On Sat, May 17, 2008 at 11:08 AM, Michael Poulin <[EMAIL PROTECTED]> wrote: > Setting WOA or Web Services as a "substyle of SOA" has two troubling aspects: > 1) SOA is still taken as a technology (otherwise Nick Gall mixes apples and > oranges), we it is not
I think SOA is an architecture (or more precisely, an architectural style). Hence the "A" in SOA. It is an architecture for IT-enabled business processes; hence it is an architecture for a technology/process hybrid. > 2) communication models - Web Services/WS*-, REST, MOM/Event-Driven - cannot > be architectural principles, its are just communication models; depending on > the task one of them might be more preferable than others. REST is NOT a communication model. It too is an architectural style at almost the same level of abstraction as SOA. It is not at precisely the same level of abstraction because it imposes additional architectural constraints above those imposed by SOA; hence it is somewhat more concrete than SOA. That is WOA's strength. By adding a few general constraints to SOA, it gives sufficient guidance to actually fulfill SOA's goals: sharing and agility. SOA is underspecified, as the discussion in this list demonstrates day in and day out. Calling REST or WOA a communication model is like calling the Web a communication system. WOA is the architecture of the Web -- both the browser to server Web and the system to system Web. Period. HTTP *might* be roughly characterized (although not all that accurately) as a communication spec (like SOAP is a communication spec). But REST/WOA is NOT HTTP. At best, HTTP is a partial implementation of REST architectural constraints. Roy Fielding discusses at length how HTTP falls short of full REST.
