Trip Report

OMG Meeting

East Brunswick, New Jersey
December 1-5, 1997
 Frank Manola (11/29-12/4), Craig Thompson (12/1-12/2)
Object Services and Consulting, Inc.

Executive Summary

OMG is working hard right now on a Component Model RFP and reviewed responses at this meeting -- so our January OMG-DARPA Workshop on Compositional Software Architectures is very timely. Kevin Tyson (ORMSC) is reusing chunks of our OMA-NG paper into the OMA green paper and will send us a copy for review before Salt Lake City. OMG made further progress in its work on business objects (essentially higher-level forms of components), and other key technologies, including Mobile Agent Systems Interoperability, and Java-IDL mapping. We chaired the Internet SIG meeting, presented our Web object model work to the Internet and CORBAgis SIGs, and contributed Internet technology references to the Tagged Data RFP to foster alignment of these technologies.

Who reported on what:  Frank Manola participated all but one day of the OMG meeting (1-4 December) and drafted most of this trip report.  Craig Thompson co-chaired the Internet SIG meeting and participated in the OMG meeting only on 1-2 December.

Outline:

Internet SIG

Craig Thompson chaired this one-day meeting -- see separate meeting minutes. We heard presentations on web-object integration, including Frank's presentation on Web Object Models. We led an active brainstorming discussion on OMG-Java convergence. Andrew Watson (OMG chief architect) says OMG management is not able to get JavaSoft's Jim Mitchell's attention to build a strategic relationship. During this session, we made good contact into the Oracle, IBM, part of the current industry open systems alliance. Also, MITRE is offering its MUD CVW to OMG - we offered to be an early adopter to populate an Internet SIG "room."

During one of the breaks, I (Frank) had an interesting discussion with Larry Smith (IBM), who had given a talk on HTTP-NG, on the subject of how much of the Web would ultimately be done in Java, as opposed to data formats like XML or HTML. Larry's view (which I generally shared) was that there was an "80/20" rule: about 80% of the Web was essentially data and would probably not be done directly in Java. There were two essential reasons for this:

I noted that the ease-of-creation issue might be improved for Java through the creation of better editors and other tools suitable for end users to create Java-based "data", but that the second issue would probably not go away. This is a major issue for the SGML community (and hence is somewhat "inherited" by XML), which is composed of users dealing with huge amounts of long-lived documentation, which has seen software technologies come and go, and has found that a straight text representation preserves their investment better than becoming dependent on representations based on a single software technology.

ORBOS Task Force

Component Model RFP submissions

Scripting RFP submissions (to work with Component Model)

DCE/CORBA Interworking RFP

ORBOS reviewed an initial submission to this RFP by DSTC and DEC (orbos/97-10-06). The submission defines a unidirectional mapping from CORBA to DCE (a CORBA application accessing a DCE service), the assumption being that the main industry interest would be for CORBA to be able to access legacy DCE. (Hence the submission only addresses part of what the RFP asks for). The submission defines: A bridge object (which is both a CORBA object and a DCE client) is defined which presents the CORBA equivalent of a DCE interface, propagates operations, and translates non-native IDL types. Another object is defined to allow CORBA clients to obtain information about bindings from a bridge object to various possible servers. The submitters indicated that they did not address security (mapping to DCE security), since they felt it was both too complex for them to handle, and also overlapped with other CORBA security work. Further work will be done on this submission for presentation at a subsequent meeting.

Possible new RFP for Components Extensions

ORBOS heard a presentation by Umesh Bellur (Oracle) which claimed that "enterprise components" require extensions to the capabilities described in the existing Components RFP, in such areas as persistence, transaction support, and security. He noted that the ORB and Object Services already provide the necessary support. He specifically wanted a declarative mechanism for specifying transaction properties (and alignment with the Object Transaction Service), and a declarative model to specify persistent state (and alignment with the Persistent State Service). No action was taken on this proposal.

Mobile Agent Systems Interoperability Facility

ORBOS heard a presentation of an updated submission (orbos/97-10-05) by Crystaliz, General Magic, GMD FOCUS, IBM, and The Open Group (there were originally two submissions, now merged). The current submission represents an update of the submission presented in Dublin, based on feedback from the Architecture Board. The submission was previously recommended for adoption by ORBOS, but needed a re-vote due to the changes. After the presentation, ORBOS voted to re-recommend adoption of this submission (OBJS was on the voting list--I voted in favor).

Objects By Value

ORBOS heard a presentation of a revised submission by BEA Systems, Iona, IBM, Netscape, Novell, and Visigenic (orbos/97-11-11). (I only heard part of this presentation, as BODTF was going on at the same time). A number of issues were raised, not so much about this specific submission, but rather about the whole issue of objects-by-value, including mixing of interface and implementation, loss of polymorphism, and trying to get more orthogonality between the type system and by-value semantics. The submitters asked for an extension of the submission date to January 19, to give people time to review the revision (which the submitters said was a significant change from the previous version), with a vote targeted for the Salt Lake City meeting. The submission date change was approved.

Java-IDL Mapping

ORBOS heard a presentation of the status of submissions to the Java-IDL Mapping RFP, specifically the status of the joint submission by IBM, Netscape, Oracle, Sunsoft, and Visigenic (orbos/97-11-29 is the latest snapshot). The submitters described the changes they had made in the current snapshot, including: They indicated that the submission was not yet ready for a vote. They requested (and got) an extension to January 19 for the final version to be available, with a vote targeted for the Salt Lake City meeting.

Tagged Data Facility RFP

ORBOS heard a proposal by Oliver Sims of System Software Associates (one of the submitters to the Business Object Facility RFP) for a Tagged Data Facility. This is intended for high-level interchange of richly-structured data between objects (of fairly large granularity, such as business objects). What it proposes is support for "data objects" (essentially complex values, since they are to be passed by value) containing tagged or labeled values, a way of nesting these complex values within other such values, and methods for accessing data values based on the labels, and for traversing and manipulating the containment structure. The RFP also calls for a way to define paths, i.e., a structured way of identifying locations within the nested data structure, and a way to traverse these paths to access deeply-nested data. A capability of this type (called "Semantic Data Objects") is contained in SSA's BOF submission. I had previously drawn Oliver's attention to the similarity of this idea to models like OEM (which was then referenced in a revision to the RFP), and prior to this meeting I had sent email to the orbos and bodtf mailing lists noting the relevance of Web technologies to the proposed facility, specifically XML, RDF, and DOM.

In discussions with Oliver during the meeting, he indicated that OEM was the closest thing he'd seen to what he was after, and that he would put references to the Web technology I'd suggested in the RFP as well. I, Ron Zahavi (Concept Five Technologies, and coauthor with Tom Mowbray of The Essential CORBA), and some others spoke to him specifically about the use of XML in relation to his proposal. In my opinion, the DOM interface to XML would provide most, if not all, of the structuring facilities Oliver wants, and the path specification mechanisms being developed as part of the XLL (XML linking) and XSL (XML Style) activities within W3C would provide an appropriate path mechanism.

ORBOS recommended issuance of this RFP. However, due to changes having been made at the meeting, it was not issued by the Platform Technical Committee (PTC). The 3-week rule was invoked, and the RFP will be up for discussion and possible vote at the PTC meeting in Salt Lake City.

Other

ORBOS also recommended issuance of the Interoperable Naming Service Extension RFP (97-12-23; this was subsequently approved by the Architecture Board and issued by the PTC), and heard presentations on: and initial presentations on: (I was not present for most of these other activities; they are included for completeness based on the summary sent to the orbos mailing list by Peter Walker, ORBOS Chair.)

Business Objects Domain Task Force

Business Object Facility Evaluation Working Group

On Saturday and Sunday, November 29/30, the Business Object Facility Evaluation Working Group evaluated the revised submissions to the Business Object Facility. The output of those meetings was a report presentation which was made to the BODTF on Wednesday, December 3. That presentation is on the OMG server as document: bom/97-12-17. (I was shuttling back and forth between this meeting and the NCITS H7 meeting, so, for completeness, I've combined my notes of both the Evaluation Working Group meeting and their presentation to the full BODTF with some of the material from the meeting minutes sent to the BODTF mailing list).

Much of the evaluation meeting consisted of presentations made by the submitters. The currently active submissions are:

The presentations are on the OMG server as: The evaluations started with architecture overviews of the submissions. The group then identified the different parts of the problem space addressed by each submission. One slide of the report presentation (bom/97-12-17) shows the problem space breakdown, and which submissions address which parts of the problem space.

Based on the co-dependencies between all other parts and the meta-model, the group began evaluating the meta-models of the submissions. The CBOF meta-model (BOA) is part of the CBOF submission, so the group started with this meta-model; the other submitters presented parts of their own meta-model, and made observations about the CBOF meta-model. No group conclusions were reached, but the meta-model discussion included these issues:

The group then began discussions on the alignment of the business object meta-model and the MOF/UML. Sridhar Iyengar of Unisys and Jos Warmer of IBM presented some alignment issues. Sridhar noted that the modeling concepts in the MOF and UML have been fairly well aligned, the two groups having worked fairly closely on this issue; the BOF has yet to achieve the same degree of alignment. The group feels that the alignment of the business object meta-model and UML/MOF is a very important issue, and requires input from multiple OMG task forces. The group felt that UML is inadequate as a meta-model for business objects, although the BO meta-model should be expressed as some form of extension to UML. The group also felt that there should not be a delay in adopting a BO meta-model just to achieve consistency with UML.

With limited time, the group had short presentations on the framework submissions. The group spent the last few hours on Sunday identifying content for the evaluation report.

The group made several major recommendations (which were not unanimously agreed to by all the reviewers) to the full BODTF:

The group also made a number of specific change requests to both the CBOF BOA and the EDS Interoperability Specification (for example, the Process and Entity BO types should be removed from the metamodel, as these will be better defined by the Workflow RFP).

During the meeting, I spoke to Dave Zenie, BODTF co-chair. He said he believes the CBOF submission includes all the capabilities of the others, and that the concepts defined by the various submissions are pretty close. He expects that it will be possible fairly soon to say CBOF is "it" pending alignment with MOF and UML; enough to let the various domain groups start building models using the concepts, while the alignment process takes place. He expects it to take two more meetings for the alignment, and to get the rest of OMG up to speed on CBOF enough to be comfortable with it.

Common Business Objects (CBO) Working Group

(I was not able to attend very many of the CBO sessions so, once again, for completeness, I am including some material from the CBO activity summary circulated after the meeting by Robert Shelton, Chair of the BODTF CBO Working Group).

The CBO WG has started work on coordinating CBO RFPs across the various DTFs (finance, manufacturing, etc.) based on a "CBO Coordination-Overlap Matrix" which was circulated prior to the meeting. The first order of business will be to obtain a definition/description of each CBO that has been proposed or roadmapped:

12 responses were received to BODTF RFI-1 (Common Business Objects). Half of the submitters presented their proposals in East Brunswick; the others will present in Salt Lake City.

A "CBO White Paper" has been prepared. This white paper defines "business object" and proposes a basic taxonomy to organize the WG's discussions/work. This paper is intended as a starting point. Both BODTF and DTC unanimously recommended issuance. Two additional white papers will be drafted by the CBO Working Group:

Agreement on terminology and taxonomy was reached so that the BODTF RFP-1 CBO submissions and the CBO White Paper will be in sync. BOF Submitters have been asked to use CBO White Paper terminology when discussing business objects. In addition, discussions under way are producing alignment between the terms used in the ORMSC / BODTF to describe business objects and meta-level matters. The CBO Working Group will apply the RM-ODP framework in its work, and will cross-reference this to the Zachman Framework.

The top CBOs on the priority list are:

Other key items include: An RFP addressing the top priority items is in the drafting stages.

Realtime, CBO, Transport, Finance, and BOF Coordination Session

Dock Allen, Chair of the Realtime SIG, indicated that the SIG was putting out a number of new RFPs, and wanted comments from the BODTF as to whether these RFPs would also address general business requirements. The Dynamic Scheduling RFP permits an organization to specify such things as when specific messages are to be sent, and to represent schedules and scheduling policies (air traffic control was one application mentioned). It must be possible to represent such things as: Both policy and quality of service information will be sent along with method invocations so that the server has the necessary information to satisfy these requirements.

The Fault Tolerance RFP provides capabilities to manage a software configuration to achieve fault tolerance requirements. This initial RFP supports:

Later Fault Tolerance RFPs will address: The idea is to look at problems of realtime systems not handled by current specifications. For example, ORBs currently have some undefined failure modes, and commercial transaction mechanisms are inappropriate for realtime applications (e.g., rollback is inappropriate for sensor data, and there are time constraints on the recovery techniques that can be used). Realtime requirements are currently handled using proprietary ORB extensions, or at the application level (however, realtime systems using CORBA are being built).

There was general interest in the capabilities being proposed, particularly on the part of those involved in transportation (e.g., not just air traffic control, but train scheduling) and manufacturing process control, although some of the capabilities are obviously overkill for many conventional business applications. The Realtime SIG said they had not spent much time on quality of service issues per se (except in the sense that realtime support, reliability, and fault tolerance are themselves qualities of service).

Business Object Facility Submissions Review

The full BODTF heard presentations by the submitters to the BOF RFP.

Cory Casanave (Data Access, and BODTF co-chair) presented the CBOF submission (a joint submission of Data Access, EDS, NIIIP, SEMATECH, Genesis, Prism, and Iona). He described their submission as consisting of:

He noted that the BOA builds on CORBA using UML constructs, providing a binding between design and high-level implementation concepts. He related business object concepts with ongoing OMG work on components by characterizing BOA concepts as defining "abstract components", with the OMG components providing a likely programming model (an implementation target).

Fred Cummins presented the BOF Interoperability Specification (a joint submission of EDS, Data Access, Iona, Genesis, and NIIIP). Fred noted that this submission is complementary to the CBOF BOA, in that:

Fred also described what he felt were the differences between the CBOF submission and the SSA and IBM submissions.

Oliver Sims (SSA) presented the SSA submission. Essentially, SSA has defined its own business object type system and more-loosely-coupled execution model at what they believe is a higher level of abstraction than CORBA, and loosely defined it in terms of CORBA concepts. For example, their "XOs" (Executable Object Components) are CORBA objects, but have only one operation ("execute"), and do their own method resolution (dispatching). They have also implemented a dynamic inheritance capability (something like roles). Oliver characterized their submission as providing both a runtime framework and an interoperability specification, and felt that their higher level of abstraction was more appropriate for business object development than some of the more implementation-oriented issues that had to be dealt with in using the other submissions.

Marc-Thomas Schmidt (IBM) presented the IBM, Oracle, Visigenic submission. This submission exposes more implementation-level detail than the others (e.g., it provides explicit locking options, an explicit way of dealing with location, distinct "data objects", an extended streaming capability); some of these seem to overlap with, rather than use, other CORBA services (e.g., transactions, persistence, externalization).

Following these presentations, the BOF Evaluation Working Group presented their report and recommendations.

Based to some extent on an Evaluation WG recommendation, the deadline for final submissions for both CBOs and the BOF was extended by the BODTF until Jan. 19, 1998 (a previous motion to delay the deadline to May 18 failed to carry). The Evaluation Working Group will meet in Salt Lake City to evaluate the final submissions. A motion (by IBM) to set up the joint BODTF / A&DTF / ORBOS / ORMSC working group recommended by the Evaluation Working Group to handle UML consistency issues was ruled out of order (basically, the problem seemed to be that the Working Group would have been chartered to develop specifications, which instead must be done via the RFP-submission process, as determined by the individual Task Forces). Cory Casanave commented that he felt that the Architecture Board should consider the extent to which OMG technology should be based on UML.

During the meeting, there was a lot of discussion between the CBOF and IBM submitters, looking for a way to merge their submissions. However, it appeared that nothing came of this. There was also some activity behind the scenes towards organizing a "no" vote for all of the submissions, should they have come to a vote at this meeting. This appeared to be based on a feeling that it would be premature to lock in specification of business objects in any of these frameworks based on current experience, as well as a lack of general understanding of the submissions by non-submitter representatives. This is consistent with Dave Zenie's observation reported above about the need for the rest of OMG to get enough understanding of the ideas behind the BOF submissions to be comfortable with them.

CORBAgis SIG

Shel Sutton (MITRE), the chair, reported on OGC (Open GIS Consortium) and ISO TC 211 activities. He noted that OGC has "task force" status with OMG, since OMG has agreed not to set up its own GIS TF. The role of CORBAgis is basically to "CORBAize" OGC specifications. He also noted that a GIS metadata SIG was being formed within OGC; they will work on a profile of a TC 211 metadata specification. OGC is also considering a Defense and Intelligence SIG (not restricted to U.S. members). Shel also noted that NIMA had a new glossary of GIS terms which was available to OGC and TC 211. I gave my talk "Towards a Web Object Model" (the same talk I had given to the Internet SIG--see the Internet SIG meeting minutes), emphasizing the potential roles of XML, RDF, and DOM in addressing GIS-related issues. Following the talk, there was discussion of how to include various standard file transfer formats currently used in GIS within the general Web framework without the need to reformat the data in XML. I noted that this would be possible using the DOM as a general interface, since the difference in actual implementation could be hidden by DOM-defined object interfaces (and typical structured file interface formats could readily be interpreted as if they were encoded in XML).

NCITS H7

The NCITS (National Committee for Information Technology Standards) H7 committee met November 29 and 30 (the Saturday and Sunday before the OMG meeting). Attendees were (off and on) Joaquin Miller (MCI SystemHouse, and H7 chair), myself (Frank Manola), Rainer Kossmann (a consultant), Roger Burkhart (Deere), and Haim Kilov (Merrill Lynch). As noted in earlier reports, H7 is the U.S. group working on the Enterprise Language for specifying the Enterprise Viewpoint (EV) 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).

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 EV (this had also been discussed at the previous meeting). Much of the EV is concerned with describing the obligations and other "contractual relationships" of various parties and components within both the Enterprise and its environment. I repeated my concern (expressed at the last meeting) that this sort of formal mechanism could become a "black hole", sucking up committee time that could be better spent getting a clear idea of the actual concepts required for the EV. The math to describe these concepts, I felt, could be better determined as a second step.

The second topic was the preparation of a list of concepts need for the EV specification, as a U.S. contribution to a January ISO meeting. This involved going through the concepts defined in the current (very sketchy) EV description, and those suggested by other national bodies, and trying to come up with a minimal set (for example, the UK had suggested a whole collection of concepts from contract law, containing much redundant material). I suggested that some general reification concept might be appropriate, in order to allow talking at the meta-level about various "object"-level concepts without having to introduce separate meta-level concepts to do so. Joaquin Miller indicated that he would continue to work on this list following the meeting, and would send me a draft for review prior to the ISO meeting.

I reported that I had sent the revised version of the H7 Object Model Final Report to NCITS for review, and was waiting their comments (which arrived after I returned from the OMG meeting). In addition, NCITS wanted the results of a formal vote to approve the report; Joaquin Miller said he would put this process in motion.

Joaquin Miller reported that he was going to be the leader of the Object Model Working Group (OMWG) of the OMG Architecture Board's ORMSC. The OMWG has the charter of rewriting the object model section of the OMA Guide, and dealing with various object model issues currently on the table within OMG (e.g., multiple interfaces, and various changes to the type system).

Later in the OMG meeting, I had a discussion with Tom Rutt (Lucent, NCITS T3 rapporteur to ISO) on the subject (suggested by T3 at the last meeting) of merging H7 and T3 to concentrate all work related to RM-ODP in one committee (I'm in favor). The move is popular within T3 due to lack of participation. Tom said that ISO was going to kill further RM-ODP work due to the lack of a Secretariat for it. I suspect another reason is because RM-ODP per se does not need much further work; further related work should primarily be directed toward specific standards conforming to RM-ODP guidelines (such as OMG's activities; and I think OMG might want to consider providing a "home" for any such activities).