[ogsa-wg] FINAL CALL: OGSA 1.5 Architecture and Glossary

[Karl Czajkowski]

On Jul 17, Subramaniam, Ravi modulated:
...
> Would it be accurate to observe that where the disconnect here is that
> the view taken in the thread is that the establishment is the key and
> hence having that bi-party is desired but I have taken the
> "binding/honoring view" where the referenced parties define the
> agreement?
> 

Unfortunately, I don't think that is the disconnect. Apologies for
this long e-mail, but I want to recap some of the issues which have
haunted WS-Agreement.

There are three different relationships, as you mention.  This same
debate was had regarding WS-Agreement, and it wasn't merely an issue
of feature minimization, but also a philosophical issue:

  1. The communicating parties in the "establishment" of the
     agreement, obviously important in WS-Agreement. These are the
     "initiator" and "responder" roles in our protocol.  One can
     actually design multi-party protocols here, but the debate is
     whether this is desirable---whether it models the underlying
     relationships and decision-making that is to be exposed.

  2. The obligated parties of the agreement. We basically map these to
     the protocol roles in WS-Agreement, but we recognize that they
     can be delegated relationships... this is part of the internal
     logic of the service to either make decisions or "broker"
     decisions. As you said, the communicating parties "represent" the
     obligating parties during the establishment of the agreement.

     This obligating relationship is actually what people are
     referring to when they say that, in practice, real world
     agreements that "appear" multi-party are really sets of bilateral
     agreements.  This boils down to how the obligated parties can
     reconcile things with an auditer or within a legal system.  This
     is complicated by the fact that new entities are sometimes
     introduced, e.g. a set of members agreeing bilaterally with an
     "association" over which the entities will also gain fractional
     control.

  3. The services which are the subject of obligations. It sounds like
     everyone here is seeing this the same way, so I won't belabor the
     point.

The two philosophical debates to be had are regarding the cardinality
of these obligating relationships (2) and also whether the obligating
relationships are important to informing the communication patterns
and semantics of the protocol.

I think the best reason to investigate real-world agreements/contracts
is to see what we can learn from them about practical solutions that
have been found.  For example, this can inform us on the important
differences between negotiation+agreement, performance, and
reconciliation which may occur in different overlapping but not equal
time periods. (For example, one might legally resolve a contract
dispute after the "service period" is over, but before the contract
has become legally irrelevant.) This has helped GRAAP-WG to delineate
the function of WS-Agreement service representations: our service
represents a basic negotiation and agreed-service monitoring
infrastructure, but retires agreements to some presumed external
entity for audit and reconciliation. Our agreement language is
(hopefully) complete enough to guide the reconciliation phase, as well
as to guide the actual negotiation and performance phases.

However, I personally belong to the school which sees this as just a
useful model for automated agreement patterns. It makes the
distributed architecture more understandable to follow the social
idioms used for human-based agreements and contracts.  There are those
who even question the legality of things like digital signature and
non-repudiation. I do not want adaptive computing systems to really be
forming legally binding obligations. With their problems, digital
signatures are still much easier to justify than the idea that we can
delegate legal responsibilities to a machine-driven decision process!

I think there are examples of machine-driven, binding decisions out
there, e.g. in the financial trading sector. But, I think many
computer systems operators/owners will be scared away from Grid
technologies if they are told that their computer will now begin
signing contracts which bind the owner... I think we need a lower
barrier to entry into this field.


karl

-- 
Karl Czajkowski
karlcz at univa.com