Understanding the Value of Web Services

Intel's Thomas says vendors, IT in danger of missing the real payoff.

What are web services? The answer depends on whom you ask. To major vendors promoting the idea, Web services are self-describing applications that live online and that, using standards such as SOAP (Simple Object Access Protocol), Web Services Description Language, and Universal Description, Discovery and Integration, can be accessed and used by any client.

To others, however, that way of viewing Web services is at best incomplete and, at worst, counterproductive. Intel Corp. Chief e-Strategist Christopher Thomas, for example, said he believes that, by focusing too soon on that ambitious view of Web services, IT managers risk overlooking the real, near-term value of Web services technologies: to solving vexing enterprise integration problems. Thomas explained his views in a recent interview with eWeek Executive Managing Editor Jeff Moad.

eWeek: Where are most IT organizations in understanding the potential of Web services?

Thomas: People are doing two things with SOAP. The first thing theyre doing is theyre wrapping XML as a message. The second thing theyre talking about—which is where most people focus their discussion—is SOAP as an [RPC (remote procedure call)] and a self-describing process. Almost no one is using SOAP as an RPC and a self-describing process. Almost everyone is using XML and XML wrapped in SOAP. [The focus on SOAP as an RPC and self-describing process is] the residue of an industry that is [used to] capturing developers mind share [and] setting the stage for the new round of products coming through and is not necessarily capable of helping IT do something today because they dont have their products ready.

This wonderful self-describing space is the nirvana that theyre getting ready for people to develop to. But what do you do between now and nirvana? ... In the first generation of Web services, all were doing is sharing files. Were literally putting XML into the messaging environment.

Im actually not sold on the next step yet ... because, as soon as I add an RPC, I have to know what Im calling, what Im connecting to and how it flows. Works great if Im connecting an end-to-end application. Doesnt work great if Im communicating with lots of apps and I dont know which are loosely coupled, which are any-to-any.

The vendors may not be solving the problem the CIO has. The standards may solve it very well, but the actual implementations that are being touted as the really cool ones may be more complex than necessary and may be more out into the future than the executive has time to work for. I believe the second step of Web services is three to four [years off].

eWeek: Is this confusing to IT?

Thomas: I think its extremely confusing. ... Vendors are saying IT has to learn [new] tools, do a new type of architecture, create services in front of their systems. This is a huge design change. ... So the designer of the system has to decide. Where is the dollar value of making this conversion? If the [current] system works, theres no dollar value.

eWeek: Unless you buy in to the idea that youre going to use Web services to turn some internal process into a revenue-generating process.

Thomas: That [idea] worked for the dot-com era. But when the dot-com crash happened and the recession hit, CIOs backed up and said, Where do I save money? ... Its in figuring out how to automate or eliminate costs internally. So what we see is XML being applied to the data integration problem.

eWeek: Will Web services bring a decentralized approach to computing?

Thomas: Inside the business ... having a central repository of data makes total sense. [But] how that data is distributed has to change from a tethered connection to that repository to a distributed connection. So I can get that data [in more ways] than just a terminal accessing it. So thats the change XML offers.

Now, everybodys trying to create these massive, centralized databases, and theyre trying to connect up their business systems so it all looks like one autonomous function. But theres a problem with that. [DaimlerChrysler AG was] talking about telematics. They said, A car [drives] up to the gas pump, they cant pay until [Wingcast LLC] and [OnStar Corp.] and everybody talk to each other so they can hand off the transaction. ... The car has a wireless connection already. Why cant it talk to the gas pump, and the gas pump can be the connection to whomever their point-of-sale connection is?

That change will happen. But ... databases are going to have to shrink to the function they provide, not the aggregation of all the data around them. Then youll have Web service engines in front of those databases, aggregating the data that makes sense.

eWeek: Youre skeptical of the ultimate state of Web services. But they make sense in that scenario, right?

Thomas: What Im skeptical about is teaching people how to design systems that are too complex for the reality of what they need today and even into the future. ... Our objective is that, as [developers] go to get trained, they learn the messaging as well as the transactional knowledge of how to do [Web services]. If they do that, the logical architectures will show up. If the transactional space—based on hype—wins, there will be really cool applications, but theyll be much less horizontal in nature and much less scalable. I fear that the excitement of how everything can work really cool together will produce a leapfrogging, and that doesnt help.