Alex: Your product has a number of software development kits [SDKs] supporting multiple languages. In your opinion, are there any inherent advantages of writing web services in one language versus another?
Craig: Our SDKs, which SOAP-enable Perl, Java, and COM applications, were implemented before there were good SOAP toolkits available. Today, we encourage developers to just use their favorite SOAP toolkit, which Grand Central will support out of the box.
Alex: A lot of companies have a lot of investments in legacy applications. How would they identify what processes they should convert to web services to share with outside partners? Should they try to convert some of these processes to a profit center? If so, what challenges would that pose?
Craig: Web services enable you to create a machine interface to an application such that you can broaden its ability to be consumed. Companies therefore need to find applications that would benefit from broadened consumption by other applicationsby other internal applications or by your partners.
This is very similar to what happened with web interfaces. Companies put web interfaces in front of applications so they could be more broadly consumed. Web services simply extend this value proposition to make APIs more broadly consumable by other applications.
In terms of making a profit center, I think that's a much higher bar that I wouldn't necessarily want to hold all web services efforts up to. I believe that most web services efforts will be about providing a new way to consume something. And perhaps if I'm already paying for a service, I may be willing to pay to more efficiently consume via a web services interface. To this end, we enable the metering data that's tracked in Grand Central to be fed to billing systems.
Alex: So they can integrate directly through something like a Metratech [an XML-based billing system] or something like that.
Craig: We haven't done that specifically yet, but it shouldn't be difficult. As I mentioned, all of the web services that we've seen deployed are to known partners and customers, and it's more about deepening the relationship than creating a new profit center.
Alex: Your technology uses a hub model to route messages. How would that scale as the number of client messages grow?
Craig: This is probably a more detailed conversation than we can effectively cover in this interview. We have deployed a two-tiered architecture in which there are a few data centers coordinating with a large number of lightweight POPs [points of presences]. We feel that this architecture will scale very effectively and while doing so allows us to maintain appropriate levels of security and reliability. By taking this approach, we can ensure the integrity and reliability of messages, but do so in a way that's extremely easy and flexible to deploy by our customers. All of our POPs, which manage the proxies that I was talking about earlier, are hosted in 24x7 network operations centers with fully redundant capabilities such that we can guarantee the security, reliability, and the integrity of the data. Because they're hosted, they're incredibly easy to deploy in front of any web service. We can swing them in front of everything from an IBM mainframe to an Excel spreadsheet, providing these services with enterprise-class security and reliability without having to install any software.
Alex: How does your firm work with some of the existing authentication initiatives such as Microsoft Passport and the Liberty Alliance spearheaded by Sun?
Craig: Well, those authentications are for user authentication. Mainly we deal with machine-to-machine communication. So our focus is more on working with folks like VeriSign. To the extent that Passport and the Liberty Alliance extend their models to include machine authentication, we would certainly embrace them as part of our framework, but it probably won't happen for quite some time.
Alex: Currently the UDDI directories are being controlled by the technology vendors, including IBM, Microsoft, and others. Do you see directories for vertical industries formed by the market leaders in those particular industriesfor example, a Baby Bell creating a UDDI directory solely for the telecommunications industry?
Craig: My stance is probably more radical than most on this issue. I think everyone's going to have a UDDI directory. You're going to use it to maintain your list of services that you regularly interact with. As such, I think UDDI will look increasingly more like LDAP than Yahoo!. I also believe that there will be public directories that will emphasize discovery and, as you've suggested, I think they will map to specific vertical industries.
Alex: One technology that has been around for a while but has gotten a lot more attention recently is enterprise portals. There are a lot of conversations around how that's going to be the first contact for most people with web services, even though the user is probably not even aware of it. Is that what you're seeing right now?
Craig: Web services are just architecture; as such, it can be applied to a variety of problems. This includes enterprise information portals [EIP] as well as application development and integration.
Within the context of integration, you're using web services to tie applications together into business processes. Within the context of application development, you're tying components together to create applications. Within the context of EIP, you're tying modules into portals. As to whether the EIP use case is going to happen first, I really can't say.
Alex: Earlier you touched on metering for billing purposes. Is it really feasible for a web services producer to charge for usage on a transaction basis? eBay is the only company that comes to mind as having been successful at this.
Craig: I guess the analogy I would use is a web site, where using the web was just a different way of producing your interface. So if you're Dun & Bradstreet and you make credit card authorizations into a web service, could you charge for it? Sure. It's just a different distribution channel or packaging. If you currently charge for tariff calculations or tax calculations, and you put a web service interface in front of it, could you charge for that? Sure. It's simply about allowing you to be more broadly consumed. So I think to the extent that people have existing service-level interfaces that they currently charge for, it's almost a no-brainer to create an opportunity for a billing interface.
Alex: The recent announcement about SAP adopting the web services initiative took a lot of people by surprise. Do you see other companies announcing? What's the broad implication of a company when it's that large, with an application that big, going to a web services model, as opposed to traditional ABAP [the traditional way of interfacing with SAP]?
Craig: I think this is a huge opportunity for all independent software vendors [ISVs]. A company like SAP or PeopleSoft may have viewed their application in the past in terms of their GUI interface [or] their human interface, but really what they've created is an application that runs a business process or multiple business processes. And just as important as creating a human interface to those business processes is creating a machine interface, especially since that enables it to interact with other business processes. By more deeply integrating within an enterprise it creates switching costs, which are very valuable to those vendors.
The big transition that I've seen in the last six months is that many ISVs assumed that web services were about delivering their software as a service. But people are starting to come around to the notion that it's not about thatalthough that's an interesting opportunity with services. The real opportunity lies in how I create a machine interface or a service interface to my application, and view my application more as the process itself, not just in terms of the human interface.
Alex: Where do you see this technology going in the next year? What major initiatives or assumptions do you make in your strategic plans, based around one year or five years?
Craig: By the end of 2002, I think you're going to see web services being deployed by early adopters to do both B2B integration and internal integration, as well as EIP-oriented deployments. Web services are an interesting model for internal application development and code reuse, but there are a lot of issues around configuration and change management that have a long way to go before this notion of dynamically assembling billions and billions of components in the network into applications. I previously worked at a company that marketed an application development platform for object-oriented components. I have a hard time imagining that having a standard protocol and data format solves a lot of the problems that you have with using external components, specifically issues around reliability and change management. If you used a component to build your application and a thousand other people also used it, how does that component ever get upgraded? What happens if there's an Internet problem? I've lived through that. I used to be a real optimist about it and now I'm a cynic, but I still believe that it will eventually happen. I just think there are a lot of issues that people have to think through. It's not just about the protocol.