OMG Trip Report

Around 200-250 people attended this OMG meeting. OMG now consists of the following groups:

Architecture Board, Object Model, Liaison, Metrics, Platform Technical Committee, ORB/OS TF, Common Facilities TF, Security SIG, Internet SIG Domain Technical Committee, Business Objects, Transportation, Telcom, Healthcare, Manufacturing, Electronic Commerce, C4I Working Group, CDIF, Finance, OA&D, Document Management, Real-time, (GIS)

OMG 1996 Buyer's Guide. This guide is now available and lists a lot of products. Got a copy. Also, there is a new book on CORBA case studies.

The CORBA Buyer's Guide contains descriptions of a number of interesting products, including: The Enterprise Engine, which "allows business designs to be specified using natural language to describe policies and procedures. These designs may be distributed and executed globally over the Internet. <http://www.engines.com>, Fujitsu's File Object Service (FOS), which "provides uniform access to file objects in a distributed environment of PC, UNIX and mainframe systems", Some business object architectures: SES Holdings' BOMA <http://www.sesh.com> and SSA's Newi.

OMG Internet Platform SIG. Craig Thompson (OBJS) chaired a one-day meeting. Thirty three people attended. See <http://www.objs.com/isig/mtg07.htm > Minutes of Meeting #7 (coming soon) (http://www.objs.com/isig/mtg07.htm ).

OMG GIS Domain SIG. About a year ago, Shel Sutton (MITRE), co-chair of the Internet SIG and also chair of the Open GIS Consortium's Earth Imaging Subcommittee, tried to form an OMG Geographic Information System (GIS) SIG, before OMG had begun its emphasis on domain technical activities. A few weeks prior to this meeting, Thompson suggested to Sutton that the time is right to try again. At this OMG meeting, Thompson and Manola participated in the formational meeting of a new OMG GIS Domain SIG. We reviewed Sutton's mission statement and discussed how to funnel the needs of various OMG domain task forces (e.g., healthcare, transportation, telecom, manufacturing) via the SIG to OGIS and how to make OGIS results more widely available to industry through OMG.

Change Management RFP. There is interest reviving in OMG Common Facilities for a Change Management/Revision Control Facility RFP. An email list is forming and we will plan to contribute requirements and a discussion. We were the original authors of the Change Management Service description that appears in the OMG <italic>Object Services Architecture</italic> and also contributed several years ago to OMG a TI Open OODB specification on change management.

Chairs' Dinner. Thompson attended the Chairs' Dinner. Discussion was on meeting mechanics, publishing timely agendas, three week document rule, better intergroup coordination. The need for these additional "procedures" is due to the increased complexity of OMG as it grows in coverage and subgroup partitioning.

Plenary Talk on ILU. Bill Janssen (Xerox PARC) presented on "ILU: A Meta ORB." Notes - may not be readable. IIOP (OMG's on-the-wire CORBA protocol) is a good protocol that could do better. [I overheard Soley say that only one ORB uses IIOP as its native on-the-wire protocol.] ILU is now five year's old. ILU commits as little as possible with an eye to moving in different directions. Design goals: multiple programming languages (C, C++, Modula-3, Common Lisp, Python, Java-wide range), layered on a runtime extensible set of RPC protocols and XDRs (RFC 1831, XNS Courier RPC, IIOP, HTTP),runtime extensible set of transport protocols (TCP/IP, UDP/IP, XNS SPP), interoperable with existing clients and servers, OS portable, free CORBA, put multiple languages into the same address space. Needed language independent network specification language, the original still in use was based on Modula-3. In an application, Application in language X - talks to - ILU-generated stubs - talks to X ILU runtime - ILU generic functions - RPC - multiple transports. DII in CORBA is mainly applicable to C or perhaps C++. Try to work with language runtime. Current design goals for RPF protocols is to assure safety. At the RPC level, there are no significant differences between DCOM and CORBA so they will add DCOM soon. Transport stack = security layers, compression layer, &#133; Transports defined by meta object. Could add queuing as another transport at this level. Can interact with Sun calendar, Framemaker RPCs, &#133; other legacy systems' RPCs to access native RPC protocols. Different optimizations for same/different languages on same/different machines. So same language gives up no runtime. Mixing C and Python is becoming popular in PARC. Another win was to use standard C due to the chaos in C++ systems (making C++ not portable). Runs on UNIX and Windows. ILU is a free CORBA ORB. ILU Object Model: pretty close to CORBA but supports asynchronous. From the beginning, supporting RPF and messaging models. OMG grief comes from unions, naming, optional objects. Trying out ideas: network garbage collection, unicode, MOO, TCL, bulk data transfer/streams, instance variables with client-side caches. No DII, interface repository, implementation repository, partial BOA, variants instead of any. Interface Description Language uses IMPORT instead of #include; everything defined in a module scope without nesting. Type system: ISO Latin-1 and unicode, 64-bit signed/unsigned ints, 128 bit floats, optional types (untagged union with NIL Network GC. Asynchronous (a marking on a) method (messaging). Users use this extensibly. Digitial libraries and Web Consortium use this. Current release is 2.0alpha8 (July 1996). Over the last year, they have doubled staff to 7 people.In use at Tivoli, Bayou, Stanford, U Michigan, U Illinois, Digital Creations, others.

Plenary Session on DSM-CC. (Handout) Don Hooper, DEC, Digital Storage Media Command and Control. Notes - may not be readable. Using CORBA IDL and IIOP. Used by digital media community, carousels of video. Operations like PAUSE, RESUME on digital media. ISO/IEC 13818-6. Using IDL as interface instead of RPCs to wrap MPEG streams. This work is now an International Standard. Uses IDL, all language bindings for C, C++, Smalltalk, Java, etc, uses Module IOP for Interoperable Object refs. User-network and user-user operations. Uses Service Interoperability Interface (SII) with a collection of ways to connect local objects to network objects. Can issue access control on individual objects. Lucent, Oracle, DEC, heavy involvement from Sun, 200 Dadac, Ninex, &#133; set tops or PCs.

During ORBOS: CORBA Security Revision Task Force. Lots of members. Comment period is open until April 1997.

Financial Domain TF presentation to Business Objects TF. Review of Financial Architecture and the working group structure they are setting up, how they relate to OMA and Business Objects. Also discussed how banks evolve legacy systems by moving to CORBA gradually. An issue is, how to you partition the space of domains (as it is forming) between many different domains and sub-domains. Frank noted that ODP viewpoints might be a model for separating areas. This is an important issue in evolution since the floor, ceiling, sides, and insides of a domain might continuously evolve. Another approach is for overlapping groups to have joint submissions for certain overlapping work.

Business Objects Domain Task Force. Comment from floor: Since BOFacility has immediately issued an open-ended RFP (too broad and open-ended) which leads to problems that people don't know the scope. There is no laundry list, no architecture, therefore no way to know what the scope and order of adoption will be. And responses to the BOF will be all over the map. Another comment - maybe we are standardizing potential interactions rather than interfaces to get reuse and avoid brittleness.

Listened to discussions on Evaluation of BOFs via a small object model the submitters agree on. There are 21 LOIs received so far.

Cory Casanave described the overall "vision" for Business Objects, and noted that the conventional "object" concept was too small a unit to be useful at the enterprise level. The idea was to create a different level of abstraction at which business developers could work. He said he expected that a Business Object Facility would, in products, be built to not require a CORBA ORB (e.g., it could use DCOM too). Discussions with the Financial DTF noted the issues of trying to manage the many potential overlaps between the activities of the BODTF and TFs associated with specific business domains (such as Financial). For example, the Financial DTF had produced its own "financial architecture", which needs to explicitly reference the BOF. A motion was passed to form a working group to (along with any other interested domain groups) address the issue of harmonizing these architectures.

There seems to be an issue within the BODTF as to whether the BOF will be a real backplane (with a lot of semantics built in), or more like a board that just provides a basis for wiring business objects together.

Business Objects Domain Task Force presentations by X3H7 and X3J21. Glenn Hollowell, Chairman of ANSI X3H7 (ODP Enterprise Viewpoint), reported on the status of an X3H7 proposal to ISO (the International Standards Organization) for X3H7 to work on an Enterprise Viewpoint (EV) component standard as part of the ISO Open Distributed Processing Reference Model (RM-ODP). In doing this work, X3H7 would work in cooperation with BODTF. The basic goal is to provide a refined definition of enterprise language terms in the RM-ODP, explaining the relationship of an enterprise specification of an information system to other viewpoint specifications of that system, with the aim of enabling the RM-ODP to be used to specify object-based application architectures. Another goal is to firm up various mappings between parts of RM-ODP and parts of OMG's OMA (OMG explicitly specifies that its specifications are supposed to conform to RM-ODP). The relationship with the BODTF is that current ideas for these mappings have Business Objects being described in the Enterprise Viewpoint, and the Business Object Facility being described in the Engineering and Computational Viewpoints (there is expected to be a future ODP work item explicitly addressing these mappings). Arrangements are being made to put the various parts of the RM-ODP specification on the OMG server, but there are some obstacles being placed in the way by ISO bureaucracy (in fact, Postscript copies can be downloaded from ftp://ftp.dstc.edu.au/pub/arch/RM-ODP/).

The meeting heard two presentations organized by X3J21 (ODP Information Management), one by Randolph Johnson (NSA) on deontic logic, and another by Haim Kilov (IBM Watson Lab.) on specification of business systems. Deontic logic is a modal logic that formally describes notions of obligation, permission, and prohibition. Deontic logic (and similar concepts) have been investigated for some time in connection with various high level information models (e.g., conceptual schemas), hence the possible connection with business modeling and business objects. Kilov's presentation discussed more precise ways of specifying business semantics, the use of specific types of relationships between objects as specified by the ISO General Relationship Model (e.g., different kinds of composition and containment relationships), and the use of design patterns in business models (from a recent paper "Business Patterns: reusable abstract constructs for business specifications"). Some of Kilov's work seems relevant to work on higher-level architectural specification techniques.

Architecture Board-Open issues in Call by Value: preservation of identity, equality, behavior; type safety (exact match or subscriptions allowed), locality (same or different language local or remote). Ramifications: pseudo objects - will they remain necessary, types - can we deprecate any, enum, union, struct, decimal, can we freeze them once and for all. Object services name hiding, &#133;<bold> </bold>

Rules Service. At the last OMG meeting in Madrid, I presented a draft Rules Facility RFP (which I had authored for NIIIP). At this meeting, I arranged for Herman Lam (U Florida) to present the NIIIP ECAA Rules Service to the Manufacturing Domain Task Force, OA&D Task Force (led by James O'Dell), and Architecture Board. The presentation was well made and a copy (will be) on the OMG server. Some of the issues raised included:

• where are rules executed? On server or client? how does that affect performance?
• how can you use rules to configure components without code?
• how are workflow rules related to ECAA rules
• what optimization opportunities are there in a rules service?
• how do you specify how rules compose in case several events are related to one object?
• rules visibility - should rules be encapsulated within a class definition or are they separate specifications and visible outside the class or both? How do you hide rules used within a system from rules that application developers might use?
• what other kinds of rules systems are there and how do we facilitate and not preclude this possibility? Should we use ECAA rules to represent invariants, business rules, condition-action rules, expert system rules, OPS5 rules, Prolog rules, deductive DBMS rules, pre/post conditions, state transition diagrams, trigger/action pairs
• how dynamic are rules-can you add or remove them at runtime? Can anyone do this?
• what happens if rules raise exceptions. Do you need a rules exception spec? what happens if there is a need for rollback? Is this related to transactions?
• should there be a separate syntactic specification for before/after methods?
• is the state of an object is modified, can we specify rules that get triggered?
• if we subclass legacy classes then can we add rules to the derived classes?
• namespace clashes for rules - how to avoid conflicts (A: same way as in IDL)
• do we need trigger references on ways to combine rules?
• we don't always want to relate event creator to event response so need to separate these two
• comment - adding rules makes systems more agile
• how is telecom notification RFP related to rules - they talk about filters where you talk about conditions. (later located Michael Greenberg, NEC and their RFP draft is on the server should rules raise and lower priorities
• should we view rules as a language mapping RFP?
• should events be tied into timing (real-time timing events or beyond that real-time constraints
• how to enable/disable methods - do you do this in a global or local context. Should you use policies and policy regions for this?
• what is the rule action language? C, C++, something you made up? Can you use any programming language you want (like IDL permits) or do you have
• why not just use rule-objects
• if you change rules, do you have to re-test - regression test everything or some things how does this rule service relate to object services such as security service's general Interceptor specification. Does it check security-one should have to address this
• shipping rules is passing behavior by value, like Java
• this is defining a meta-object protocol. That it is in the form of a rule is separate.
• some of this we already have in the security interceptor architecture
• your rules management looks like code management. It looks like an interface repository for rules.
• another area dealing with this is mobile agents
• how would you write a rule to sum a set of b instances into a total, whenever a b-instance changes.
• how to state an integrity constraint
• Jeff Lewis - something that complicates the object model semantics is probably bad. Your method invocation is an method that invisibly kicks in. It make be more convenient to make it implicit but makes it more complicated.
• Jeff Lewis - reuse IDL and not new language
• is the rules language (partial language, twisted technology) a special case of a scripting language business rules separate from the code is a well established requirement. These ECAA rules are narrow but might be broadly useful
• OMG standardizes interfaces and not languages
• should the rules service be treated as three RFPs-one for a rules language, another for a rule management service for adding, deleting, and modifying rules, and the third for ???

Telecom DTF. Trisha Johnson (HP and NMF) gave a presentation on Network Management Forum activities. NMF is a consortium of organizations involved in service management, with a focus on service providers. NMF output consists of:

• SMART - a set of business agreements for process integration; this is based on a common Service Management Business Process Model that can be used as a basis for alliances between service providers
• OMNIPoint - sets of technical specifications of service interfaces (e.g., interfaces for specifying trouble-ticketing); these specifications would use IDL for CORBA-based implementations
• SPIRIT - a scalable distributed computing platform

She said that Internet Service Providers are starting to sign up as members, and they will probably drive some of the work on new services. She also mentioned that NMF had produced analysis and design documents which used ODP viewpoints, and implementation documents which mapped to the ODP Engineering Viewpoint. (Manola reported this to X3H7 as an example of viewpoint usage they may want to build on). These documents use a modified version of OMT as a neutral (graphical) notation, and they are working on translators to GDMO, SNMP, IDL, and SQL. She said that defining a set of common object specifications is currently a major activity.

Discussion indicated that there were a number of possible ways for OMG and NMF to cooperate. Colin Ashford (Nortel) noted that there was a limit as to how far modeling activity should go; rather than defining endless interfaces in IDL based on a single model that was claimed to work for everyone, it would be better to provide tools to allow bilateral aggrements between vendors and consumers. Tom Rutt (AT&T/Lucent) said that the Telecom DTF should focus on things not already done or being done by others, and that work on an architecture and object model for Telecom would be redundant.

A number of documents were also discussed. These included:

draft Notification Service RFP, draft Topology Service RFP, a white paper on CORBA-based TMN (Telecommunications Management Network), a white paper on "Using CORBA for Session Management in ATM Networks", a white paper on "Intelligent Networking with CORBA", a white paper on "Issues in Designing Scalable CORBA Applications" (a paper which described HP's experience with using CORBA to build applications dealing with very large numbers of objects, and argued that scalability issues in CORBA are not due to inherent shortcomings in CORBA technology, but rather to how it is applied).

ANSI X3H7. Most of the meeting was spent editing the proposal to ISO to approve the new work item on developing an RM-ODP Enterprise Language, and drafting the official U.S. response to the proposal. Frank Manola (OBJS) contributed to the wordsmithing of both these items. Some clarification of the proposal was needed because the current wording made it sound like a proposal to work on an application architecture in addition to an Enterprise Viewpoint language (the intent is work on further describing the relationship of the Enterprise Viewpoint to other viewpoints of the RM-ODP architecture, not develop a new architecture). Haim Kilov submitted a paper on how to represent business rules in the enterprise viewpoint, and there was discussion of procedures for fitting such technical submissions into the draft outline.

Sridhar Iyengar (Unisys) gave a presentation on the MetaObject Facility work he is heading. They are focusing on the object analysis and design domain, and so are working on a MOF for OA&D metamodels, with work partitioned between the MOF and OA&D TF so as to coordinate.

The next X3H7 meeting is scheduled for December 6,7, in Boston, at which point they expect to have a number of submissions toward parts of the Enterprise Language standard.

Issue on OMG spec evolution. What do we do if we later add "pass by value" or "rules" later after some other specs should have depended on it? Separate issue: sometimes we see an apparently related issue that is not one since it is used in a different context. That may be true for security's interceptor specification and the rules service.

Object Model Subcommittee.

• Object model for services WG - what is the OMA framework-they may be candidates to review the componentware glossary - a draft paper is available (this discusses the general issue of describing "object frameworks": specific configurations of objects that provide services)
• PIDL and pseudo-objects WG - lets get rid of PIDL - green paper is available
• Type System WG - a green paper is being prepared
• ODP relationship with OMG Object Model - this involves work on identifying correspondences between ODP and OMG concepts, and generally carrying on liaison between OMG and ODP-related activities
• Object Model Architecture WG - nothing so far, wait for Nice

ORBOS. Issued an SSL RFP. Presentations on Multiple Interfaces and Composition RFP responses.

Common Facilities. Reviewed Sankar's and Dan Chang's Mobile Agents submissions. It recommended the TC to adopt Tom Mowbray's Data Interchange Spec (but the AB later voted to table tis to Nice.

Reception/Lunches/Contacts. Talked to:

• Bob Hodges (TI) who is going to SEMATECH for the next two years, a more stable environment sinceTI software has less money due to the tight semiconductor market.
• Dave Zenie (NIIIP) who is doing the Workflow Management Consortium IDL interface for NIIIP-he tells me some of the WfMC interfaces are more conceptual than well specified. Also, as NIIIP winds down, NIIIP Smart and others wind up but it is not clear that they will use much NIIIP reference implementations.
• David Hartzband (Riverton Software Corporation, Cambridge, MA), really a smart guy, used to be in charge of DEC ObjectBroker, was employee #5 in a startup now 1-2 years old with 34 employees living in Powersoft world and delivering their first product. The thesis is that companies do not want to put logic in DBMS stored procedure calls and need a middleware layer. They do not do DBMS stored procedure generation but understand the problem. He is their software architect-I introduced him to Frank Manola. Hartzband wants to keep in touch. Good.
• Denis Bagsby (Southwestern Bell), connected into Erik Mettala/Barry Leiner's MCC projects. Open OODB and some SW Bell technology forms the core of a project and they are looking for a project lead. TI is a member. Need to call Erik since this may be an OBJS opportunity.
• Neil Wagoner (Lead Security Engineer, MITRE), covering OBJS Virtual Office distributed security needs as well as NIIIP Consortium Virtual Enterprise security requirements.
• Bill Janssen (Xerox PARC), covering the need for a CORBA core extension mechanism. Maybe the general purpose Security spec "interceptors" (need to look at that) are relevant and can be used in the Rules Service
• Tim Brinson (Protocol Systems Inc. tim@protocol.com) is chair of the OMG Healthcare SIG. He was hoping Rusty Lewis (VP HBOC) would attend (he did not). Also, saw Juggy Jaganathan who has now started his own healthcare startup. Also noticed that NIST has issued an Advanced Technology Program (ATP) on Information Infrastructrure for Healthcare, with a public meeting on 9-10 October 1996.
• There was a presentation, which I missed, on Real-time Mach made by Raj Rajkumar (CMU).

Manola talked to:

• Joachin Miller (MCI Systemhouse, and X3H7 Vice-Chair). He expressed interest in Manola's RISC Object Model documents on the GTE server, with the idea of trying to apply those concepts to object analysis and design models. The idea is that these models keep getting more complicated (by having new constructs added to them), and the RISC concept might provide a systematic way of coordinating such extensions.
• Selden Stewart (NIST). He works on manufacturing-related standards, and is involved in ISO TC184/SC5/WG1. Regarding handling configuration information for software.
• Sridhar Iyengar (UniSys). Heading the MetaObject Facility work; we spoke after the X3H7 meeting about some architectural stuff. He'd heard about my RISC model work; we discussed the relationship of the MOF to other metadata sources within the OMA (e.g., the Interface Repository), and when it would be appropriate to separate particular sources (e.g., on the grounds that one type of metadata requires a particularly lightweight implementation). We also discussed the use of MOF to support service metadata (e.g., the transaction models supported by an extended transaction service).

Platform Technical Committee (Summaries)

Since Madrid: these specifications - Collection, Trader, IDL Fixed Point, CORBA 2.0 Revision - all passed the board but some still need the TC vote to pass.

LOIs outstanding on Mobile Agents, COM/CORBA/Part B, Messaging, Multiple Interfaces and Composition.

ORBOS - recommended Realtime RFI for Realtime SIG, recommended Secure Sockets SSL RFP, COM/CORBA part B dates set initial submission Feb 14, presentation from Andrew Watson on CORBA packaging (a marketing opportunity), presentation from Open Group on CORBA conformance test project, ORB Portability RFP Evaluation Criteria, accelerated COBOL RFP, Telecommunications TF Notifications RFP, presentation on Extensible Interop across Security Domains

Common Facilities: briefing on Workflow Management Coalition (Dave Zenie), initial submissions briefed on Mobile Agents, revised Data Interchange submission was recommended for adoption (10-0-0) but vote held up until next meeting,

Other groups reported: OA&D Task Force (OA&D RFP looks like it will get 10+ submissions, Real-time SIG, Internet SIG, Japan SIG. Note that Real-time SIG has formed a QoS working group.

Votes during this meeting: Extend CORBAservices RTF deadlines, Add Visigenic to OROS Technology RTF, Create CORBA Security RTF

Domain Technical Committee. These groups gave reports: Business Objects, Finance, Manufacturing, Telecommunications, Healthcare, Electronic Commerce, Transportation.

A vote to create the GIS SIG was deferred to Nice but so far interest is favorable. Finance RFI-1 Insurance and RFP-2 Currency, Manufacturing RFI-2, and Electronic Commerce RFI-2 were issued.

Healthcare is creating working groups on Architecture/Roadmap, Security, Computerized Patient Record, and Vocabulary.

Next OMG Meetings.

• Nice, Nov 4
• Tampa, Jan 13
• Austin, Mar 10