Lifecycle Discussion based on the current spec. Lifecycle's everything between creation and termination creation done via the factory, and leads to the negotiation phase termination is handled via the terminationTime OGSI operation either party can cause termination termination does not release parties from all commitment, particularly penalties States negotiation states (considered -> observed) in previous discussion idea of commitment added during yesterday's discussion, perhaps we'll come back to such discussion satisfaction of terms as from the spec: satisfied, violated, completed, partial satisfaction may be less severe Agreement state diagram considered -> satisfied satisfied -> violated or completed can we go from violated back to satisfied? Ming, yes. Jon, yeah, but it's not quite the same since a violation did occur? Asit, must maintain some state about violation having occured for post mortem saving or notification of the parties. Can be used for triggering penalties. Karl, how are our operations changed by going into violated state? Really, we need the audit trail that this has happened at some point. Karl, is observed really a mega-state. Perhaps the states are really just properties or current snapshots. Ming, all we really care is that we have an agreeement or not. Alain, violation requires that we can talk about guarantee semantics. Karl, what's the diff. between term violations & agreement violations. Kate: summary, we must capture an attribute that violation occured Kate: What about completion? Karl, comments related to termination states and whether the agreement ever really had meaning. Asit, remember this is about whole agreements rather than about terms (which was discussed yesterday). Use of terminationTime termTime for purposes of limiting the time we have to reach agreement, but it could be part of negotiation as well and we're trying to use it for a special purposes. Alain, why not termination time on terms? Kate, price example shows how a term might have limited lifetime Karl, perhaps we look at this at term grouping and phasing, so the lifetime of an offer could be tied to a term grouping. That is, whatever granularity is used for offer/counter-offer negotiations. Kate: this implies that this term time (on the group) is distinct from the overall agreement term. time Ming, this generally feels too complex (he might be tired as it's late in the day) Asit, this can be optional or there can be some default behavior that provides the simple case Karl, how can you do fine grain negotiation without this? But, if a provider doesn't support fine grain neg. then we probably don't need that. Karl, thinks of "active" has to do with attributes on terms wrt when they are enforceable. XXX, Question, is there any guarantee/assertion that terms are in a consistent state? Kate: not presently, but that's a good idea Ali, what about state diagram for individual terms? XXX, If terms are not asserted to be mutually consistent, then there's no point to some of this like term. times XXX, How do lawyers frame negotiations and terminology related to them? Karl, termination times are really about offers, not about terms. XXX, Can you add terms during an agreement? Karl, we've discussed this somewhat, and there may be a way of doing that. How is agreement termination time related to termination time of the target? Must the agreement outlive the target? Chris, we must be careful with the use of ogsi:termTime because we might do record keeping, so the completed state is important. Karl, some of this was discussed at interim wrt finalization and putting something into an audit system even if the real agreement is no longer around. So, we need to negotiate for the needed finalization activities. Asit, back to the original question, they're really distinct things, so these should be independent Kate, not really what she's asking, what about the example where a reservation completes after the agreement completes is really the question? Ming, think about kerberos tickets, I have to renew those to keep access to my services via kerberos Summary: if the agreement terminates before the active time of any terms complete, then those terms/agreement are completed. The old active time is lost. Karl, in this case you SHOULD be able to renew agreement to extend to the end of the active term. XXX, finalization gets involved here because it might need some of this "residual state" / agreement terms even after the agreement itself is gone. Ming, one approach to finalization is an "auditing service" that gets callbacks as the term/agreement states update. Kate, do we define such a service in our scope? Group, no? Kate, do we provide requirements? Chris, probably some of this is domain specific, so we probably cannot do this. JonM, like the resource usage service. XXX, there are international changes in requirements as well. XXX, all this finalization fine-print must be captured somehow, but a lot of this stuff is outside scope. And someone might need to draft these higher level things. XXX2, IETF has banned the term "Agreement" from SLA and use specification due to legal interpretations. Asit, ultimately this is just language, and what we get here is not a legal contract. WRT termplates Karl, has been looking at this wrt JSDL. Templates actually easier than supportedAgreements because it looks already like the structure of an agreement. Kate: Fusion uses the same approach Alain, DAIS, also looking at templates Karl, perhaps templates ends up replacing supported agreements? Chris, but what about for discovery? Asit, on expiration time: by adding another term. time it is not based on OGSI Karl, that's completely wrong. the OGSI notion already captures, so it makes no sense to add another thing with the same semantic meaning? Finally, commit handshake? Consensus remains that it is needed. Karl. the spec. is broken, and it needs this, so we have to solve this. Switch to Asit's presentation on templates as we talk about terms, remember it is not a single value, it could be an entire clause with lots of meaning/semantics, etc. Customization is kinda like car buying. You set up different "option packages" and that ends up having different constraints, results such as price, and you can negotiate on any of these. Coming up with these option packages is complex, and the template helps with this. Also, terms are not all SLA things, they can be parameters for the service (e.g. naming the subnets when we're negotiating a network connectivity service) Ming, enumerating all possiblities can be hard/impossible. Asit, sure just leave everything open and put no constraints, that's always possible Kate, so you end up with an empty template Asit, sure, but then you maybe don't even need to talk to the provider. Jim, so, any reason the templates cannot go into an SDE. Ming also asks? Asit, yeah, we think it can go into SDE Ming, mostly this is when the client hasn't a clue how to specify terms? Asit, at least when they don't know the meaningful way of doing it Karl, so it basically mixes in the schema and the policy. Kate, don't have to get template. Perhaps they reuse an old agreement. Asit, sure. Asit, Data svc. people asking about this, and we just don't have them yet Karl, also, for scale, etc., these could be collected into a central service for discovery and could contain with them names of providers that can support them. Karl, for repositories, we may need a simplified version of template for these repositories relative to what provider's factorys can say about themselves. Chris, why? Karl, the sort of query is gonna be complex, and doing that on arbitrarily complex templates is gonna be difficult. Question, how do we make reference to WS-Policy if it is not in any standard body? Because it is publicly available, so we can refer to it. What about IP issues though?