On Apr 27, 2007, at 10:49 AM, Mike Glendinning wrote:

> --- In [email protected], Stefan Tilkov
> <[EMAIL PROTECTED]> wrote:
> >
> > I was in a panel discussion at a conference this week, and was
> > surprised to notice there's still no consensus about whether or not
> a
> > process engine (or rather, support for automated BPM) is a "must"
> for
> > SOA.
>
> Stefan,
>
> when you mention BPM, Workflow and BPEL you are covering quite a wide
> range of concepts and technologies, perhaps too large for your
> question to make any real sense.
>
I know, but I choose not to care ;-)
> At one end of this range you have the definition and management of
> long-lived horizontal workflows, usually human-based and with all the
> attendant business issues of the process re-engineering fad, etc. In
> my view, this is entirely orthogonal to SOA and requires a different
> approach and different set of technologies.
>
I agree this is orthogonal, and this is what I was referring to in my  
view above.
> At the other end you have the implementation of automated, short-
> lived vertical workflows, otherwise known as composite services.
> These, I think are an essential part of SOA and its unlikely that any
> meaningful SOA will not find the need for these kinds of services at
> some point. As others have pointed out, composite services can be
> implemented in a variety of ways, for example directly in Java or
> using a BPEL tool (process engine). But this choice is exactly what
> it says - an implementation decision - and is therefore nothing to do
> with the SOA itself since one of the major goals of SOA's "loose
> coupling" is to separate service interface (or contract) from
> implementation: you should not know or care how a services is
> implemented.
>
> Therefore, I think that although a process engine "may" be the right
> way to implement composite services in a SOA, this is definitely not
> a "must have".
This is exactly my opinion (which is a nice coincidence for various  
reasons ;-))

Stefan
>
> Regards,
>
> -Mike Glendinning.
>
>
> 

Reply via email to