[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [dist-obj] DOP & web-services & n-tier

Paul,

> Declarative workflow, then?
>
> I'm very interested, can you elaborate?  As someone who hasn't looked
> deeply into any of this, when I hear state machines / choreographies /
> etc. workflow naturally comes to mind.  Otoh, I mentioned this to one
> prominent web services proponent who dismissed it, saying "you can
> call anything workflow"... I didn't find that response too satisfying
> ;)
>

I don't know what you mean by declarative workflow, but here is a shot at
the way I have been parsing the hype. I don't think it is very insightful,
but maybe some body else can clarify further.

You are not the first to be confused. I run into this many times without a
satisfying answer. In the W3C web services workshop, so called experts were
fighting over these terms. I think that part of the problem is that anything
has been called a workflow and folks who are in the 'orchestration' camp try
to dumb down and upsell  'scripting of async. and distributed services' as
'business process automation' (which ofcourse makes customers think of
workflow for the same reason you thought of it).

I think many of the vendors when they talk about orchestration (or
choreography), what they are talking about is the ability to write a
'procedural/functional program' (perhaps declarative, perhaps not) that can
interact asynchronously or synchronously with applications as services. Adam
Bosworth shows this bias, MSFT's XLang shows this bias, etc. On the other
hand, 'rule interpreters that orchestrate' workflow have been as simple as
event-condition-action rules to as complex as production/prolog like rules.
I have not seen this second group of vendors participate that much in
definition of 'scripting' (orchestration if you will) of web services.
Maybe it is just a matter of time...

The problem with the rule based systems is that as they get more complex,
they become hard to debug and typically the 'searching' kind of rules that
one implemented in centralized knowledge based systems don't make much sense
in business process automation where folks try to remove search from the
process.

In this thread, Sundar brought up two interesting issues that seems to have
not gotten picked up: Granularity and async/sync interleaving. In addition
to this is the (availability or lack thereof of) declarative specification
of orchestration and what does that exactly mean in the distributed
application coordination sense.

Sankar





==========================================================================
To manage your subscription, mailto:dist-obj-help@distributedcoalition.org 
Archives, FAQ, etc.     http://www.distributedcoalition.org/mailing_lists/