OMG Agents Working Group

No one is making much money with agents
No compelling case to use agent technology instead of a conventional approach
Not a safe harbor like objects and RDBs
No compelling-sounding RFPs
Not fully baked (e.g., ACL syntax, architecture)
Agent-object and Agent-XML coexistence and interoperability not well understood.

Options for Agent WG:

Kill Agents WG - not warranted, too much material and promise to take this route
Escalate via becoming an OMG Task Force and issue RFPs - not yet warranted, too few industry players will respond to RFPs at this time
Keep it low-level in 2000, but still be useful in various ways

Suggested direction for 2000:

Become a Platform Special Interest Group (PSIG)
Flesh out RFPs (via whate paper)
Education (e.g., ACLs, agents vs objects, critical infrastructure issues for agents)

Within Agent WG plenaries
Speaking to other OMG groups
Speaking to outside groups

Bring in new blood (new members, new experiences, new specs, possible RFPs) e.g., DARPA and industry players
Sponsor agent issues forums (e.g., why ACLs, how can we make agents pervasive by 2002)

FIPA Liaison Report, Frank McCabe, Fujitsu

FIPA (http://www.fipa.org) started in 1996 with 50 organizational members (originally 50/50 industry/academic, now 75/25, more European/Asian). It was originally dominated by broadcasting and telecom companies (with a PDA vision). It has around ten standards. FIPA is going through a sea change. The old regime left in October and now they have a new process and new officers. The world has changed. The focus continues to be ACL and cognitive agents that know what they are doing, not ant communities, mobile objects, or stand alone agents. FIPA continues to do work on its abstract architecture as well as some work on higher level services. The next meeting is in Portugal in April.

Presentation: Agent Social Networks, James Odell

Jim Odell (OMG document internet/00-03-05) gave a presentation on Agent Social Networks. Small world networks focus on the phenomena of closeness in networks. The Bacon Phenomena is actor knows actor knows … Kevin Bacon. Urdish number - involves nearness in citations . The entire web is only separated by 19 clicks of separation. Analysis: Caveman network is lots of little islands and full connectivity in a cave. Now interconnect into a ring. Star wheel - fragile if you lose center. Fully connected is inefficient. A few local connections and then 1% random connections and you get good benefits. Marines are setting up interconnection experiments. Marines are using this to locate resources - pool a bunch of guys across the services and then when a need occurs they ask do I know someone who can do this. Also, applied to western US power grid. Questions: domain of applicability? Business benefits? This might be applied to the federated directories problem. Disbursement of information, viruses, information, diseases. Q: how is this related to Patti Mayes Collaborative Filtering.

Agents Technology Green Paper, James Odell

Jim Odell reviewed the latest revision of the Agents Technology Green Paper, version 0.91 (OMG document ec/2000-03-01). Revised material appears in Section 7 on the relationship of agents and objects and in Section 11 on the Other Standards Organizations.

Discussion: The Relationship of Agents and Objects, Craig Thompson, OBJS

This topic is shaping up to be very important for OMG if we are to find good and widely useful ways for agents and objects to co-exist and interoperate. Based on a discussion at OMG Mesa, Craig Thompson revised Section 7 of the Agents Technology Green Paper. [Inadvertantly, some useful older material was removed and so we will add it back in soon.] There was some discussion of whether comparison (similarity/dissimilarity) was the correct way to communicate the relationship of two technologies. There was some strong feeling that agents are not objects, and an opposing view that agents are just objects with some new things that should be added anyway. There was another view that we should just envision and describe a way of thinking about autonomous, emergent, interactive entities—and label this as agents. Then, determine how objects figure into this world. More viewpoints and debate are needed. [Note a revised version of the section on The Relationship of Agents and Objects is now available as OMG document internet/00-03-09, containing material from both earlier versions.]

Agent White Paper and RFP Roadmap Discussion, Craig Thompson, OBJS

The purpose of the Agent Technology White Paper and RFP Roadmap, OMG document Internet/00-03-03, is to provide a rationale and roadmap for future RFPs. [A new version of the document reflecting the meeting discussion is OMG document internet/00-03-08.] The current document reflects discussions held at the OMG Cambridge meeting because we did not really discuss this topic at the OMG Mesa meeting. The current list of higher priority RFP candidates is:

Agent Identity

Message transport (aMail)

Agent discovery/matchmaking

Agent communication language

Ontology

Content language

Agent security

Agent/object mobility (triaged)

UML profile for agents & ACL & agent platforms

There are many other possible RFPs identified in the white paper. This list indicates that the Agent area has several areas where work leading toward standards is needed—and certainly could justify making the Agents WG into a Task Force. However, there are not (yet) enough resources to take on such a list. Also, many of the items might be addressed by other standardization groups (e.g., FIPA). Frank McCabe suggested that it’s not that it does not need doing, but who can/should be doing it. Frank McCabe suggested that we initially select a small number of RFPs that put agents on the OMG roadmap. Criteria that can affect the ordering of RFPs include:

interoperability

ACL communication

security

mobility

distributed, robust, large scale

It was decided that we don’t want to commit to a specific RFP roadmap just yet, i.e., we don’t want to commit to certain RFPs developed on a paticular timetable. This is because we need to build a stronger case for these RFPs, how they relate to OMG, and who in industry will respond to RFPs and develop compliant products. However, we still currently recommend that the first two bundles of RFPS might be:

Agent Interoperability RFP

Agent Communication RFP

We are attempting proactive standardization. There is not yet a great body of agent technology to abstract and standardize; we want to prevent the potential jungle of approaches. We want to benefit the industry as a whole.

After discussion, we agreed to the following changes to the roadmap document (that Craig Thompson will make):

Add content for the sections describing high priority RFPs. This may initially expose scope and issues and later be refined into RFP descriptions.
Better clarify that the long laundry list of possible RFPs is not in scope initially.
Back off from the current aggressive schedule for RFPs until we make a stronger case for the core ones.

To make progress on the RFPs, we continued discussions on the high priority candidates.

Agent Discovery and Matchmaking, Craig Thompson

We discussed this RFP at the Mesa meeting - see http://www.objs.com/isig/wg-agents08-minutes.html#AgentDiscovery and http://www.objs.com/isig/wg-agents08-minutes.html#Discovery. We decided that the ECDTF RFP on Resource Discovery was broad enough to cover most of our needs. Craig Thompson will write this up as a revised section on Agent Discovery for the white paper. Frank McCabe will list some additional requirements he thinks the ECDTF RFP does not address but are still needed to provide a discovery service useful to the agent community. Steve McConnell provided this update concerning the Electronic Commerce Registration and Discovery RFP: the RFP was successfully issued during the Mesa meeting in Arizona (see http://www.omg.org/cgi-bin/doc?ec/00-01-05.pdf). The LOI deadline is March 10, 2000 (Friday of the OMG meeting week). Currently there is an LOI from OSM. CommerceNet intends to support OSM in its submission (CommerceNet is not yet a member as as such cannot LOI directly). The initial and revised submission dates are May 23 and November 6 respectively.

Agent Identity Issues, David Levine

Agent identity is an important property, but a difficult one to implement in a secure manner. Assigning an identification is easy; hopping from identity to identity is problematic. Identity issomething that serves to identify or refer to an entity. In this way, an agent could be referred to by its name, a role that it is playing, or the fact that it is a member of some organization, and so on. An agent, then, can have multiple forms of identity. For example, a particular agent could simultaneously be a purchasing agent working on behalf of user Rolf Smith; be playing the role of a bidder in a negotiation with E-Widgets; having its software composed of elements from company Exdeus; and having the serial number 98734501. Each of these identities might be important in different interactions.

Some identity related notions:

Authentication – Using some credential model, ability to verify that the entity offering the credentials is who/what it says it is.
Credential – An item offered to prove that a user, a group, a software entity, a company, or other entities is who or what it claims to be.

Examples: X.509 certificate, a user login and password pair, a PGP key, a response/challenge key, a fingerprint, a retinal scan, a photo id. Obviously, some of these are better suited to software than others!

Credential Authority – An entity that determines whether the credential offered is valid, and that the credential accurately identifies the individual offering it.

Examples: X.509 certificate can be validated by a certificate authority. At a bar, the bartender is the credential authority who determines whether your photo id represents you.He then may determine your access permissions to the available beverages!.

Credential model – The particular mechanism(s) being used to provide and authenticate credentials.

David Levine believes we are not far from issuing an agent identity RFP. Issues are:

how does agent identity differ from object identity
how does agent identity stretch the security models already identified for objects
which companies are able and ready to respond to such an RFP.

David Levine will write this discussion up as a section in the Agent Technology White Paper.

Update: OMG SECSIG's Interoperable Security RFP, Donald Flinn, Concept5

Don Flinn (coming soon: OMG document internet/00-03-07) reviewed the Security SIG's work on the Security Service (certificates, secure invocation, authorization, accountability, security administration, security unaware and aware applications, trust delegation, interceptors) and recent work on interoperable interfaces. They are working to fit EJB into this framework. Their next step is a roadmap discussion on next steps.

we discussed whether some of the security services are progenitors for similar services that use interceptors to provide replication, versioning, and other services and whether the work on policy languages and security properties should therefore be genericized to service policy management and service policies. Of couse, all these, like security, must be assured so perhaps security is a way to go this.
we discussed what possible extensions to security are needed to enable agents:

one area might be in making security policy and decisions clear to end users. For instance, if an agent application appears on your desktop and wants to access some of your files, do you let it? How does it tell you its needs? This is partly a user interface and understandability issue that might roadblock the use of agents if they are to become pervasive.
another area is, if rather than thinking of security as client server and domain oriented, we think of an agent as implementing the entire security mechanism so it is self contained and not just provided by some separate infrastructure, does this break the security model. We are not sure that it does. It might mean that a domain contains just one agent, which would make agents heavyweight. There is also an issue of how much the agent depends on its platform (the thing that executes its code) - does it get the security services from the platform? does it trust the platform.
agent mobility appears to be a related area where the security model may not provide everything the agent community will need.
we briefly discussed scaling up to millions of agents,
we also discussed (security) enclaves and policy management.

It is clear from the short discussions that got started on Agent Security and Agent Identity that there is much more discussion needed. We need more open ended time slots to discuss these topics. These are planned for OMG Burlingame.

Next Meetings

Oslo, Norway - June 12-13 2000

Jim Odell, Frank McCabe and David Levine might all be able to attend and organize this meeting which might provide some education and/or OMG outreach and also discuss selected agent issues. An agenda will be forthcoming from Jim and/or Frank.

Burlingame, CA - September 11-12 2000

This is likely to be an important meeting. We will try to make a lot of progress on (a) educating OMG about agents and (b) resolving some key issues roadblocking our progress. We also expect to progress Agent WG to become an OMG SIG.

educating OMG about agents

tutorials

agents and agent wg - Jim Odell - 1 hour
ACL, ontology, and content languages - ask Tim Finin - 2 hours
FIPA Abstract Architecture - David Levine
Agent Construction/Development -

presentations to other OMG groups

Security - David Levine?
EC
Manufacturing - Jim and Nenad
other OMG impact: EDOC

discuss key issues - provide input to green or white paper

ACL issues
agent vs object - Craig, David, Jim
agent identity
aMail - Frank
agent security
mobility
teams

Green Paper work

revamp FIPA section (Frank)

Agent WG -> SIG

revise mission statement - Craig, Jim

Presentation on Agents to BODTF

See Jim Odell presentation, OMG document internet/00-03-06

Presentation on Agents and Agent WG to ORBOS TF

See Frank McCabe presentation, OMG document internet/00-03-04