TAG 21 June 2005 telcon

21 Jun 2005

See also: IRC log


ER, NDW, HT, VQ, DO, NM_(partial)


I get forbidden for http://www.w3.org/2001/tag/2005/06/21-agenda.html

But the ,access* trick tells me it's world readable.

<Ed> worked for me :(

date: 21 June 2005

<scribe> scribe: NW

<scribe> scribenick: Norm

absent: Roy

Telcon: 28 June?

Possible regrets: NW

Next telcon: 28 June 2005

Scribe will be Ed

Approving minutes of the f2f

HT asks for time to review them

NW concurs

We'll approve the f2f minutes next week

Technical discussion: TAG future directions

DC published summary at http://lists.w3.org/Archives/Public/www-archive/2005Jun/att-0010/tag-directions.html

subtopic: Grid Services

<ht> HST would be happy to discuss his message to XML Schema about naming things if we have time

VQ: What seems interesting architecturally is the interaction of grid services with both web services and the semantic web

ht: My feeling is that there is an impact, but I don't know what it is
... I'd be happy to get educated in this area "on TAG time" as it were, if we could identify someone who has some reasonable understanding of the architecture of the grid and might be prepared to suggest some things we ought to look at

VQ: agreement

NW: Sun has some grid services work going; I can look for someone to come chat with us

ht: I have connections with Yaron Goland that I'll be giving a talk to about naming in ten days time. I'll look for interest there.

<ht> UK National e-Science Centre

dorchard: I can do a limited job, or one of our distribued web management folks, there are a couple in our standards teams, that would be glad to talk about some of the work BEA has been lookin gat in this area

VQ: any other proposals?

VQ asks DO to give a brief intro

dorchard: What we've been looking at is primarly an issue of management of resources. In that regard, there are questions of how do you manage resources, detect change in state, change state, subscribe to resource state changes, how does notification occur. All in this area of managability of resources.

<ht> HST understood the GRID to be something quite different from what DaveO is talking about

dorchard: There's roughly two competing camps: at OASIS there are three TCs: WS-Notification: Pub/Sub, Brokered notification, and event descriptions; WS Resource Framework which has introduced a number of SOAP/WSDL based specifications to describe operations that a resource will support (e.g., query by qname, create, delete, etc.);

VQ: This sounds very similar to web services.
... Are you aware of the vision of the grid that is more about distributed computing, where resources are more like computing resources, CPU power, or data storage

<ht> Global Grid Forum

dorchard: The term "grid" has many different meanings. I'm describing, briefly, one of those meanings. The output of the Global Grid Forum (Forum?) ended up going this way.

DO: What I think we're going to have here are a few different views of what the grid and grid services mean

<ht> I thought they had their own as-it-were rec-track. . .

<ht> http://www.ggf.org/

VQ: suggests going around the table to see if we're all on the same page

ht: From the particular corner of the intellectual space that I sit in, it's much more the scientific side that VQ mentioned.
... The focus seems to be sharing extremely high concentrations of computing power
... Problems like quantum chromodynamics and the calculations demanded by the superconducting supercollider
... This is simply more computing power than they can possibly have at one or two sites. It's in there general interest to make sure that everyone can get access to the available computing resources.
... there is a bunch of open source software out there for finding and integrating into the authentication model of the resources to "run jobs"
... they're also very interested in tracking the history of datasets.

dorchard: I think the area that people are seing from the WS side is the distinction of managing WS resources and using WS to manage resources
... Our customer base is concerned about managing computing resources on the order of thousands of resources
... Big customers need to be able to manage things like drive failure, manage devices, manage computers, etc. They want to have a big set of "things" avialable that they can take out when needed and to which they can add new devices.
... there's obvoius potential relationships between that view and the web view

<ht> The GGF does indeed have a pub process, modelled on ours it looks like: http://www.ggf.org/ggf_docs_process.htm

dorchard: points to Yaron Goland one of the authors of webdav. Formerly at BPEL. This is one possible starting place.
... can you fill in "Yaron Goland" for me, please

VQ asks NW about the grid

<dorchard> http://www.goland.org

NW: I really don't know more than the name. I think I've heard both sorts of discussions.

ER: We have a center inside HP that's working on Grid computing. I'd be happy to invite them if it isn't redundant.

VQ: We should try to find the right person to be invited first to give us an introduction to the grid

VQ asks HT about his connections

ht: I'll make some inquiries

ER: Suggests we each send a paragraph to the TAG outlining what our contacts are working on. As a way of figuring out who to ask.

<noah> IBM is doing a lot of work on grid. Are you looking for experts who can orient us? If so, I can almost surely find someone good.

VQ: Sounds good, please send them.

NW: Noah seems to be indicating an ability to get someone as well.

<noah> OK, I'll check in IBM. Are we looking for someone to dial into a TAG call? What's needed and what's the likely format?

We're just looking for a single para of intro so we can pick the right folks to invite

subtopic: P2P

NW points VQ to Noah's remark earlier

noah: Note that I think we covered P2P and streaming under schemeProtocols last week. The ball is in my court to propose next steps, I think.

VQ: I'm not sure the problems are all related directly to those areas. Noah has an action that's relevant, but I'm not sure that both issues are streaming and multimedia and P2P are all just scheme and protocol related
... What other directions should we explore?

Scribe lost the beginning of that thread...

VQ: P2P and streaming could be related.

ER: I don't think P2P and streaming are really related

VQ: As far as protocols are concerned, they're using different protocols than the ones typically used on the web. Also, the way they carry information seems a bit different.
... The notion of media type isn't handled the same way.

NW: So they aren't related to each other so much as they are both different from the typical web

VQ: Is the only issue related to web architecture the issue of schemes and protocols

NW: It seems to me that folks use P2P to get long streaming resources

dorchard: The notion of P2P and security and safety is interesting. How do you deal with the fact that people can inject errors into the system. Azareus now has a plugin called safe peer that does IP blacklists based on comparisons with master copies
... People are now trying to verify the representations that they actually obtain. This is an effort to heal the network's view of what the resource is. This treats the information space as if it was sort of local.

VQ: Asks Noah if he thinks there are other issues related to P2P and streaming

NM: I don't know and I certainly don't have any problem with discussing other issues. I just thought they were on the whiteboard at the f2f because of the stream/protocol stuff.
... I think the balls in my court to summarize the feedback that I got, both at the f2f and in email

dorchard: The thing I mentioned was the issue around retaining the state of the resource as it's distributed
... This is the security and distributed caching aspects I just talked about

NM: I think those aspects are related to the scheme/protocol discussion. To what extend do the scheme or protocol play a role in preserving that state? For a given scheme, it may be important. The social conventions around the http: scheme provide one view of the system that the P2P schemes don't. Or at least do so differently. I do plan to discuss some of that.

VQ: In addition to the issue of schemes and protocols, it's not completely clear how to address the specific aspects of P2P and streaming. I'm not sure how to explore further to avoid missing other aspects of web architecture.
... I don't see anything else to do in the short term.

<dorchard> btw, the global grid URI that I see is http://www.gridforum.org/

ht: Suggests that the schemes/protocols finding is a very general question, the question of, for instance, is it possible to integrate P2P transport mechanisms into the web, is perhaps a separate question and we should take it up in due course.
... I'd like to see that in a separate document, I think.

<ht> ack

<Zakim> ht, you wanted to say that I'm more concerned to hear from further afield, i.e. the 'Big Physics' GRID as opposed to the WS-Grid and to support the separation of the peer-to-peer

VQ suggests we proceed with the finding and if we discover that other issues are raised, then we can decide how to address them, perhaps separately.

NM: I think it's important to agree on a cople of things: we should be able to iterate on the scope of the findings and HT's paraphrase is a little narrower than I had in mind when I raised the issue. Some of these P2P/streaming issues are strong motivators for opening this issue.
... If they're decoupled, we should separate them, but I'd like license to find out if they are.

ht: Of course

VQ asks NM about attendence

NM: I'll be here next week, then on vacation for two weeks

VQ: Several other folks are missing today. My proposal is to stop this discussion now and focus next week with NM on the scheme/protocol issues in more details

NM: That's fine, but I've had a lot of feedback so it may take time for me to get it all together

VQ: Any other issues?

ht: I'd welcome the opportunity to ask for some feedback about the naming discussion
... That's namespaceDocuments-8 and ...
... ... abstractComponentRefs-3x
... I'm writing a longer piece on this for the TAG to consider
... I think the point I'm starting from is that a change in the way we think about namespaces has slowly evolved
... It crystalized in the discussion about whether the XML Core WG had done a bad thing by "adding a new name to the XML namespace"
... After pushing that back and forth for a while, I think that position is incoherent
... The namespaces REC provides a disambiguation mechanism, it's a syntactic device. That's *it*
... In particular, the primary conceptual unit that emerges from the namespaces REC is that of "expanded name" (What QT calls "expanded-QName")
... It doesn't make any sense to ask "is this name in this namespace". Namespaces don't have constituents. All we ever need are namespace names. The next level of discussion is about what particular things can be associated with expanded names. Typically, a spec/program/whatever uses a set of expanded names that have the same namespace name
... it doesn't make any sense to talk about type definitions or graph nodes or information items being in a namespace. That's not the sort of thing a namespace is. A namespace is just shorthand for a namespace name.

NW: I think that's right.

vq: You didn't receive any feedback or answer to the message you sent yesterday evening?

ht: No, but the user experience workshop is probably occupying some of the folks I'd have expected feedback from

<dorchard> HT, I'm a little confused. Why can't we talk about adding a "new name to a namespace", when we say that a namespace *is defined by* a particular specification?

NW points out DO's comment and notes that DO is no longer on the phone

HT indicated his intent to reply in IRC for Dave

<ht> Well, I don't think we should say that, rather that a spec assigns definitions to names which share a namespace name

VQ: Any other topics for today?
... Adjourned

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.115 (CVS log)
$Date: 2005/02/15 22:31:37 $