Interoperability with Sun/Oracle/Fujitsu WS-CAF (Composite application Framework)
Hi,
Can someone tell me if there are still two standards emerging, diverging or converging in the Web Services transaction space? A year or so ago, I was reading about Oracle's and Sun's Composite Application Framework as a rival to what Microsoft, Bea and IBM were doing. Is this still the case? Is there a winner/loser or hybrid emerging?
Is it a case of Java and .NET demarcation flowing through?
I am specifically interested in the WS-AT subset of transactions so feel free (in fact please do) confine your answer(s) to this realm.
Heuristic txn resolution is anathema to me (and by the looks msdtc) but Oracle/Sun (and curiously IBM) seem to live for it! They got "heuristic hazard" all over the place and are quite happy if a subordinate TM votes PREPAREd and then changes its mind and aborts the txn :-(
So I guess my question is "What's the state of play? Are there still two horses in the WS-AT standards race? And, if so, who's winning?"
Regards Richard Maher
I am a member of the Oasis WS-Tx TC, so do not expect an unbiased answer. :-)
My view is that the measure of success for these standards is boots on the ground. In other words, real products implementing a standard and providing real product-level interoperability, creating real benefits for the customers and the industry. That's the measure of a successful standard.
WS-AT is a great protocol because it has all the necessary features while remaining very simple. That means that anyone in the industry who owns a transaction manager can implement it easily, which makes it all the more likely that it will be adopted. And with Microsoft's WCF and IBM's WebSphere implementing WS-AT 1.0, and several more companies participating in the TC and signed up to implement WS-AT 1.1, I would say that there's some real product-level momentum gathering behind it.
(As an aside, you are correct to indicate that neither WS-AT nor OLE Transactions support heuristic outcomes.)
Hi Max,
Thanks for the reply.
So, would it be fair to say that a transactional web service in something like Oracle iAS cannot participate in the same atomic web service transaction as IBM's WebSphere? Would the same restrictions/conflicts also apply to WS-Coordination and those (lamentable IMHO) Business Activity transactions? (Does no one else spit their coffee out at this whole compensation-transaction idea?)
In a perverse sort of a way, I hope that WS-Coordination is also incompatible 'cos (again IMHO) WS-AT is already pushing it up hill on the Web with a whole bunch of people who don't believe that Legacy Atomic Transactions model true Business Activity at all well, and certainly not on the Web. (Eg: Travel agent ignores a failure to book a hotel but still proceeds with Rental Car Hotel and Credit Card)
What's your gut-feel for the appeal and take up of WS-AT as oppose toe WS-Coordination among the Web-Services devotees?
Regards Richard Maher
PS. You didn't mention BEA WebLogic. I thought BEA was also on your side. Are they on board for 1.1 or did you just omit them 'cos your list was obviously not meant to be all encompassing?
Richard:
I don't know enough about Oracle's infrastructure to really answer your question. Obviously, if Oracle were to implement a WS-AT RM or TM, they would likely be able to interoperate with WebSphere's WS-AT support, as well as ours. Similarly, I don't have a solid sense of BEA's plans in this space - you'll have to ask them about their intentions here.
I can tell you that neither Oracle nor BEA have brought an implementation to our WS-AT interop events.
WS-BA activities are a different animal than WS-AT transactions. Obviously the former are more applicable to 'cloud' scenarios, while the latter are more applicable to glass-house enterprise scenarios. In other words, WS-AT does not create new scenarios for atomic transactions - it provides an interop protocol for the same scenarios.
WS-Coordination, BTW, is just a base framework for other protocols. Both WS-AT and WS-BA build on top of that framework. So the question isn't really WS-AT vs. WS-C. It's which interop protocol will win in the AT space and which will win in the long-running activity space.
