I'm showing my age, but a number of years ago ELO released an album entitled "Discovery." One of the songs was entitled "Shine a Little Light," which is apropos for this month's editorial since our feature focus for this edition is Discovery.

I took part in a panel discussion on Web services during our Web Services Edge East show in New York City at the end of September. A number of very august industry representatives joined me, including James Gosling, Rick Ross, and David Litwack. The panel covered a number of topics that are on the minds of anyone considering Web services, and one of the topics that received a great deal of attention was Discovery.

Discovery is probably one of the most important differentiators for Web services as it positions itself as more than just the next incarnation of an RPC mechanism. It is the ability to locate a service without prior communication between service users that markedly advances the entire concept of assembly of systems from Web service components.

At least, it does in theory. The panel was skeptical on a number of issues. Our current idea of discovery mechanisms consists of UDDI and WSDL - two technologies that have a bit of overlap but more or less work together. UDDI is XML-based and as such is textually oriented, massively redundant in its meta-information, and extremely bandwidth intense. One of the issues we debated on the panel was the usefulness of a global UDDI. As we discussed the purposes and probable first adoption scenarios, it became clear that most of the industry expects Web services to be adopted first within the enterprise, rather than between enterprises. Not that we don't expect it to happen, but given the flurry of development that's gone on over the past few years, we saw industry trying to catch its collective breath by first consolidating their own internal systems.

So the usefulness, and even the implementability, of a global UDDI was brought into question. There was significant debate over the way UDDI would play. Most of us, myself included, thought that UDDI would first be deployed on an industry-by-industry basis, perhaps by third-party brokers in a fashion similar to the Net Market craze, or perhaps by key industry players in a pre-emptive strike. In doing so, the size of the overall database, and the network bandwidth required to support it, can be more thoroughly controlled.

Not everyone agreed though. Some thought that Web services should be like HTML, where anyone who wanted to expose something neat, something new, should be able to do so easily. Napster comes to mind as an example.

And speaking of Napster, the peer-to-peer model was discussed as well. James Gosling hit it on the head when he pointed out that peer-to-peer is just another, broader method of discovery. Once you identify - discover - the music you want, it's still a machine-to-machine connection. Which holds up well with my opinion that peer-to-peer is just another Web services option, not something completely distinct from Web services (although it can be).

We also discussed the versioning of Web services. The question of what to do when you've discovered a service and the owner changes it was on many people's minds. That actually led to a proposition about service-level agreements as an integral part of the UDDI entry, pointing out yet again how important discovery will be. For example, if the entry guarantees support for the service for one year, once you've found it via discovery you wouldn't need to look it up again, you could just use it. You would also be confident that the service interface would be stable for a period of a year, so that even if the owner of the service introduced an improved interface, you would have that period of time in which to investigate the improvement and adapt to it. In theory, a service is well defined and only the underlying implementation will change, but we all know that reality is never that simple.

Hopefully, this shed a little light on the topic of discovery. We have several articles this month intended to put it into the spotlight. Enjoy!

Author Bio:
Sean Rhody is the editor-in-chief of Web Services Journal. He is a respected industry expert and a consultant with a leading Internet service company. sean@sys-con.com