I'm a big fan of testable definitions.  In fact, if it's not testable
I don't consider it a definition.  So let's have a look for testable
assertions in here...

On 1/22/07, Selwyn Akintola <[EMAIL PROTECTED]> wrote:
> "SOA is a business centric software design paradigm characterised by
> the utilisation of well defined standards and protocols

Ok, standardization is testable, but some more precision would be
useful because if messages contained one non-standard extension, would
it still be SOA?  *What* is standardized is important.

> to create
> services and compose applications from services.

That's not testable.

> SOA mandates that
> services are loosely coupled

Loosely coupled how?  Exactly what concerns need to be separated?

> and communicate through the exchange of
> messages

So if services communicate via a third party (e.g. blackboard or
pub/sub style) that's not SOA?  Indeed, hiding the identity of
services provides an important degree of loose coupling.

> thereby allowing resource sharing and reuse.

Not testable.

> Interoperability and platform independence allow the composition of
> applications from services created using heterogeneous resources and
> hosted on heterogeneous technology platforms.

That seems mostly testable, except for "interoperability" which seems
to be redundant due to the aforementioned requirement of
standardization.

> SOA is a long term
> organization wide cross functional collaborative activity whose ROI
> will be achieved by service reuse and efficiencies gained by better
> alignment IT with business."

This doesn't even appear an attempt to define anything, so you could
probably remove it.

Mark.
-- 
Mark Baker.  Ottawa, Ontario, CANADA.         http://www.markbaker.ca
Coactus; Web-inspired integration strategies  http://www.coactus.com

Reply via email to