Gregg,
 
> The Jini Configuration mechanism in concert with the JERI stack
> provide a great mechanism for deciding transport at deployment. 
>
I agree with this design approach (I am wondering, if my earlier post left
an an impression, that I am recommending transport binding at a stage
earlier than deployment?)
 
> You can just plug your appropropriate transport into the configuration
> and go.  This is one level of abstracting that many SOA architectures
> completely miss out on. 
>
Some SOA Architectures (and certainly Fiorano's) supports exactly this.
So, I must say that not all SOA architectures are "missing this".
 
> I think it's very short sighted to say that ESBs will be based solely
> on WS-* technologies.
>
Once again, I agree completely! (and I am again not sure how you got
an impression that my definition of ESB, restricted ESB to WS-*).
 
Just because a service INTERFACE can be represented by a WSDL,
it does not imply that the ESB should limit the invocation semantics
to those limited to WS-* or SOAP....
 
For e.g., even though Fiorano SOA Platform 2006, exposes each
configured "Service Component" as  WSDL, it still allows multiple
such Services to participate in a single ACID (distributed or simple)
transaction - without necessarily having to go over the WS-* semantics.
 
Thanks,
Amit Gupta
Fiorano Software Inc.

 
----- Original Message -----
Sent: Wednesday, January 11, 2006 11:16 PM
Subject: Re: [service-orientated-architecture] RE: Basics

Amit Gupta (FSI) wrote:
> Service implementation should not use "transport specific" API's, so that it remains
> completely agnostic to transport used for delivery/invocation of the service.

The Jini Configuration mechanism in concert with the JERI stack provide a great
mechanism for deciding transport at deployment.  You can just plug your
appropropriate transport into the configuration and go.  This is one level of
abstracting that many SOA architectures completely miss out on.  The fact that
SOAP ties the transport to the API and content is a big inhibitor to choosing
the appropriate wire representation and transport of your enterprise data.

> ESB vendors provide (or will provide) infrastructure for following:-
>
> a) Configuring Services (representing a configured service by a WSDL)
> b) Deploying Configured Services to a machine (network endpoint)
> c) Instantiating 1 or more such services over a set of machines
> d) Associating services with transport bindings (WS, JMS, In-memory invocation,
>     MQ series, MSMQ, Tibco, HTTP, FTP etc.)
> e) Creating a business process by connecting multiple such services (which are
>     bound to a transport) by linking output of a service to input of 1 or
>     more services.

I think it's very short sighted to say that ESBs will be based solely on WS-*
technologies.

I still continue to be amused by the fact that people think that WS-* has made
them, somehow language/platform independent.  Its just the transport
implementation that is open.  All of your software solutions and vendor tools
that wrote that software have positioned you to be completely dependent on the
technology, its risks and its costs.

Gregg Wonderly




SPONSORED LINKS
Service-oriented architecture Computer monitoring software Free computer monitoring software


YAHOO! GROUPS LINKS




Reply via email to