W3C Technical Architecture Group Teleconference of 4 Dec. 2008

04 Dec 2008



Dan Connolly, T.V. Raman, Stuart Williams, Jonathan Rees, Noah Mendelsohn, Ashok Malhotra, Henry Thompson, David Orchard
Tim Berners-Lee, Norm Walsh
Stuart Williams
Noah Mendelsohn


Approval of last week's minutes

<DanC> to wit http://www.w3.org/2001/tag/2008/11/20-tagmem-minutes 2008/12/04 16:55:52

RESOLUTION: The minutes of the TAG telcon of 20 Nov 2008 at http://www.w3.org/2001/tag/2008/11/20-tagmem-minutes are approved

Scheduling of next meeting

SW: Next telcon will be on 18 Dec. 2008, our last before the holidays

<DanC> 18 Dec ok for me

AM: I can scribe

SW: Any regrets for that?


Issue httpRedirections-57

<DanC> action-178?

<trackbot> ACTION-178 -- Jonathan Rees to prepare initial draft of finding on uniform access to metadata. -- due 2008-11-25 -- PENDINGREVIEW

<trackbot> http://www.w3.org/2001/tag/group/track/actions/178

SW: Jonathan has made some progress on actions 178 and 184.

<DanC> . http://www.w3.org/2001/tag/doc/more-uniform-access.html

JR: On 178 I've prepared something and it's linked from the agenda. See http://www.w3.org/2001/tag/doc/more-uniform-access.html . Not sure this discharges my action but I think it's worth discussing.

<DanC> (I see a bunch of whitespace at the bottom; browser bug?)

SW: As to the action, I'll leave it as PENDING REVIEW. Should we discuss now?

JR: We can, but I'm not sure people have had time to read it.

SW: We'll discuss at the F2F, and will discuss reviewers later.

<DanC> action-184?

<trackbot> ACTION-184 -- Jonathan Rees to write to Lisa D of IESG, cc www-tag, to explain about 303, with cool URIs and webarch as references. -- due 2008-11-25 -- PENDINGREVIEW

<trackbot> http://www.w3.org/2001/tag/group/track/actions/184

JR: On action 184 I sent mail to Lisa and have not yet received a reply. This has to do with the link header draft that he's circulated.

DC: There are URI's for relationships. If those go to an IANA Web server that responds with 200's, then that claims they are information resources.

<DanC> I'd like the action to be "contact" rather than "write to"

JR: I'm not sure I've followed all of the discussion on link header. Perhaps the next step is to contact Mark Nottingham?

<DanC> action-184 due next week

<trackbot> ACTION-184 contact Lisa D of IESG, cc www-tag, to explain about 303, with cool URIs and webarch as references. due date now next week

SW: I'll leave the action OPEN. Perhaps you would update the due date?

JR: Sure.

SW: There are some things cross posted from the IETF working group and www-tag, but more is on the IETF list. I think the draft is a bit vague on whether relations are between representations and resources our between resources and resources.

<Zakim> DanC, you wanted to sugges that the way to continue an action is to move it back to open and move the due date; by default, just say: action-nnn due next week

DC: I'm not sure that, on the draft, PENDING REVIEW is the right status.

DO: Joining the call now.

SW: More discussion on this topic?

HT: First of all, my apologies for arriving late, I'm here now. I like some of what Jonathan has written. Are we ready to discuss today?

SW: We can.

<DanC> ht, until you got here, noone had read jar's draft of 25 Nov

<Ashok> I have read it!

<ht> http://tools.ietf.org/id/draft-nottingham-http-link-header-03.txt

<DanC> note -03 is recent: December 1, 2008

<ht> Note that link-values may contain multiple relations; for example

<ht> Link: <http://example.org/>; rel="index start";

<ht> rel="http://example.net/relation/other";

<ht> rev=copyright

<ht> Here, the link "http://example.org/" has outbound links of the types

<ht> "http://www.iana.org/assignments/relation/index",

<DanC> sorry to presume, Ashok

<ht> "http://www.iana.org/assignments/relation/start", and

<ht> "http://example.net/relation/other", as well as an inbound link of

<ht> type "http://www.iana.org/assignments/relation/copyright".

HT: OK, a question then. In Mark Nottingham's document at http://tools.ietf.org/id/draft-nottingham-http-link-header-03.txt is quite anodyne, though doesn't clearly distinguish resources and representations as clearly as we might like. From the draft:

   Note that link-values may contain multiple relations; for example

       Link: <http://example.org/>; rel="index start";

   Here, the link "http://example.org/" has outbound links of the types
   "http://www.iana.org/assignments/relation/start", and
   "http://example.net/relation/other", as well as an inbound link of
   type "http://www.iana.org/assignments/relation/copyright".

HT: Does he really means that links have links?

TVR: I think he meant target.

<Stuart> FYI... my participation in the ietf thread begins at: http://lists.w3.org/Archives/Public/ietf-http-wg/2008OctDec/0294.html

<DanC> ("outbound" seems wrong/backward to me too; the link is from the resource in the HTTP transaction to <http://example.org/> )

HT: That doesn't make sense either. I think he's talking about the link. Having a clearer ontology would be very helpful. I think he means that he means this syntax to signal four links, 3 outbound and 1 inbound.

JR: I suspect Mark might accept that.

SW: Henry, did that get to the point?

HT: Yes, though with a residual issue. I have not been a participant in this discussion, and am not sure whether I should respond or ask Stuart to respond.

TVR: He cross posted to www-tag, and you can respond there.

SW: Yes, though note that the discussion has forked and more is on the IETF branch.

HT: I'm happy with responding on www-tag.

<ht> http://www.hueniverse.com/hueniverse/2008/09/discovery-and-h.html

HT: I read http://www.hueniverse.com/hueniverse/2008/09/discovery-and-h.html and was at least somewhat convinced by it. I might question Jonathan's formulation of "use link headers, but for efficiency do this on request", perhaps changing to work with dynamic resource mapping (scribe isn't familiar with this) falling back to link headers when that doesn't work.

JR: A Google group has been set up and discussion is progressing of things like site metadata. I don't have anything written. We could discuss at the F2F.

SW: We have a F2F session scheduled. The session on this will be Wed. 11-12:30 at the F2F.

DO: I want to participate, please.

<DanC> (http://www.w3.org/2001/tag/2008/12/09-f2f-agenda#uniformAccessToMetadata says 11:15; skw acks that as a bug)

SW: Have we recruited you as a reviewer, Dave?

DO: (Laughter). Uh, sure.

JR: Answering Henry. My hesitation is that more apparatus is needed to make a URI -> AboutTheThingURI mapping work. It's being worked on by Mark, et. al. Lines of communication are quite good, and better than a few months ago. Had I put link header ahead of URI rewrite consciously, that would have been my reasons.

HT: Understood.

<Zakim> ht, you wanted to ask a Link header question

HT: It will have an impact on update too.

JR: Yes.

AM: Wanted to ask Jonathan about Google group, but update next week will do it.

<DanC> # Metadata discovery David Orchard (Friday, 21 November)

<DanC> http://lists.w3.org/Archives/Public/www-tag/2008Nov/0103.html


The Google group is at: http://groups.google.com/group/metadata-discovery/browse_thread/thread/d34e7fd9c9387a97

TVR: The Google group will have an ATOM feed and an RSS feed.

JR: Not sure site metadata in particular is being discussed on the Google group, but there are probably pointers.

Issue uriBasedPackageAccess-61

<DanC> close action-196

<trackbot> ACTION-196 Notify webapps wg that TAG is unlikely to meet the 3 month timeframe and ask for notification when tests relevant to widget URIs become available closed

SW: I have fulfilled my action to warn the Web applications group that we will not make our 3 months target, and to ask that we be notified of test cases pertinent to widget URIs. Both are done, and they have established their own action regarding notifying us of such test cases. I've done some more thinking about this, and I feel that the right extension point is media types and fragid syntax. I tried to work that up in the message athttp://lists.w3.org/Archives/Public/www-tag/2008Nov/0119 . Who has looked at it?

<DanC> (I read that sketch only once, but not carefully.)

AM: Looked reasonable to me. You do a good job of using a hierarchical scheme to address a hierarchical structure. You got pushback, and I'm not sure why you did. Do you understand the concern?

SW: I think the pushback, mainly from Marcos was to the effect that they are not interested in network retrieval of packages. The widgets are in some way installed in the Widget engine.

<jar286> they could use tag: URIs if network reference isn't an issue

<Stuart> ?

AM: They have a different and more limited use case, right?

SW: They need to resolve relative URIs within the structure into absolute URIs.

<Ashok> Marcos Cacares

DO: I'm starting to understand what they're looking at. Perhaps we should look for general solutions. Architecture astronautics is a risk. Stuart, I'm glad you followed through. I think it's worth looking at whether we can do something that will bring communities together.

<Zakim> noah, you wanted to ask why use URIs at all?

<ht> scribenick: ht

NM: I'm just starting to pay attention to this. Has anyone said why they need URIs at all? My understanding from this call is that they don't need to get outside their closed world Google/Firefox use funny URIs for their local config data is because they want them to appear in URI contexts. So I'm missing a careful statement of requirements which among other things makes clear why these things need to be 'on the Web'. Sure, general is good, but only if there's prospective synergy.

<noah> DC: The Web app groups needs URIs because they need to use DOM APIs.

DC: They have explained -- WebApps need URIs, because their system context is the DOM API, e.g. <img src="...."/>, where that src can be a reference within the widget.

<scribe> scribenick: noah

<Stuart> http://dev.w3.org/2006/waf/widgets-reqs/#r36.- probably is the requirement that applies

TVR: So if not going to the network, they can use widget:

<DanC> does r36 refer to the DOM API? double-checking...

DC: I'm not sure Web Apps requirements are shared by anyone else.

SW: They almost need a skolemized URI scheme.

<noah_> Yes, exactly re: skolemized.

DC: I got conflicting input on that at the F2F. Some seemed to imply these never escape from the local environment, some seemed to imply a need for cross site linking.

SW: They did ask the TAG to look at their requirements, which is why I put it on our agenda.

<Zakim> DanC, you wanted to answer why use URIs at all

<Zakim> DanC2, you wanted to note http://lists.w3.org/Archives/Public/public-pkg-uri-scheme/

DC: There is a new mailing list, Marcos asked for it, and Mike Smith gave it to him. http://lists.w3.org/Archives/Public/public-pkg-uri-scheme/
... Stuart, would you like to send your proposal there?

SW: Yes, I will. Will suggest further discussion on that list.

<Zakim> ht, you wanted to ask why normal base URI rules don't do the job

<DanC> (worth an action?)

HT: I tried during the week to understand why Larry's points weren't well taken. If there's a package, there's a base URI. Why isn't that the obvious answer?

SW: Let's say the package is installed in a widget agent in a user agent.

DC: In the meeting they gave the case of two clock widgets, and you want one to be set to analog and one digital. They allow script writers to associate preferences with the name of the widget. If they have the same name, you can't set the properties on the two clocks separately.

HT: Seems to me we need to frame this as a what's the base URI problem.

NM: Yes, and if we're using fragids, what's the media type.

<Stuart> yes... indeed "...a what's the base URI problem."

TVR: I agree with Noah to an extent, though if you say the URI is the HTTP that's not always enough given what's happened with text/html and scripting.

<Stuart> +1 to raman scheme/prot and media-type/frag are orthogonal

DC: Back to some of Henry's question. Some might say that the base URI is the file where I found it...

TVR: That isn't good enough for the clock case, as we've seen with Google gadgets.

DC: Some of the implementors did follow the file: path, and it turned out to be a privacy concern by giving script writers access to names in the filesystem.

<ht> Stuart, yes, so this now sounds like what we know in the Schema WG as the 'left hand side' problem, for Schema Component Designators -- the components are distinct from the document fragment which gave birth to them

DC: One of the requirements, I believe, was to get away from file:.

<ht> I hear the words "install-time identifiers" as useful

<Zakim> noah, you wanted to say you can't squat on scheme space even if you don't expect it to escape

DC: Another angle was to leave this unspecified, but the worry was that some implementors wouldn't notice the concern and use file: anyway.

<ht> scribenick: ht

<DanC> 2 requirements I'm not sure they've written up clearly: (1) install-time naming to give distinct names to the 2 clock widgets (2) prevent leaks of the form that file: experience shows

NM: TVR said that as long as we knew these wouldn't escape to the network, a widget: scheme would be OK, But I don't agree, in that this amounts to squatting on scheme name space -- if someone registers a public widget URI scheme, no widget user could get at them, because they already have a private, conflicting, meaning for a scheme of that name.

<DanC> (is chrome: registered? yes, in the esw registry... http://esw.w3.org/topic/UriSchemes/chrome )

TVR: So, we should use random digit strings?

NM: Well, some of these (chrome?) do get typed by people

<DanC> (I asked whether widget: was going to get written in files and such, and the answer I got was "no." then somebody else said "well, maybe")

<Zakim> DanC, you wanted to note file: experience and arguments against it

NM: So I think the right thing to do is register these things, even if they are not expected to escape

TVR: agree

<scribe> scribenick: noah

SW: I think we will come back to this at the F2F, focussing especially on their direct request that we consider addressing requirements.

<DanC> "To develop a response to a direct request for feedback on widget addressing requirements" -- http://www.w3.org/2001/tag/2008/12/09-f2f-agenda#uriBasedPackageAccess-61

<DanC> where "direct request" is linked

Agenda for the F2F

Draft agenda for the F2F next week is at: http://www.w3.org/2001/tag/2008/12/09-f2f-agenda

<DanC> "2-4: Will review inputs on the session reading list"

<DanC> -- http://www.w3.org/2002/09/wbs/34270/F2FPriorities200812/results

NM: How did I wind up on URNS&Registries?

SW: You voted for it.

NM: Yes, because I think it belongs on the agenda.

<DanC> action-183?

<trackbot> ACTION-183 -- David Orchard to incorporate formalism into versioning compatibility strategies -- due 2008-12-05 -- OPEN

<trackbot> http://www.w3.org/2001/tag/group/track/actions/183

SW: We are waiting for versioning input.

<DanC> (one computer is not enough. I keep 2 in working condition and 95% of my work on the net)

<DanC> ("redundancy in all critical systems")

DO: My Mac drive crashed. Just got it back. Will try and get finding out this afternoon.

SW: I'm concerned about the nature of the discussion we should have. One option: press on reviewing document as we have been doing, but I sense fatigue setting in for other TAG members. Or, we could step back, and ask how to conclude in a way that leads to a result. Do we want the meta discussion of minimum to declare victory?

<Stuart> ?

DO: A possibility would be to proceed to cut things down, perhaps to just forwards-compatible versioning. If there next few F2F's are West Coast, I could possible travel to them. Another possibility would be to get as far as we can, and then volunteer to work past my term, as I did in the past. Yet another possibility would be to add an editor.

SW: I've left space for up to 2 sessions. I would like some guidance from the TAG as to the direction they'd like to go.
... What are your thoughts?

HT: The only thing that's sparked my interest is the formal approach. That's not on the table.

DC: Why not?

HT: There is no current proposal for a future for that?

SW: There is an action to incorporate it.

HT: Not what I said. The document I find exciting. Cramming it into another document isn't as exciting.

<Ashok> I agree that would make a better finding

NM: What Jonathan has done is, if I speak honestly, the simple clean framework I had in mind from the start. If we went down that path, I'd want to also tell some stories with examples, etc. for people who can't handle the math directly. That said, I would really like to get value out of the hard work already done on Dave's draft; if there's a slice that's nearly ready to go, I'd be glad to wrap that up and ship.

TVR: I suggest publishing what we have, not necessarily as a finding, but as background that we can use if we start with Jonathan's. The passwords in the clear finding was much simpler, and it took us 2 years after Ed left. I'm very concerned doing without active editor.

NM: I don't think we should back into making versioning a priority in the coming year. It will be one of the 3 or so big things we do if we move ahead with it. If it makes the cut as one of the 3 most important, then I'm excited about doing it. If not, we should put it down, and publish Jonathan's as background work too.

<DanC> ftf agenda says "due 2nd Dec 2008"; should read "due 5 Dec"

<noah_> I would like to be dropped as a reviewer for URNSandRegistries. Never meant to volunteer.

<DanC> likewise for urns and registries, ftf agenda says "due 2nd Dec 2008" should say 5 Dec

HT: The document I circulated today is not formally for F2F reading.

NM: I would like to be dropped as a reviewer on this one.


JR: I think uniform access to metadata is set.

AM: What do we do here? Publish a TAG finding? Start a dialog.

SW: At least we can be involved in the community. We can publish in varying forms. You were one of people who asked for discussion. Where would you go with this?

<DanC> (I'm decreasingly excited about tag findings, but the TAG agreed to address issue-57 ... hmm.. is the "HttpRedirections" name passe?)

<DanC> (workshops are not out of the question)

AM: I have argued that this is central to the Web. I want to get the right people together to work on this, but am not sure best how to do it, except maybe something like a workshop. Not sure whether that's practical.

JR: Maybe we can add as a F2F goal to bring people up to date on what's happening in the community. I'm very optimistic that people like Eran are working well together on this.

<Zakim> DanC, you wanted to say I'm happy for AM and other TAG members to be creative about how the TAG addresses issues; findings are somewhat traditional but not in any way the only

DC: Findings are traditional, but not the only choice. On Javascript security. I have not yet read, but am hoping to read, mashing with permission.

SW: URI based package access I put on. Review material is linking into specific widget requirements. You may want to look at surrounding context too.

<DanC> said email doesn't seem to be cited

<DanC> action-194?

<trackbot> ACTION-194 -- Noah Mendelsohn to revise Self-Describing Web Draft Finding in response to September 2008 F2F meeting -- due 2008-12-05 -- PENDINGREVIEW

<trackbot> http://www.w3.org/2001/tag/group/track/actions/194

NM: New self describing Web draft is out at http://www.w3.org/2001/tag/doc/selfDescribingDocuments-2008-12-03.html

<DanC> (noah, stuart and I sometimes leave actions pending review until we have some *other* action on that issue)

SW: Tag soup got a lot of support as a discussion topic, but nobody has suggested a structure for the discussion.
... Meeting adjourned

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.134 (CVS log)
$Date: 2008/12/05 15:14:34 $