Lists Home |
Date Index |
> [ Simon St.Laurent ]
> Does anyone know of a current and hopefully complete
> picture of the (SOAP) Web Services stack?
> [ SOAP, WSDL, UDDI, and then what? ]
I think that the real question about the stack is which layers people are actually building on.
SOAP and WSDL are definitely legitimate.
I would qualify the presence of UDDI -- "free love" on the business internet is a long way off. Security doesn't begin to cover trust, and that's far more important for business transactions.
> [ Choreography / Orchestration ]
To understand the various options, you have to understand which vendors are sponsoring the various alternatives. The most important thing to realize is that none of this is actually new ideas -- it was all covered thoroughly in the late 1980's and early 1990's by people thinking about how to weaken the various subproperties of ACID. ("I" is typically the sticky wicket.)
There are any number of implementations of the necessary underlying concepts (application-level faults, "sagas", open nested transactions, etc.), most slanted at some particular angle.
I'll just give you a vendor map for starters.
BPEL is proprietary (to IBM, BEA, and Microsoft), and when I've tweaked people at Microsoft about it, their response was more or less "we've already given plenty of stuff to the community -- now we want to get paid." (And who can blame them! I like getting paid, too!) <blatant.tweak>The question of whether any RAND licensing would be defensible boils down to whether an XML veneer on a well-established body of academic research is defensible. (Much of the academic research focused on and motivated by business processes, so I don't see the addition of XML as constituting a new application. I assume that the folks at the USPTO are clueless, however.)</blatant.tweak>
BTP is more open, at least ostensibly (See OASIS IPR policy.), is hosted on OASIS, and has folks like BEA, Oracle, and Sun on the TC.
WSCI has BEA, SAP, and Sun as the gorilla chorus.
And then there's ebXML BPSS, BPML, etc.
The more interesting part comes when you dig into the models, try to map them to common semantics, and see where people are being silly (like doing two-phase commit over async communications -- "Prepare, hold those resources, and I'll call you back.").
The answer to the question "Is there a royalty-free specification of web services choreography that legitimately combines and acknowledges the established body of knowledge on long-running transaction models, business process semantics, and cross-domain implementation realities?" is "No."
Just some thoughts.