Steve Jones wrote: > On 08/12/06, Stefan Tilkov <[EMAIL PROTECTED] > <mailto:stefan.tilkov%40innoq.com>> wrote: > > I have no idea what you might mean here. The WS-* universe has built > > its own protocol stack that can be used on top of some others > > (without really sucessfully hiding their details). REST is an > > abstract architectural style, the most popular implementation is the > > HTTP/URI protocol stack. > > There are others?
The issue for me is separation of invocation layer and transport. To provide compatibility with current applications, only HTTP works as the invocation layer/application protocol. I've not seen an alternate implementation either. This is the foundation of my arguments about transport vs transfer. If I use JERI endpoints to use HTTP transfer as a transport layer, then I've reused the existing infrastructure without changing the application. The JINI/RMI invocation layer is RESTful in that it cares nothing about types or number of parameters and return values at the implementation layer. When you use HTTP in a RESTful application, directly, you are directly bound to the semantics and implementation details of HTTP. Some type of virtualization layer is needed to keep that from happening. With RPC stacks such as JERI, there is a virtualization layer that abstracts the whole invocation and transport/transfer layer so that application portability is maintained even in light of service implementation differences. Several languages and platforms are equiped with such. Technology proponents only need to be prepared to take advantage of them. Gregg Wonderly
