|
I was sent a presentation this morning outlining some of the various approaches proposed within the WS-I community for Service Oriented Architecture (SOA) service interaction. And a few of the graphics just jumped out at me:
I find it remarkably amusing that the notions of “registration”, “discovery” and “binding” are so challenging for the industry... we were doing this with CORBA 10+ years ago and a full 7 years ago with Jini. What am I missing, is it just the need by competitive marketing groups/pre-IPO's to re-name things just to get funding? I just feel like this is a NIH problem, or worse that some competitors wanted to introduce discontinuous technologies so that they could fill the void. Even Adam Bosworth , who ran in that crowd for a while, points out here: (thanks Simon)
'Vendors such as IBM and Microsoft, in proposing the standards, were big, institutionalized companies trying to protect themselves, [Adam] Bosworth (formerly of BEA and now a VP at Google) said. “They were deliberately, in my opinion, making something hard,” he said, citing specifications such as Web Services Reliable Messaging.' InfoWorld
With the simple use of proxies/stubs (which of course, can be dynamically generated) we eliminate the need for intermediate message formats, with leasing we pick up dynamic binding and as a plus, can get resource management, and with Jini Transactions and Spaces, I get the ACID properties... and put the control of recovery back where it should be, with the calling (initiating) service, and without a “required” coordinator I can drive to distributed/peer oriented scale. What the heck am I missing, except the fact that everyone thinks that Jini, by default = Java and RMI?
And now let me really get up on my stump... language neutrality vs. platform/protocol neutrality.... why the heck should we use a description language to write code... even business code. There are terrific code generators for Java, moving from UML! Do Microsoft and IBM think that we, as a developer community, are that nieve?
A colleague of mine, Murali pointed out there there is little new ground in the WS space, but it's still popular because, let's face it Java isn't interoperable. I suggest that there are people who didn't want to be wrong in their dismissal of Java and Jini who now must invent their own “versions” of this venerable technology... let's just take a look at a problem decomposition solved by WS and by Jini:
So what's the hitch... Murali suggests “The fundamental thing we need to do is to externalize Java objects into a compactly encoded binary XML, through serializers with FastInfosets and JAX-FAST implementations. We then need to implement SOAP messaging within JERI, to provide pluggable transport. Then it doesn't matter what platform the service-endpoint is or how services are invoked. You have the best of both worlds: JINI and WS.”
We can soon prove to the WS-I community, neigh, the developer community at large, that its a lot easier to discover and reuse interesting services within this platform as opposed to wasting time in politicized standardization efforts. Furthermore, recognizing that more homogeneous environments wouldn't have to take the lowest common denominator approach; would allow us, as a community, to create hugely performant and well behaved service oriented networks.
Oh well, Tim, count me into the Illuminati, the Loyal Opposition!
Permalink
Trackback:
http://blogs.sun.com/dhushon/entry/like_old_wine_in_a
|
