Just read a compelling article from Grady Booch about the perils of believing the hyperbole surrounding Service Oriented Architecture. Personally I think SOA is a great idea but I’m struck by how similiar to the OMG’s CORBA it is and how, in effect, the real evolution is using a arbitrarily firewall traversing port and protocol for messaging. That semi-solved the messaging problem, delegated the security problem and the rest, as they say, is history… Thinking about Steven Vinoski’s comments about a service oriented approach I’m a little dissatisfied. The temptation to want blueprints and cookbooks for all problems is great but there’s a large inherent value in having these. Not to mention the peace of mind that comes with applying an industrially accepted architecture to a thorny issue like service meta-data. I’m too jaded to want to roll my own 🙂
To this end SOA standardisation is being seriously advanced by Steve himself and the other participants in the OpenSOA Collaboration which promises greater standardisation and hopefully industrial acceptance for SOA data layer and service composition technologies. Principally SDO and SCA. The value of service composition and the possibilities for workflow automation are compelling but what as Grady so eloquently puts it, what works so well in the powerpoint sandbox doesn’t yet translate to real world software systems.
Categories