Mark Baker wrote:
> On 4/5/07, Jef Newsom <[EMAIL PROTECTED]
> <mailto:jef%40improvingenterprises.com>> wrote:
> > I think that it is difficult to argue RMI v REST. Perhaps you should
> be arguing IIOP or JRMP v REST. At the wire level, JRMP has a uniform
> interface defined by three messages/operations/methods:
> >
> > Call
> >
> > Ping
> >
> > DgcAck
>
> "Uniform interface" refers to application layer semantics like
> "getStockQuote".
>
> As described at the URL you provided, "A Call encodes a method
> invocation" and therefore "Call" isn't an application layer semantic
> because "getStockQuote" is what you call.
Okay, this is really becomming frustrating. Everyone here is demanding that I
not talk about programming paradigms when talking about SOA. Yet, here, Mark
is
demanding that an SOA system be considered only a programming interface. The
RMI programming model is about allowing a software developer to say
Quote q = resource.getStockQuote(symbol);
in their software, instead of coding all kinds of transport and transfer
specific activities. What the 'resource' object is, how it interfaces with the
world, is not the important thing.
In order for resource, to be an arbitrary implementation, we have to consider
the operation that is supports to only be INVOKE. The notation for what to
invoke, is expressed in the programming language constructs.
That means that it is UNIFORM. It doesn't matter what the underlying
invocation
implementation is, nor what the underlying endpoint protocol is. INVOKE works
the same everywhere.
Mark, I guess you and Jan don't get my responses, because I'm not seeing
replies
to those, only replies to others comments on my replies. It would help me in
this debate if you could reply to the points that I am making so that I could
really understand where the issues lie.
We seem to go all over the map here...
Gregg Wonderly
