Todd Biske: Outside the Box

Yet another good discussion from Dana Gardner’s group of independent analysts at Briefings Direct:SOA Edition. In this edition (podcast / transcript), the panelists (Dana Gardner, Steve Garone, Joe McKendrick, Jon Collins, and Tony Baer) discuss the year in review and their predictions for 2007.

The first comments that I liked were from Jon Collins of MacehiterWard-Dutton. In commenting on the consolidation in the vendor space that occurred in 2006, Jon points out “the one thing that’s been lacking so far … is integration, which is to me the ultimate irony, because it’s exactly what SOA is about as a concept.” Well said. He goes on to state that they need to become service-oriented in terms of how they put together their products. What I find great about that is that this trend toward the superplatform will make vendors think about their management interfaces. After all, this is where the integration must begin. If the integration points don’t exist, or are only available via a user-facing console, the integration cost will be higher and take longer. Prior to the superplatform, this expense was incurred by the end consumer. Now, let the vendors take that charge. Let’s hope it’s done properly however. All too often, they perform this integration, but then fail to expose these services to the end consumer. Back when I was working for a large enterprise, I had a vendor come in and tell me how their management console was built using their portal product. I said, “That’s great! Are the individuals components exposed as JSR-168 compliant portlets so I can built my own custom console with portlets from other products that operations needs?” Their answer, “Well, no…” Let’s get it right this time!

I was surprised at Tony Baer’s comments on the convergence of the Registry/Repository space. The only explanation was his comment that “at runtime, you don’t want a repository with a lot of overhead.” I’ll agree with this, but the reason for the convergence was not runtime, it was design time. Perhaps this was more intuitive to me because of the research I had done into component reuse in the 2000-2001 timeframe, and I understood the importance of metadata to support it. I think it’s only now that enterprises are starting to figure out an appropriate role for the registry/repository in the run-time environment.

In the predictions portion of the podcast, Joe McKendrick made the boldest statements, stating that “SOA as a term has crested.” He went on to discuss event-driven architecture, consistent with one of his recent blog posts, correctly calling out the ability for it to integrate with business intelligence solutions. Personally, I think an enterprise that is able to leverage SOA, EDA, and their BI systems is in the upper echelons of SOA maturity. Joe was also the most conservative regarding the rate of SOA adoption, expecting only a 20% increase from 2006.

Tying the vendor consolidation comments with Joe’s comments on the cresting of SOA, I think it is true that the superplatform vendors need to position their solutions not as enterprise SOA infrastructure, but enterprise infrastructure. As Jon suggested, a service-oriented view helps break down the infrastructure needs into capabilities. These capabilities aren’t specific to SOA, however, they are enterprise needs. For SOA to be successful, it does need to get to point where it is simply what we do, rather than being some special initiative within IT while the rest of IT continues to operate the same way it has in the past.