Trip Report

OMG Meeting

Montreal, Canada
22-27 June 1997
Frank Manola and Craig Thompson
Object Services and Consulting, Inc.

Executive Summary

Probably the most important thing that happened at this meeting is that OMG is moving toward adopting componentware technology (via a Component RFP and Scripting RFP brought to you by the Gang of Four - IBM, Netscape, Oracle, Sunsoft) and beginning to think harder than ever before about Compositional Architectures (based on DARPA-funded OBJS work on our Scaling OSAs to the Internet contract).

Who reported on what:  Frank Manola participated in the entire OMG meeting (22-27 June) and drafted most of this trip report.  Craig Thompson co-chaired the Internet SIG meeting and participated in the OMG meeting only on 23-24 June before being called away to act as reviewer for the DARPA AITS (formerly ISO) Architecture Review held in Washington D.C. on 25-26 June 1997.

Outline:

OMG Internet SIG (ISIG) and Joint Meeting with Security SIG

See Thompson's Minutes of this meeting  -- worth reading.  We succeeded in getting our Compositional Architecures work to be a hot OMG architecture topic and also helped create a place to work on Information Access/Object File Systems -like architectures like the I*3 reference architecture (we'll let Dave Gunning know).  We also comandeered a 2 hour meeting with the Security SIG to talk about many of the architecture topics on Scaling Compositional Architectures and created an opportunity for the DARPA Security Reference Model to be connected in. (and let Sami Saydjari know).

ORBOS/Common Facilities Joint Meeting

Craig Thompson and Shel Sutton briefed this meeting on the Internet SIG meeting.

John Sebes (Trusted Information Systems) briefed this meeting on his Firewalls RFP.  It passed a vote to be issued.

Dave Stryker (Netscape VP of Core Technology) presented a paper on "CORBA Component Imperatives" (orbos/97-05-25), motivating the addition of component and scripting capabilities to the OMA, on behalf of IBM, Netscape, Oracle, Sun, and Visgenic. The basic idea is to make CORBA development more accessible to non-wizard programmers, by incorporating technology similar to JavaBeans and ActiveX controls that supports visual development tools and scripting languages. Better integration of CORBA networks and the World Wide Web is also a goal. They don't want to simply use JavaBeans, however, because:

Dave was clear in indicating his belief that the sorts of "components" they were talking about were at a much more basic level that the business objects being discussed in the BODTF, and that he saw no conflict between them; he suggested that possibily different terms should be used for the two technologies. (After the meeting I briefly spoke to Dave; we had been colleagues at the Naval Research Laboratory.

Jeff Mischkinsky (Visigenic) then presented the CORBA Component Model (orbos/97-05-22) RFP. An explicit evaluation criterion is functionality in a World Wide Web environment. After some JavaBeans-specific wording was removed from the mandatory requirements, ORBOS recommended issuance of the RFP.

Jeff then presented the CORBA Scripting Language (orbos/97-05-24) RFP. The RFP explicitly calls for "a scripting language", and there was some discussion about modifying the RFP to address the issue of scripting components using multiple scripting languages. However, a motion to amend failed; most people seemed to want to separately address individual proposed scripting languages (and consider any interoperability problems raised by the use of multiple scripting languages at that time). ORBOS recommended issuance of the RFP.

Following this joint meeting, ORBOS and Common Facilities split up. As part of the subsequent ORBOS meeting, ORBOS recommended issuance of the revised Persistent State Service RFP (orbos/97-05-14), a request for technology to replace the current Persistent Object Service.

Common Facilities Task Force

Manola attended part of the CFTF meeting considering an RFP for a "Semantic Data Facility" proposed by Oliver Sims of System Software Associates (one of the submitters to the Business Object Facility RFP). A Semantic Data Facility is defined within the Common Facilities Architecture, and is intended for high-level interchange of data between objects. It might more accurately be termed a "tagged data facility"; what it proposes is support for "data objects" containing tagged or labeled values, a way of nesting data objects within other data objects, and methods for accessing data values based on the labels, and for traversing and manipulating the containment structure. A capability of this type (called "Semantic Data Objects") is contained in SSA's BOF submission, and Oliver noted that the Electronic Commerce DTF had found a need for this facility, and had helped him draft the RFP. Shel Sutton also indicated the usefulness of such a facility in the GIS domain (e.g., for passing metadata around). Prior to the meeting (via email), I had drawn Oliver's attention to the similarity of this idea to models like OEM, and I commented to the group that there were a number of variants of this idea being used in Internet applications (I explicitly mentioned SOIF and Netscape's Meta Content Framework). On the other side, there was a lot of concern that the proposed facility mostly overlapped the Property Service (possibly in combination with Pass By Value). Andrew Watson also suggested that a better OMG type system would subsume this; hence these requirements should be translated into type system requirements, which could then be addressed by ORMSC. The end result was that Oliver was basically told to work up, with help from several members of the AB who were present, an analysis of why the Property Service (possibly in combination with other OMA capabilities) was inadequate to deal with the requirements he had in mind. I suggested that it would be useful to have this analysis "on file" as a quasi-OMG statement of how to address these sorts of requirements using OMG capabilities, since they keep coming up all the time in different application domains. I also volunteered to help with information on the relationship of the "Semantic Data Facility" ideas to the Internet technologies I had mentioned (I had already done a certain amount of analysis of this prior to the meeting). This area of work within OMG might also be of interest to, e.g., Netscape, since they are working with both SOIF and MCF.

Object and Reference Model SubCommittee

The ORMSC met both in plenary sessions, and as several subgroups.

Thompson met with Jishnu Mukerji <jis@fpk.hp.com>, chair of the Object and Reference Model Subcommittee (ormsc@omg.org), who agreed to reserve Thompson a slot for a 45 minute discussion on compositional architectures at the next OMG meeting (in Dublin).

Colin Ashford (Nortel) gave a presentation on the TINA-C Business Model (TINA-C is a telecommunications industry consortium). This is a reference model based largely on RM-ODP concepts, but with some differences in the viewpoint definitions. The basic idea is that the TINA-C model provides a concrete example of how RM-ODP might be used by the OMG in developing its reference model.

Reference Model Working Group

Thompson met Kevin Tyson, chair of the OMA RM WG.  This brand new subgroup of ORMSC is working on The overall goal is to try to provide a coherent picture of all OMG TF activities by defining a "reference model".

Their Mission Statement calls for explicit levels of abstractions, well-specified rules of composition, and separation of concerns but is no more specific.

I suggested that they do an informal RFI for OMA-NG by requesting position papers/green papers to get community ideas for extensions needed to the OMA (what I am calling OMA-NG) (Kevin later reported to the ORMSC plenary that he intended to follow this suggestion). Our composable issues mini-green papers and componentware glossary will fit here exactly. I suggested we want to be careful about whether and how we extend the OMA - by modifying it or by adding appendices.  As under-specified as it is, it has served as a very useful community architecture for seven years so maybe we don't want to muck with it directly.

It was noted that there was related work going on within the BODTF, and there was an issue of who should be doing what. Kevin noted that, according to its charter, the RMWG was the place to bring issues from other Task Forces for resolution, and that while RFPs must continue to go out from the separate TFs, that it is still necessary to try to clean up the architecture. He noted that there may also be an Object Model Extension WG chartered (since the ORMSC only does actual work through WGs). Kevin indicated that he expects a key issue will be specifying behavior.

A problem with this group is that it is somewhat ODP-centric though the charter is OMA-NG. The basic idea seems to be that, since the group is commissioned to provide a "reference model", and RM-ODP is a "reference model", that the latter should guide the former, particularly since OMG has already more-or-less adopted the position that its specifications should conform (in some generic sense) with RM-ODP. ODP frameworks will also fit here but ODP (in my opinion) provides a poor guide for how to grow OMG over the next five years though it contains some good ideas like ODP failure transparencies and viewpoints. In particular, translation between RM-ODP concepts (which are well-thought-out but often rather abstract) and the detailed considerations that OMG must deal with may be rather difficult.

Semantics Working Group

The Semantics Working Group discussed a draft green paper (ormsc/97-06-02) written by Haim Kilov and Kevin Tyson, which attempted to: After some wordsmithing, this paper was adopted by the WG, and will be forwarded to the Architecture Board as background material. (Some of the discussion suggested that the appearance in meeting reports of phrases like "the meeting generated fruitful discussion" should be interpred more literally as "participants threw tomatoes and other overripe fruit at each other"). The WG also commissioned two additional green papers, on: Drafts of these papers are expected to be available at the Dublin meeting.

There was also discussion of a proposed RFI on ways to precisely specify model semantics, with specific emphasis on ways to distinguish between system and business specifications of semantics. It wasn't altogether clear what the intent of the RFI was, and there was some objection on the grounds that it could be interpreted as conflicting with ongoing work within the BODTF. As a result, it was decided that there would be joint work between the Semantics WG and the BODTF that would analyze the current JBOF specification language (CDL) based on the ideas in the Semantics WG green paper, with the goal of producing a deliverable to both groups, instead of an RFI.

Quality of Service Working Group

Chris Shuman, QoS WG chair, reported to the ORMSC on the activities of his WG. He said their goal was to produce a green paper that would be the start of work to identify what specifications were necessary to insure the compatibility of the various QoS-related activities going on within various OMG groups. In particular, the group wants to define a framework, common terminology, and a reference model identifying: The goal is to specify the minimum necessary to provide QoS within the OMA. Individual RFPs will then address specific mechanisms to provide QoS support. He said that the green paper (ormsc97-06-04) was progressing well, but that it was the result of a relatively small group. Thus, he wants wider participation in order to get wider representation of ideas, and buy-in from vendors and other OMG groups, as well as more exposure of QoS requirements within OMG. The goal is to have a more complete green paper with which to address the Architecture Board at the Dublin meeting on ways of getting QoS requirements into OMG's technical agenda.

There was then a brief presentation on concepts for the QoS reference model, including:

The presentation referred to the BBN work on a QoS architecture for CORBA networks that we have already looked at. The group has a mailing list at QoS@omg.org.

Other Reports

The ORMSC also heard brief reports on MOF/OA&D progress (with emphasis on object model issues) by Jim Odell, and work on object model extensions by Andrew Watson. Jim indicated that there was a potential for MOF concepts (coming from UML) to get into the Core Object Model. MOF is now the UML default meta-metamodel, with the two being aligned by both sharing of common structural ideas and explicit mappings. There is an intentional divergence in relationships (with the UML being much richer). The MOF/OA&D folks see increased rigor of specifications and additional concepts for expressing relationships and behavior as issues for OMG object model work.

Andrew noted that the Core Object Model is a chapter in the OMA Guide, and that an additional printing of this is now necessary. This creates a timing issue as to whether Core Object Model extensions can be completed in a reasonable time for inclusion in a new printing, or whether printing will have to go ahead without these changes. He issued a call for volunteers to work on those extensions; in particular, to look at the existing text, see what changes needed to be made immediately, etc. He said that he and Jishnu would put together a draft plan for discussion at the Dublin meeting.

Business Objects Domain Task Force

The BODTF discussed the following general issues: There was a fair bit of discussion at the BODTF meeting about BODTF having a "PR" problem with respect to the rest of OMG (particularly the platform side), due to a perceived lack of progress (and, in BODTF's opinion, a lack of understanding of the complexity of the problem compared to typical platform problems). Some people apparently feel that the BOF RFP was really an RFI, and they should admit it and withdraw. (On the other hand, there was no explicit "anti-BOF" activity taking place that I could see). A number of suggestions were made as to how to (positively) clarify the situation of the BO work to the rest of OMG. (It seems to me that there should be explicit work going on somewhere in OMG that shows or suggests how to combine existing and yet-to-be-specified OMG capabilities to build business applications, independently of the work on specific platform capabilities and vertical domains; and that this is generally what the BODTF is (or could be) about). It was also suggested that those developing BO definitions need a common language in which to describe them, and that the work of the BODTF on CDL is addressing this requirement (although it was noted that a more precise roadmap as to how to progress CDL was needed). A matrix was circulated showing the different business object definiitions being developed by various domain groups (Account, Currency, Party objects in Finance; Actor, Assembly, Person in Manufacturing, etc.), and their relationship to CBOs (Involved Party, Resource, Agreement, etc.).

The effect of the Components RFP was also discussed. Some people apparently saw this as a political response to get the concept of "components" out of the domain side, rather than a response to general industry interest in "components" as an underlying software technology. The justification for ORBOS "Components" (as presented by Netscape, et.al.) explicitly stated that it was much lower level than Business Objects, and essentially a different thing. From that point of view, it can be interpreted as filling in some of the previous large gap between existing OMG technology and business objects, and hence make the latter more plausible. On the other hand, the actual wording of the Components RFP could be generically interpreted to subsume the BOF, and hence opponents of the BOF activity could use it to argue that separate BOF work is no longer needed; instead, business objects should simply be developed in terms of Components. (Since there are a lot of individual domain groups developing business object definitions, those wanting to make that argument will have to be careful that the domain groups agree with them). Cory Casanave (BODTF Chair) said that Data Access (his company) would probably submit a response to the Components RFP. He also said he wouldn't mind a BO being a JavaBean, and that he saw that as an implementation issue. (Further work needs to be done to verify whether ORBOS "components" fully satisfy the actual requirements for BOs, or whether additional capabilities or further constraints are necessary).

A work-in-progress paper was distributed on a domain metamodel for business objects. Bill Cox, CFTF chair, brought in a draft RFP for a Calendar Service (for groupware, personal scheduling, meeting coordination), which was approved for release by the BODTF.

Architecture Board, Platform Technology Committee, and Domain Technology Committee Plenaries

Architecture Board

The Architecture Board (AB) plenary meeting reviewed liaison activities, and ORMSC activities. A new mailing list, rmwg@omg.org, has been set up for the Reference Model Working Group; anyone attending part of that group's meeting will be automatically on that list. OMG is a candidate within ISO for being an organization whose specifications will receive fast track status.

Platform Technology Committee

ORBOS reported recommendation of RFPs on Components, Scripting, Persistent State Service, and minimum CORBA. The Common Facilities Task Force (CFTF) reporting issuing the Firewall RFP, and adoption of Initialization and Time Facilities. They reported that there would be a single MOF revised submission for discussion in Dublin, as well as presentations on Workflow. The OA&D TF reported that all submitters were going to merge into one submission, and that this was still being worked on. They have pushed the final submission date back to Sept. 1. They also reported that there may be an RFP for a CDIF-like file stream definition for CASE tool interchange brought up in Dublin. The Real Time SIG reported they had received 20 submissions to their RFI; the minimum CORBA RFP has been voted out, the Real Time CORBA RFP1 and Enhanced Time Service RFP are in rework, the Dynamic Scheduling RFP is in final draft, and they expect a green paper on fault tolerance for discussion in Dublin. The Internet SIG reported on their compositional architecture work.

The plenary then voted to approve the Component, Scripting, Persistent State Service, Firewall, and minimum CORBA RFPs (these had been previously approved by the AB). During discussion on the Scripting RFP, there was a motion to amend the RFP to indicate that this was the first in a possible series of CORBA scripting languages. This was voted down, largely on the basis that the AB would have to re-approve the revised RFP, thus effectively delaying it.

In a surprise move, a motion was made and passed to dissolve the Common Facilities Task Force. A separate motion (which also passed, after discussion of specific assignments) parcelled its RFPs/ongoing activities out to other existing groups. Facilities work that was obviously in some particular domain went to that group (e.g., Financial Facilities went to the Finance DTF). A lot of general technology work went to ORBOS (Internet was one example, the Printing RFP was another). MOF went to the OA&D TF; Workflow went to BODTF. There were attempts made to move both MOF and Workflow to ORBOS, but there was a lot of opposition to both ideas. There was a lot of opinion expressed that suggested that people (at least on the domain side) saw a continued need to keep Enterprise-level work separate from more generic stuff (for example, they saw a need to keep workflow separate from scripting, feeling that workflow was primarily going to be used to define Enterprise processes, and the workflow products were based on the use of business-level objects, as opposed to the lower-level scripting stuff).

Bill Cox, the CFTF Chairman, was apparently told at 9AM the morning of the plenary that these motions were going to be introduced, and it was a complete surprise to him. The "official" reason given (by the proponents) was to reduce the number of distinct meetings that people would have to attend in order to track items related to "the same subject" (and I could see some substance in that argument). On the other hand, many people on the domain side apparently felt this was another effort by the platform folks to rein in the domain side, and keep things more under the control of the platform side. My own view is that, while there might have been excellent justification for this action, the way it was handled appeared underhanded (procedurally correct, but not really up-front) and smacked of political manuvering, and could very well give rise to all sorts of adverse interpretations that may or may not be correct. I don't think this mode of operation, procedurally correct though it may be, will help OMG's cause in the long run.

Domain Technology Committee

The DTC voted to accept the groups that had been allocated to it as the result of dissolving the CFTF. It also reported issuance of RFIs on BOF and Clinical Decision Support, and RFPs on Health Claims, Negotiation, and "Party" (from the Insurance/Finance groups). Technology in response to the RFP on Control and Management of A/V Streams was voted out by the Telecom DTF and approved by the AB. DTC adoption will be by email vote. (This technology may involve QoS). The Telecom DTF reported on submissions to their Topology and Notification Service RFPs, and noted there would possibly be an RFC by Lucent on the subject of GIOP over OSI for managing SONET networks. They also indicated that there would be a one-day conference at Dublin (on Monday) on management of CORBA networks, including the topics of TMN/CORBA interworking, logging of management events, and alternate transport technologies (ATM, SST, etc.), and possibly a later one-day workshop on CORBA/Intelligent Network interworking. It was also reported that a BioInformatics Working Group would hold its first meeting in Dublin (contact Jon Siegel at OMG for information).

ANSI X3H7

X3H7 (now NCITS H7, since X3 has renamed itself the National Committee for Information Technology Standards) met the Saturday and Sunday before the OMG meeting. As noted in earlier reports, H7 is the U.S. group working on the Enterprise Language for specifying the Enterprise Viewpoint within the ISO Reference Model for Open Distributed Processing (RM-ODP). RM-ODP is a general reference model to which OMG specifications are supposed to conform (OMG RFPs generally contain an explicit requirement for RM-ODP compliance; and CORBA IDL has been made an ISO standard for object interface specification within RM-ODP). Most of the meeting was spent preparing a U.S. contribution for an upcoming ISO meeting in Helsinki (contributions from the U.K. and France had also been circulated for review prior to the meeting). Discussion at the meeting focused mainly on two topics.

The first topic was the potential use of deontic logic (a form of modal logic) for describing issues of "obligation" and requirements within the Enterprise Viewpoint (EV). Much of the EV is concerned with describing the obligations and other "contractural relationships" of various parties and components within both the Enterprise and its environment. Deontic logic has been discussed within the Semantics Working Group of OMG's Object Model (now Object and Reference Model) Subcommittee, and a tutorial on deontic logic was given as part of that Working Group's activities at the last OMG meeting. Manola argued that it was important in using any formal method, such as deontic logic, to have a good intuition about the modeling requirements; otherwise, there was a tendency to let the formalism constrain the requirements, rather than letting the requirements govern the use of the formalism. Hence, he argued that effort should be made to thoroughly develop the requirements for expressing obligations, and other concepts describable by deontic logic, with the recognition that deontic logic represented a formal tool available for use when that became necessary. Others noted that there were a number of variants of deontic logic, and that they generally had formal "problems" (e.g., possibility of generating contradictions) of one sort or another.

The second topic was a discussion of the concepts of "scope", "purpose", and "policy" within an EV description (the EV is supposed to describe these aspects of an information system). The discussion led to some resolution of what these things were supposed to mean. Joachin Miller, the H7 chair, is preparing a draft contribution to the Helsinki ISO meeting incorporating these ideas.

Manola announced that he expected a revised version of the H7 Object Model Features Matrix would be available at the OBJS Web site by the end of this OMG meeting (this version is in fact now available, and an email message was sent to the H7 mailing list announcing this fact).

H7 also discussed the possibility (suggested by X3T3) of merging with X3T3 to concentrate all work related to RM-ODP in one committee. T3 is generally in favor due to an apparent lack of resources within T3 to pursue its many projects. There was some opposition within H7 to this move, as it might involve H7 in a number of extraneous activities and overhead, and involve the need to have EV-related material voted on by those not familiar with the subject area. The issue was ultimately tabled for further review later.

Action Items/Notes of Interest