Middleware Matters

The latest versions of each of the two magazine columns I write are now available.

My latest "Toward Integration" column, entitled "WS-NonexistentStandards," discusses the WS standardization situation. In WS-* we have lots and lots of WS specifications, but we unfortunately have very few actual WS standards. You can read either the online version or the PDF version. (Note that the column name as printed contains a space character due to an editorial mix-up -- please just ignore the space.)

My other column, "Object Interconnections," is one that I've been co-authoring with Doug Schmidt for 10 years come January 2005. (That has to be some kind of record!) In this issue we continue our multi-part coverage of the CORBA Component Model. As is traditional in this column, we provide lots of details and code examples. If you missed the earlier parts, you can find them on my home page.

The 2nd International Conference on Service-Oriented Computing (ICSOC) will be held next week, November 15-18, in New York. I have two jobs at the conference. One is to serve as industrial co-chair, which means I will chair the industrial paper session on Wednesday.

I'm really looking forward to my second job, which is to moderate a panel on Thursday November 18, 2:00-3:30pm, about real-world service-oriented systems. We've assembled a panel of industry people, each of whom has significant experience in delivering real-life, actual, working, production-quality service-oriented systems. These guys will be able to talk about what actually works, what doesn't work, and lend us their perspective and advice on where they see things going. The panelists are Mark Davydov of Bank of America, Thomas P. Kozempel of Verizon, Ali Arsanjani from IBM Global Services, and either Mark Luppi or Hugh Grant from Credit Suisse First Boston. Maybe you've heard of them, maybe you haven't. If you haven't, it's likely because they're too busy actually building working stuff. Come along to the panel and bring your hard questions!

On Wednesday, November 10, at 11am EST, I'll be giving a webcast entitled What's New in Orbix 6.2? In it, I'll talk about the new release of Orbix 6 coming out in December, and will talk mainly about Orbix enterprise features, including security, high availability, configuration improvements, system management, new platforms, and extended support. A number of the lead engineers that worked on these features and extensions will be on hand with me, as will one of the Orbix product managers, to answer any questions you might have about the new release.

Go on, register so you can join in and listen. See you then.

I just returned home from attending ICSOC in New York. I wrote some of this as I was listening to Tim Berners-Lee's keynote. I've heard Tim give a similar talk to this one before, at WWW2003, but this version is better and fairly compelling. He's looking back over the development of web services and is presenting pros and cons for WS design approaches at different levels. It's quite interesting to step back and look at the big picture that way. Of course, with Tim it all leads up to the promotion of the Semantic Web as the right answer. His slides are here.

Tuesday night after the conference was fun. My friends Joey Hudoklin and Rob Fried are a guitar duo called Bluefish, and they play acoustic blues and ragtime (ala Hot Tuna). They are damn good. They play regularly at the Alphabet Lounge (Ave. C and 7th St). On Tuesday night I sat in with them on harmonica, and it was a blast. It was a small place with a small crowd very familiar with Bluefish's music, and in such a setting, the potential problem with anyone sitting in is that they might change the music so much that the regular fans dislike the result. Fortunately, that didn't happen. The regulars agreed that the addition of the harmonica definitely enhanced the music. I haven't played very often in public, so it was nice to get that kind of positive feedback. Now I have to find reasons to visit New York more often, so I can sit in with Bluefish more regularly.

I've mentioned the following here before, so forgive me for repeating myself, but I can't say it often enough: Joey and Rob are also some of the best frisbee freestylers in the world, with Joey being widely acknowledged as the greatest freestyler ever. How mint is that?

Unfortunately the late night Tuesday meant missing Adam Bosworth's Wednesday keynote -- doh!

In the afternoon today I was the moderator for a panel on industrial service oriented architectures. My panelists were Hugh Grant from Credit Suisse First Boston, Thomas P. Kozempel from Verizon Communications, and Francisco "Paco" Curbera from IBM Research (T.J. Watson) whom I pressed into service after one of the panelists didn't show up. They were all excellent. Thomas in particular gave a great overview of what they're doing with services at Verizon, how getting there required a few non-technical and cultural changes, and how they're now much more able to measure return on investment. It was a great SOA success story. Hugh had some good insights too, and spoke in detail about the many shortcomings of UDDI for his organization. And despite being a researcher, Paco is definitely in touch with what the real world needs in terms of SOA and web services, so I'm glad he agreed to join the panel at the last minute.

Overall, the conference was pretty good, and I recommend it for researchers and practitioners alike. I had a variety of interesting conversations with people I normally would not see at the middleware and distributed systems conferences I typically attend, including finally getting to meet Sanjiva Weerawarana in person. (He also introduced me to Tim, whom I had not met before either.) A number of these conversations have centered around the notion of multi-protocol or multi-access services, which I've worked on now for over a decade. I'll write more about those conversations in a separate blog entry.

Take a look at your business card. Along with your name and title, it lists a number of ways you can be contacted: your phone number, fax number, mobile number, email address, mailing address, home page, and perhaps your instant messenger ID. Someone who wants to reach you can choose any of the methods listed on your card, and despite the fact that those contact methods are quite different with respect to the qualities of service they each offer, they all lead to you. You're basically the same person regardless of which method someone might choose to reach you.

I sometimes use the business card as an analogy to discuss the merits of web service references. A multi-port or multi-access reference type is one of the issues currently being addressed by the W3C WS-Addressing Working Group (WG), of which I'm a member.

The WS-Addressing endpoint reference (EPR), as currently defined, allows access to a service over only a single port or mechanism. Using the business card analogy, that's like mandating that a business card can specify only a single means of reaching the person it references, thus forcing anyone who wants to be reachable over multiple means of access to have to issue multiple business cards. Can you imagine attending a business meeting where you have to give business cards to five different people, and you want each of them to be able to reach you over desk phone, mobile, and email, so you wind up having to give out three separate cards to each person, 15 cards in all?

As Rich Salz points out in his review of WSDL 2.0,

In WSDL 1.0, a service could implement multiple ports. This fits with the conventional model for implementing IETF/Internet servers: each TCP port defines its own language (SMTP, NNTP, etc.), and a single executable can implement multiple protocols, by using the port number to de-multiplex (or dispatch) to the appropriate handlers.

As Rich implies, the multi-port service model is nothing unusual -- it's been around for years. A service might offer itself over multiple ports so it can offer different qualities of service to its consumers. For example, it might accept compressed messages over one port, encrypted messages over another, and perhaps management messages over a third. Alternatively, using capabilities like those provided by our Artix product, a service might accept CORBA messages on one port and SOAP messages on another.

Is a service reference, rather than or in addition to an EPR, something that the WS-Addressing standard needs to provide? I think it should.

More on this topic later.