identifier schemes and document metadata.

Noah: I'm not sure how much introduction is needed, I know Ashok and Jonathan a bit. I'm no longer on the Protocol WG. I am still involved in XML Schema.
... I can't say I have a technical hot button, I just think the web is really important and at its best the TAG has an opportunity to explain things that are subtle.
... We can also promote clear thinking.
... The web is something like a telephone system, it has to keep working in 30 or 50 years.
... I'm wrapping up a draft on the self-describing web, which doesn't have an issue.
... I tried to take a crack at the relationship between schemes and protocols, but I've put that down for a bit.

Stuart: I've been co-chairing for a while. My strong interests are in the semantic web. I can't seem to leave issues related to identifiers alone. I find some of the ontology aspects really absorbing and hard.

Tim: Generally the semantic web. I think it's great that we have a subgroup doing semantic web architecture.
... We need to be able to write these things in RDF and describe relationships between them.
... My current 'tabulator' project makes some of these issues urgent for me.
... All sorts of other things hit me at glancing angles: versioning in HTML and XML.
... There have been discussions, for example, about XML being upgraded. That's an example of one of the many times we've messed up versioning. We've got a lot of material thanks to Dave but we haven't boiled it down to truths.

Ashok: I started on Schema in 1999. I worked with Noah and Henry on it for many years. I also did XML Query where I worked with Norm. Most recently, I've been doing WS-Policy where I'm working with Dave.
... Now I'm focussed mainly on web services. I've been doing lots of OASIS work on web services: WS-Policy, etc. The other thing I'm trying to start is an incubator group to map relational data to RDF and OWL.
... That's taken a little while to get started, but once it starts, I think the TAG might have some wisdom to offer.

Norm: I'm co-chair of the XML Core WG and chair of the XML Processing Model WG so XML issues are always on my mind. I'm interested in the tag-soup nexus of issues. I'm interested in issues related to URIs and resources and the semantic web as well.

2008 f2f schedule

Staurt: I'd like to make a formal decision about the two meetings following Vancouver.

Stuart: There's been a WBS poll for a while now. The September proposal is pretty strong.
... For Bristol, we are at risk for not having TV, Dave, and Dan for some or all of that meeting.

Dan: The risk for me is a semweb conference on the west coast that looks really cool, but I guess I could miss it.

Dave: Monday is a public holiday in CA, so we're likely to have plans, though we don't have any yet.

Stuart: Does anyone have reservations about us meeting w/o those participants.

Henry: Given how hard we've tried to find another date without success, I think we should go ahead.

Dan: My risk is negligible, let's ignore it.

Norm: I'm with Henry, it may not be ideal, but we can't find anything better.

Stuart: I propose that we adopt those two sets of dates.

Dave abstains, no objections.


<timbl> * Spring: 19th-21st May 2008 (Mon-Wed), Bristol UK, hosted by HP Labs, Bristol (Stuart)

<timbl> * Summer: 23rd-25th September 2008 (Tue-Thu), Kansas City, USA, hosted by W3C (DanC)

RESOLUTION: The TAG will meet 19-21 May in Bristol and 23-25 September in Kansas City

Issue tagSoupIntegration-54 and contentTypeOverride-24

Stuart: Noah posted a note about the use of META tags to trigger standards-compliant rendering in browsers

<Noah> http://blogs.msdn.com/ie/archive/2008/01/21/compatibility-and-ie8.aspx

Noah summarizes his message and how he came to discover this topic.

Noah: Roughly what's going on is that users got dependent on how older versions of IE rendered pages.
... But there is a desire to move forward. Some versions keyed off the presence of the DOCTYPE declaration.
... For a combination of reasons, they feel that's no longer working. If they did the same thing in IE8, it would break a lot of content tailored for IE7 and IE6.
... The proposal that's been floated is to use a new http-equive meta tag.
... I think the spin on that is that a site-wide HTTP header can set a global optoin.
... If you don't use the meta tag, you get quirky interpretation. If you do use the meta tag, then you identify the level of IE that you believe is best for your content.
... I have at least two concerns: the first is whether this is in any way, shape or form a good idea. The other is, what happens to follow your nose.
... I don't think it woudl break webarch at that level if (scribe: iff?) the HTML spec says something about that meta tag.
... Without that in the HTML spec, I'm not sure it's legitimate at all.

Dave: I think this is a great thing to discuss. This is effectively a kind of browser sniffing as TV pointed out.
... I guess there's a bunch of different aspects that are ... interesting.
... One is that if there's a version attribute, it'll be the *browser* version.
... Then there's where it's going to be, in the meta tag instead of a version attribute on the HTML tag or as a parameter on the media type.
... Then there's the fact that the default is going to be IE7 mode. The expectation is that a lot of people are going to forget to do this, so they'll be frozen indefinitely at IE7.
... Then there's the question of whether or not anything can actually be done about this.

<Noah> Norm: I don't think this is a great solution.

Norm: I appreciate that there are some hard problems here, but I think the proposed solution is awful.

<Zakim> DanC_lap, you wanted to think about economics and information theory of the http header

<timbl> How about an HTTP spec where you can quote the tracker URI of a bug you require?

Dan: David Barren gave a pretty coherent argument about the economics of putting the version identifier inside the document.

<timbl> So we have a tag for "Best viewed by" at last .. sigh.

Dan: If the HTML WG decided that this was the right thing to do then, Firefox version 12 would contain versions 11, 10, 9, etc.
... This is only practical for the guys with the biggest guns.
... I found this pretty compelling argument against a version attribute in the language

<timbl> Maybe the HTML spec should give a set of "Best viewed with" which are automatically inserted when this ttribute is found.

DanC: On the other hand, having the version inside or outside the document is important.
... The spec documents say I send you a request, you send a document.
... In practice, you send me some bytes and you expect those to be interpreted according to the dominant browser at the time.
... So if you want your document to be interpreted per the specification, you're in the minority.
... It makes sense from an economic sense that the minority should pay a few more bits.
... If we get to the story where the deployed software obeys the specs, then you can throw away the HTTP header.

<Zakim> ht, you wanted to query dan

Henry: I don't understand how what you just said renders less signficant Dave Baron's observation.
... I thought you were going to say that if you move it into the HTTP header, then you can just launch the right browser.
... But then I thought you said it worked equally well inside or outside and that doesn't work for me.

DanC: What I mean is that if you have a version flag that can be used in either place, you can have a marketplace where some browsers ignore the flag and just go as close to the specs as they can
... and other browsers obey it and the web gets better over time.

Henry: I don't see the connection with inside or outside

DanC: If it's outside, then the document doesn't have to change as the browsers evolve.

Some more discussion

DanC: I'm not interested in supporting users who write code for a specific browser.
... MS can't ship a browser that obeys the standards because it won't get uptake.

<Zakim> Stuart, you wanted to ask folks how we feel abouts a situation where we have to deal with versions of interpretation/implmentations rather than the spec.

Stuart: We're now in a situation where we're concerned about the interpretation of a particular version of a spec. That seems weird.

<ht> HST wonders how serious the pushback was to the IE7 move which sparked this

<Zakim> Noah, you wanted to ask about range of user agents

DanC: Everything is weird about the HTML space. It's about economics and biology more than computer science.

Noah: The rule of least power encourages users to write content that is idependent of particular user agents. That's a good thing when you can get ther.
... The simplest HTML is sort of like that. There are headers and paragraphs, and exactly how that's interpreted is up to the UA.
... Certain kinds of commercial work demanded greater fidelity.
... When you see this meta thing, if we could say that the core abstractions were the same, but that the meta would promise that corners on tables wouldn't be rounded, that'd be one thing.
... But I don't see any bound on it. I'd love to see a stake in the ground that says "here are the things you can't change in the meta tag".
... As long as I stick to certain things, I'll know that everyone is going to interpret it the same. If I go beyond that, to CSS corners or broken markup, then maybe the meta value will matter.

Stuart: Increasingly with subscription environments, the question is less about what pixels go on the screen and more about what DOM gets built.

<Zakim> Stuart, you wanted to stay that it goes way beyond screen rendering

Noah: The punchline for me is, when I see a meta tag, are all bets off or is there some level of functaionlity that I can rely on.

DanC: The hardest part about this stuff is that you don't find out what the tokens mean until well after they're issued. The browsers see the "Mozilla" token and so they send CSS. So IE sends the Mozilla token. And then some labels become labels for sets of bugs.
... What the label stands for is really hard to figure out in advance.
... Another kind of code tries functions and based on return values makes decisions about functions it can actually use.
... Consider the GNU autoconf stuff. It starts with now information and probes for various things.

<jar> danc, i think you meant 'autoconf'

Stuart: Is there more to be said now?

Dave: I wonder how this relates to our work on the versioning finding. I haven't really thought that through.

TimBL: This would definitely be a good story.
... So would the XML 1.0 5e story.

Stuart: I don't see a particular action to leave dangling here.

Issue passwordsInTheClear-52

Stuart: Dave had an action to publish it and solicit comments.

Dave: The edits that I did were slightly more than I was asked to do. Because I picked up the ball recently, I wanted to make sure that the group was happy with my changes.
... I had hoped to get a diff out. Norm offers to diff them.

Norm: Diff 20071124 with 20080124?

Dave: yes.

<scribe> ACTION: Walsh to create a diff of passwordsInTheClear [recorded in http://www.w3.org/2008/01/31-tagmem-minutes.html#action02]

<trackbot-ng> Created ACTION-97 - Create a diff of passwordsInTheClear [on Norman Walsh - due 2008-02-07].

<ht> http://www.w3.org/2007/10/htmldiff?doc1=http%3A%2F%2Fwww.w3.org%2F2001%2Ftag%2Fdoc%2FpasswordsInTheClear-52-20071108.html&doc2=http%3A%2F%2Fwww.w3.org%2F2001%2Ftag%2Fdoc%2FpasswordsInTheClear-52-20080124.html

Dave: I'll listen until Wednesday and send something out if no one objects.

Stuart: Ok, that's what we'll do then.


Henry: It turns out that the XRI TC has published a Committee Specification for XRI resolution 2.0.
... The comment period closes tomorrow.

<DanC_lap> (ends tomorrow? when did it start? ah... 2 Dec. hmm... who is our oasis liaison, I wonder...)

<ht> http://www.ltg.ed.ac.uk/~ht/xri_notes.html

Henry: This is what I wrote on the basis that it's been a long time since we talked about it.

<jar> do w3c and oasis coordinate?

<DanC_lap> (the main place where XRI shows up on my radar lately is near OpenID)

<DanC_lap> (oasis liaison is Karl, says http://www.w3.org/2001/11/StdLiaison#OASIS )

Henry: For reasons I have to say I don't understand, they've gotten themselves written into OpenID 2.0.
... Implementing OpenID 2.0 mandates implementing XRI.

Noah: Can you explain that?

<dorchard> I had understood that it was optionally in open id.

Henry: You have to be able to decode XRIs and implement the authority lookup protocol in order to find out what the OpenID is.

Danc: Folks are saying http:// is too ugly, let's have =danc instead. And then people ask about email addresses. The subtext is "oh, no, no, no, we want to be able to collect money when people invent these"
... I've heard that one of the reasons the OpenID folks didn't go to the IETF is because the IETF would expose this.

Henry: It's very hard to find the current, relevant bits. Lots of stuff on the web is old.
... I was told I could register =henry, =henrythompson, and @ibm!
... I

<jar> dns costs money too... ??

Henry: I'd like to talk about this more, but the fundamental architectural proposal behind this is to introduce a mandatory level of indirection into all addressing.
... The core operation you can do is to retreive metadata about a resource.

DanC: So the design mandates an extra round trip on the network. That's the number one thing to avoid in a protocol.
... I'm happy to say that on behalf of the TAG by tomorrow.

Henry: That will put a stake in the ground, but it's fundamental to their design.
... at the end of the docment I pointed to earlier, you'll see a list of the services you can get on an XRI
... With the right bits, the redirection would have been automatic.

Tim: If we haven't said it strongly enough, we should say again and again that conneg should only be used for two different representations of teh same thing.

Henry: Yes, that seems to be broken here too

Stuart: I've heard two things, one on the content negotiation, and one on the mandatory round trip.

Henry: I don't fully undertand all the dimensions of this yet. There's a distinction between URIs with and without service identifiers, for example.

Stuart: It is possible to express concerns in a general way and ask for more time?

DanC: We can also ask them as questions in the meantime.

Tim: Can't we do both?
... Lodge the complaints we really have and ask for clarification elsewhere.

TimBL: Isn't the privately owned naming scheme a problem to OASIS?

Henry: I don't see how we can make that argument given taht you have to pay someone to get a DNS name.

Stuart: Another possible technical question, XRI has been injected into OpenID, does that mean that XRI URIs are special in OpenID. So you're not treating URIs in a general way.

Henry: On the wiki's and things, they use XRIs so the agents do have to be able to recognize and interpret them.

Noah: Is the lack of URI syntactic compatibility another issue? Let's say that XRIs happen, can I put them in the same slots where URIs can go or is that another issue?
... Do you really always know that when you want an XIR you don't want any other kind of URI or vice-versa?

<ht> Something like "Do we understand that XRIs _without the xri:// part_ must be recognised as alternatives to http: URIs for OpenID2.0 implementations?"

Stuart looks for volunteers to submit these comments

Henry says he's draft it now

<Noah> I think it's more than Open ID. "To what extent is it expected that there will be use cases in which a choice of URI or XRIs without the explicit scheme name to be allowed in, for example, the same attribute value or input field? If so, then how are the syntaxes to be coordinated to avoid collision?

<ht> "Is it a consequence of the spec., as it appears to us to be, that a) All access to resources identifies by XRIs requires (at least) two round trips and b) that content negotiation is used to return metadata or resource representations?"

Some discussion of the optional nature of the xri: part of the URIs.

<scribe> ACTION: Noah to craft comments and send them on our behalf. [recorded in http://www.w3.org/2008/01/31-tagmem-minutes.html#action03]

<trackbot-ng> Created ACTION-98 - Craft comments and send them on our behalf. [on Noah Mendelsohn - due 2008-02-01].

Vancouver F2F Agenda Requests

Noah: As promised, I'm mighty close to a new draft on self describing web.

Noah: I'd like that on the agenda.
... I've also been thinking about http-range and 303 and that might be ready in time.

Stuart: I think namespaceDocument-8 is really on the brink of closure, we should try to get that closed.

Norm: I'd like to see xmlFunctions-34 on the agenda.

Stuart: Dave's not here so I can't ask about logistics.
... Are folks generally happy with the logistics?

For the meeting, yes.


Summary of Action Items

[NEW] ACTION: Noah to craft comments and send them on our behalf. [recorded in http://www.w3.org/2008/01/31-tagmem-minutes.html#action03]
[NEW] ACTION: Walsh to create a diff of passwordsInTheClear [recorded in http://www.w3.org/2008/01/31-tagmem-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2008/01/31 19:36:56 $