[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [orms] 2008-05-02 Mtg Minutes (plain text)
ORMS TC Minutes 2nd May 2008 (continued from previous day's meeting) Scribe: Bill Barnhill Minute taker: Bill Barnhill 5 Review of the TC charter -- the statement of purpose, goals, deliverables, schedule, etc. 5a. Drummond Reed's proposed changes (sent to mailing list) Bill Barnhill moved to adopt Drummond's charter change Hal seconded Approved without objection 5b. Bill Barnhill's proposed changes (sent to list via Drummond) Discussion of meaning of score(s). Position noted by Tony Natalin and Nat Sakamura that they feel the proposed change broadens the scope. Hal made motion to amend 'assessment' to 'rating', not seconded Chris made motion to amend assessment to 'calculation' .Seconded by Hal Hal read the section on clarifying vs expanding the scope Jeff H. noted contextual identifiers may be output Chris, Daniella noted that contextual identifiers may be covered under relevancy Jeff noted the assessment word occurs multiple times, so calcultion would need to replace it in several locations. Chris clarified his amendment that it be changed in all three places Hal seconded it Tony, asked if discussion finished and asked if anyone opposed to the amendment, calling the question. Tony clarified that this was only the amendment to change the word assessment to calculation in the proposed new charter text, not a vote to adopt the proposed changes to the charter. No objections, amendment passes. Now voting on accepting changes to the charter text,. Tony asked for objections. Tony and Nat made objections, Hal made motion to table this motion and submit the amended text to TC admin for an advisory opinion as to whether this would constitute a clarification or a scope change. Seconded by Chris. Hal noted we can take action after we get advice if we table it for now. Hal restated his motion: Table the motion and submit Tony: Calling the question on tabling. 3 objections: Les, Jeff, Frank Motion to table until after we get advice passed, so it will be sent to OASIS for advice. Tony asked if there were any objections to accepting other changes (Drummonds), no objections. So discussion moved to Use cases. 7. Go over Possible Usecases & Scenarios Daniella went over her slide deck, noted it will be contributed to TC document area in PDF form. Chris then went over his slides: Peter asked Chris: Sounds like you're defining defining space to be relevant inputs Chris agreed. Peter: Relying party is the one that's defining this space. So it's interesting, is the RP defining the space or is the OP defining the space Nat noted if you calculate a score it doesn't become a reputation until you publish it Peter: Definition of context is what's really important. Context of the aggregate result is what we're really getting at. Reputation evaluation is going to be based on my particular context. Jeff: If you take the model that's in the XACML then the policy decision process is well defined. Hal: When you get the score you need to get the inputs that were used to create the score. In some circumstances knowing the output and all the inputs would allow you to reverse engineer the algorithm Note (Bill): Addressing all the inputs may involve considerations of privacy, if you give someone a negative vote will they know? Chris then continued with slides Peter: To a great extent all the statements on Chri's slides are related to the verifiability of the claim. Chris: Correct Peter: Seems very related to PEP (Policy Enforcement). System that gets reputation score as input, never mind what it's doing with it's data. Expectation is output will be a policy decision. Bill: WS-Policy analogy, using custom assertions ie Trusted Hal: (ref: Bill's analogy) Not how I see WS-Policy working Hal: On slide 4 are you saying the stuff outside box, policy is what we're trying to address? Jeff: No Hal: In which case I may not disagree. My interest is what's in the box Drummond: What I was seeing was that you could take a bunch of these boxes and connect the outputs Bill: Reputation pipelines Peter: Use case that was described was really about inputs to a PDP. How I actually tell the inputs that they're wrong is something different. I think we need to solve that, but it's something different. The context is about controlling the inputs as a RP..the context may be pictorial outside the defining space. The RP can dictate some of the inputs to the algorithm. JJ? : In it's simplest form the inputs might just be weights. Peter: Scoping question: Do we care whether an RP can tell the algorithm what to do? I can either tell Ebay to give me a #, or I can tell Ebay to give me only the number coming from these users. Peter: So the question is whether the RP gets some kind of dynamic set of inputs to the algorithm or whether it will be static inputs. Jeff: The weights are Historesis (ref control theory, and check spelling) aka feedback to the algorithm. Chris: Essentially this is personalizing the algorithm. AI for Bill: Write up his use cases in a doc and publish to ORMS TC docs. AI for Bill: Work with OASIS TC Staff to fix permissions on TC Doc uploading Nat: Anything more to discuss on slide 4? No one said. Nat: Next time we really need to define reputation. Peter: I think the input of the producer or the defining space is a policy information point with respect to the relying party, Most of the use cases are relevant to inputs into a policy decision. An example is a linkedin invitation to join a social graph, and I look at their links to decide whether to include them. Bill: Have we clarified that trust and reputation are two separate things? Jeff, ?: No, and that's something we really need to do. I would say we need to generate a UML reference model and we shouldn't invent it out of whole cloth. Jeff: Wiki is poor tool for hacking on documents. But great tool for publishing stuff. Peter: Some combination of email and wiki will be used for terminology. Nat: Who's leading terminology work? Jeff H: I'll do it. Nat: Something by next week. Perhaps have strawman of the list of terms to define. Terms we need to define, strawman: Context, Score, Reputation, Defining space, Input, domain, reputee, reputer If we have definitions that we've come up with then we should email Jeff H with them. May have an adhoc meeting, no decisions as no quorum, at IIW. Nat: If you have any input to the TC please submit it. Peter: First real work will be a use case document. AI[Bill]: Check with Mary if there is a template for a Use Cases document AI[Bill]: Come up with Use case template document by next week, but not fitting everything in by next week. Nat presented slides of his presentation: Nat: We don't want our model to be too much of a moving target Bill: True, and we also don't want to go the other way by defining too stationary or rigid a target or we risk alienating a large portion of our potential adoption audience Dorothy: Input variables should be user defined. It's important to keep in mind the capability to represent subjective inputs. Need to include what meaning of value is. Chris: Degree of uncertainty needs to be captured. Jeff: Confidence level as seen by the provider, and confidence level as seen by the reputer. Chris moved to adjourn, Drummond seconded. --- end
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]