Once a lone voice in the wilderness extolling the virtues of Enterprise Service Bus (ESB) technology, Sonic Software now finds itself with lots of company. The idea of ESBs is to offer a standards-based integration platform that brings together service-oriented architectures (SOAs), messaging, Web services, and XML with a distributed deployment model. Perhaps the most vocal evangelist for this technology has been Gordon Van Huizen, Sonic's CTO, who has played a central role in developing his company's flagship product. Van Huizen recently spoke with InfoWorld Editor At Large Ed Scannell about the still evolving competitive nature of the ESB market and where his company's technology fits in between the offerings of IBM and Microsoft's still off-in-the-distance Indigo technology, which will be stitched into Longhorn.
InfoWorld: As a company heavily invested in Enterprise Service Bus technology, how do you view Microsoft's upcoming Indigo technology?
Van Huizen: We find Indigo really fascinating from an SOA [Service Oriented Architectures] perspective. With enterprise SOAs there are two main aspects to consider. There is the notion of an application as an end point and how should it be thought of within an SOA in terms of what sort of interface should it have and how should it behave. Then there is the infrastructure for tying together a bunch of applications. [They] tend to be fairly distinct areas of endeavor. Historically, the application-centric vendors like Microsoft and BEA have a programmer-first view of SOAs. Instead of changing the interaction model or interface model to something that would be optimal for connecting a lot of systems together, they just kept the traditional client/server model in place and made that Web services oriented. So we anticipate Microsoft continuing on that path. But Indigo actually represents a really fundamental shift in terms of what the interface style should look like. But it really is nothing more than a communications sub-system.
InfoWorld: Isn't Microsoft focusing a lot of attention on various collaboration and communication improvements with Longhorn?
Van Huizen: Exactly. I think they have correctly identified that the way you carry out these things really matters. This has significant architectural implications. Instead of the same old same old client/server-based RPC Web services model, with Indigo they are moving to an event-driven, message-driven, loosely coupled services-oriented model, which we think is vastly superior. We think it is really a key to architectural flexibility.
InfoWorld: Doesn't Microsoft bundling this capability into the operating system worry you?
Van Huizen: You have to be careful not to view Indigo as anything more than what it is. It does represent an important architectural shift in how you think about a Web service. But it doesn't really bundle in capabilities you would normally associate with messaging brokers and middleware.
InfoWorld: You don't see your ESB and Indigo as competing development platforms?
Van Huizen: We see them both as platforms, but platforms of different types. Windows [Longhorn] with Indigo as part of the stack, a first class apps platform. We view the Enterprise Service Bus as being an integration platform. The [set of] requirements for middleware that connects lots of systems together is really a different set of requirements than hosting an application. Time will tell whether Microsoft becomes competent at developing a distributed infrastructure. It is just something they have not done a lot with historically. BizTalk Server is there, but it is not really any sort of a distributed platform for connecting hundred or thousands of applications together.
InfoWorld: How close a relationship do you have with Microsoft?
Van Huizen: We have a healthy relationship with the technologists. Instead of using a proprietary interface to connect into their messaging broker, for example, we are working with them collaboratively on open protocols that would not require use of a proprietary API. So the direction we believe in, and Microsoft at least seems to be going along, is interoperability over the wire. I think some of the early predictions around Indigo were misguided. I don't think they will promote it as an ESB. They certainly could adopt an ESB strategy, but it wouldn't involve just Indigo. What Indigo does give you is a really good SOAP stack for asynchronous, loosely coupled interactions.
InfoWorld: On the other side of you is IBM. What do you make of its ESB strategy?
Van Huizen: To be fair, they make a reasonable amount of money from their existing technologies, so why would you cut that short while you are in the process of building out this new approach? The price tag and services component and all the other stuff surrounding more traditional integration technologies is pretty substantial. ESBs and SOAs, in general, promote a very different sort of value proposition. I think they will be very careful about how they transition into that.
InfoWorld: So you aren't worried about being the little sapling in the ESB forest and getting your sun blocked out by Microsoft and IBM?
Van Huizen: Well, the more serious IBM gets about this technology and the more pressure they feel [from] the market to move in this direction, the more there is for us to worry about. But what they will do for the foreseeable future is -- and I think the announcements they made over the past few weeks validate this -- they will incrementally move in the right direction. They will smooth over the gaps through statements of directions, lots of white papers and consulting services, until they have the real deal. And the real deal is a very hard thing to build.
InfoWorld: Longhorn is a very ambitious project technically. Has Microsoft bitten off more than it can chew?
Van Huizen: It is possible. You have to constrain the problem domain in some way to get releases out. It is a hard set of things to solve simultaneously. The area that IBM has explored for much, much longer is: How do you broker information across machines. They have invested way more in the way of On Demand strategies. When you look at the specifications they tee up for Web services, IBM works more on the distributed side and Microsoft works more on the single machine, application framework side. Microsoft is good at building coherent frameworks for building applications as well as developer tools. What they have not been as good at is getting off of the [stand-alone] box and thinking more from a network perspective. In an enterprisewide SOA, what you are really talking about is lots of boxes with lots of apps using a shared architecture and being able to communicate. This is a thing Microsoft historically has had a tough time with. So when you look at the stuff that is scaled back in WinFS….. it [WinFS] will exist but only for a single machine. Right now there is no network file system in WinFS.