Daily Archives: May 21, 2006


Edinburgh – the last stop of the trip. It’s raining again today.
I’m staying in a hotel a brief walk away from the conference hotel, the Radisson SAS on the Royal Mile. Luckily I have a jacket with me, but my head was getting wet and it was hard to see through my glasses after a while. So it was a very welcome surprise to discover that the meeting gift was an umbrella, instead of the usual T-shirt or bag.
Beating Retreat in the Rain at Edinburgh Castle, May 20
Since I am IONA’sW3C AC rep, I am here to attend the semi-annual Advisory Comittee meeting. We are advisors: the membership does not directly control what W3C does (although of course we have considerable influence).
I was invited to chair a discussion on the Web of Services, which was a follow up to the discussion I started (summary blog entry) at last December’s AC meeting in Montreal.
I was joined by folks from BT, Chevron, IBM, SAP, and Yahoo. I would have to say that after almost two hours’ discussion on the topic of what’s wrong, if anything, with the Web of Services (W3C has included this in their goal statement (i.e. building a “Web of documents, data, and services”)) the picture is not much clearer than before, although we did have a very good and lively discussion.
What I had said, essentially, in December is that the Web of documents is an unquestioned success but the Web of services is not. So what, if anything, should W3C do to get it to be? How do we get the Web of documents connected up to enterprise data sources? Web services would seem to be the answer but there’s a good contingent who basically say let’s build up from where we are with HTTP.
One of the key differences between the HTTP world and the WS-* world of course is that WS-* specs are designed to work over multiple communications protocols, not just HTTP. One reason for this, of course, is that HTTP doesn’t match the capabilities of the communications protocols currently in use within the enterprise, such as various message queueing and RPC-oriented technology (and nor should it since it was designed for a different purpose – another factor behind all this discussion).
A lot of the discussion focused on whether or not “the Web” and “the Enterprise” have sufficiently differing requirements as to justify different architectures. In that case the question would not be so much how to fix one or the other, but how to connect them up effectively.
This is of course another variation on the “REST vs. SOAP” or XML/HTTP vs. SOAP discussion. Several commented that both are valuable and important, and should and could be used in conjunction to solve a range of problems.
Another view is that what we’re seeing in the industry is more of an implementation issue than a specification issue – that is, how to ensure compataibility, consistency, and interoperability (including security for example) across multiple implementations?
Maybe you can’t legislate this, as someone else said. Or maybe it’s a foolish discussion since market forces are in control, not standards, as someone else suggested.
But at the end of the day, the cost of continuing in a nonstandardized enterprise software environment is significantly larger than a standardized environment would be, and the widespread and consistent adoption of enterprise software standards would pave the way for industrial and scientific methods to be applied to software development. This represents a significant potential change to the industry as a whole, but it seems inevitable. The big question is probably where, or how, or who will emerge as the catalyst. Will it be the W3C? Stay tuned is all I can say at this point…