OMG Java Controversy. The meeting began with controversy. Just before I left the U.S, Bob Marcus published some of my comments on OMG/Java convergence and standardization strategy (e.g., OMG should switch to Java from IDL or at least permit it as an alternative specification language). There was a flurry of email on the 19th and on arriving in Dublin everyone I met kidded me about the controversy. But speakers were defensive, including the keynote speaker Annrai O'Toole (Iona's CTO).
ORMSC (Sunday). I participated in the ORMSC meeting all afternoon. The topic was "identity." OMG has never taken a consistent stand on identity. Most of the OMG specifications implicitly assume a system notion of identity, including a closed world assumption, no different than closed programming languages or DBMS systems. But there is a growing need to more explicitly define identity:
A few other points were made about identity:
Other topics they plan to discuss are location transparency and behavioral specifications.
OMG Internet SIG (all day Monday). See OMG Internet SIG Minutes.
ORMSC (Tuesday). Another session of the ORMSC discussed ODP viewpoints but did not resolve much except to educate the audience.
GIS SIG (Tuesday). Shel Sutton presented two talks in one on the NIMA US Imagery and Geospatial Information Service (USIGS) Technical Reference Model and Open Geospatial Exchange (OGE) Services. NIMA is working closely with the Open GIS Consortium. The OMG GIS SIG is basically a front for OGC at OMG that is effectively a domain-oriented group.
QoS Reference Model Working Group (Tuesday). Chris Sluman chairs this working group. New work in ODP and existing work in OMG needs to be aligned and it is being. The OMG green paper has become a key input (40 pages) to the ISO document (now 57 pages) which is now being rolled back into OMG. The OMG green paper will be about the application of the standard to the ISO environment. The current reference architecture is ISO 13236 - I got a soft copy from Chris. It covers dispatch, management, negotiation. There are some 70 attributes in the spec. QoS has been in existence statically where you buy the RAM, processor, … but now we are moving to dynamic QoS, in a distributed environment, with variations of ownership and control. The ODP notion of viewpoints is fundamental in the QoS RM. The RM-ODP is becoming very general and covers a lot of ilities. There is a thing that QoS applies to, it had an identity. More below.
Plenary Talk: "CORBA Branding Program," John Morris, The Open Group (Wedneesday). The CORBA branding program will launch in January 1998 covering C++ test suites for CORBA 1.0, (1.1 in 1999) covering ORB AI, IDL syntax, IIOP, DII, DSI, interface repository API. They use an ORB proxy between client and server for tests so they can test interoperability of clients and servers from different vendors. Java test suites are still ahead. The brand places a legal commitment on vendors to register, fix conformance problems, and ambiguities in interpretation of syntax.
Plenary Talk (Wednesday): "The World is in a State of Chaos," Annrai O'Toole, CTO of Iona. His point is that with so many platforms and products, the need for glue is inevitable and that's what ORBs provide. He says Microsoft owns the desktop so he is in the server business. He says, Java may be trendy but so were the Osmonds. Still, it is OK to be Java friendly. He claims you cannot seamlessly distribute a normal sequential program, it takes design (yes but … there are many more reasons than distribution to interoperate, for instance, simply crossing language boundaries). He quoted an Irish poet trying to write a haiku (3 line 17 syllable Japanese poem): "to write a poem in seventeen syllables is very diffic …" One interesting point, he thinks the IIOP wire format and future extensions to add transactions, security, etc., e.g., wire formats, is the way to go for interoperability, which puts him squarely in the IETF philosophical camp. He thinks the next step after "Corba is good stuff, hard to use though" has to be frameworks. The web is here to stay since its a utility.
Announcement (Wednesday). We knew Chris Stone had left OMG to become Senior VP at Tandem. So now Richard Soley has been elected Chair, CEO, and still CTO. Bill Hoffman, who ran Object World, is President.
ORBOS (Wednesday). This session was partly about Call by Value and partly about the Java to IDL Mapping.
Jeff Mischkinsky (Visigenics Software, jeffm@visigenic.com) presented the call by value RFP response status and asked for an extension, granted, to the next meeting December 1-5. The current document accounts for externalizing graphs of objects, like Open OODB used to do, but requires a new type added to IDL called value which you apparently inherit from (a means of marking) and it has the copy by value instead of object reference meaning. You will be able to add you own custom value streaming policies and algorithms. This is a prerequisite to one of the Java to IDL Mapping specs.
Simon Nash (IBM, nash@hursley@ibm.com) then presented the Java to IDL initial submission from IBM, Netscape, Oracle, Sun, Visigenic. It seems to provide two separable capabilities under one hood:
Martin Chapman presented Iona's Java to IDL initial submission. He claims the other team is not really focusing on Java to IDL for arbitrary Java but just a subset that has RMI classes. He claimed, wrongly, that we don't really need the RFP since we don't need a standard way to map from Java to IDL, its just a tool. Many did not agree. He said we must not encourage Java as a design center for CORBA. He did a rather poor job of explaining their submission.
Quality of Service Working Group (Wednesday). Chris Sluman ({cms or chris_sluman}@omg.org, qos@omg.org) continued the discussion on the OMG QoS green paper, the ISO QoS work, and the RM-ODP section 5 on QoS. I got hard copies of the first and third so I have an up to date set of these documents. I asked Chris who the key players are - none are US - Jeremy Tucker (UK Logican) and Laurent LeBoucher (France Telecom) - and the three efforts are well coordinated. I asked if the BBN crew, SRI, Teknowledge or anyone else in the US was a key contributor. The BBN crew (Dave Barko and John Zinky) alone have provided academic papers but no reviews of the various standards documents. I asked what other communities are also at work on QoS. Sluman said several: telcos, multimedia streaming standards, IETF RSVP (an insular, hard to penetrate community), real-time, more. He listed user stimulus (NATO, manufacturing), application stimulus, and technology stimulus as drivers.
My next question was scope: QoS is usually thought of as controlling performance of network bandwidth. But the green paper seems to be focused on preserving a host of ilities, as if quality were the top ility with all others as expansions of it for some aspect. Chris generally agrees that the framework is generally intended as reusable (so far as it goes to date) - though, when I asked him about security, he says that security is such a separate community that the QoS guys do not try to penetrate this. I claimed and he agreed that the QoS stuff should abstract out the reusable framework of the security spec. Also, I pointed out that the interpretation he is placing on QoS is the same sort of one I am but is not what the QoS community has traditionally viewed QoS to be, which is much narrower bandwidth negotiation and guarantees. He agreed.
He said their next step is to move from concepts to an architecture for QoS. I am planning to try to help them with this.
By the way, the reason QoS has been performance-driven is that one can isolate just this concern and it is necessary to do so for many purposes so it has been explored.
We talked about tradeoffs and care-abouts and the idea that there is no highest ility for all purposes - for instance one group's concern is safety, another is performance, another is throughput. This partly explains why ODP is in interested in QoS. It is a property of a system of parts that must be guaranteed across the parts. ODP Enterprise Viewpoint is concerned with relationships and constraints and statements one can make across parts of a whole.
By the same token, Chris mentioned that telcos are reluctant to publish any interfaces to their QoS outside their system, just using it internally. I agreed, that federation of QoS is possible technically, but there are business reasons why not to publish these controls. I pointed out that the same is true for componentry in general - that component vendors may be reluctant to publish the metadata and management interface because it exposes implementation details and also capabilities.
QoS in ODP. I read part of this document on the bus. Its reasonable though not quite right in parts.
Joint meeting if Internet SIG and Security SIG (Thursday). Not many came to this early morning session but the few that did had a discussion about how security was just one ility and could be generalized into an architectural properties specification where security is just a specialization.
ORMSC (Thursday). I game my presentation on "OMG meets the Ilities". The audience was small but good. The talk was well received, people seem to agree with most of the points and believe this is a good direction.
Then the working groups reviewed their progress:
ORBOS (Thursday). The Mobile Agents RFP was passed by vote. Though I heard: Crystaliz dropped out early from contributing; General Magic led for a long while then dropped out, seeing no business potential. IBM pushed the spec through the vote. A new RFP is being proposed for a Garbage Collection Service and an Interoperable (Federated) Name Service.
Architecture Board Plenary (Friday). I missed part of this but came in just in time to hear the ORMSC presentation; Jishnu stated that my presentation was important to the OMA future.
Platform Technical Committee Plenary (Friday).
ORBOS - the Internet RFP 1 did not have responders, not surprising since it was boring, wrapping FTP. The GC RFP is deferred until the next meeting. IDL to Java final mapping in January. COM/CORBA Part A and B, UML, MOF - technology adoptions. Multiple Interfaces. Tagged Data ???, Mobile Agents did not make it through the Architecture Board this meeting.
OA&D - Mary Loomis resigned and Sridar is new co-chair. UML 27-0 vote to adopt. MOF.
Real time - dynamic scheduling in 2 meetings; fault tolerance in three meetings; other working groups on performance metrics and realtime comm, time services
UML revision Task Force - created.
Internet SIG - I gave the 5 minute report.
Domain Technical Committee Plenary (Friday).
Lots of DTC activity. The balance in OMG is shifting toward this side! The following groups reported: Finance, Electronic Commerce, Manufacturing, Healthcare, Medical (lexicon query services), Telcom (logging), Transportation (rail, air), C4I ( Tom Mowbray of MITRE is trying to get together a framework and requirements for C4I by next meeting), GIS (Simple Map Features), Distributed Simulation, Business Objects (common business objects).
The business objects RFP on workflow generated a lot of responses, now grouped into three main camps. David Zenie (good friend from NIIIP) was elected co-chair.
The Distributed Simulation is preparing an RFI.
A new Autonomous Decentralized Service Systems SIG was created, though I pointed out it makes more sense as a platform or architecture SIG as described.
A Utilities SIG was created.
A Life Sciences SIG was created and an RFI is planned.
Action Items.