RE: Executing OWL-S/Profile - a comparison of existing mechanisms

Hi, 

So finally you mentioned Jess, which I supposed as better than Jena regarding
inference. That makes OWL-S fantastic rather than solely simulating, and better
than BPEL in executing *rules*.

Cheers

Jun
 
-----Original Message-----
From: public-sws-ig-request@w3.org [mailto:public-sws-ig-request@w3.org] On
Behalf Of Bijan Parsia
Sent: Thursday, September 23, 2004 10:34 PM
To: public-sws-ig@w3.org
Subject: Re: Executing OWL-S/Profile - a comparison of existing mechanisms



On Sep 23, 2004, at 1:26 AM, Pranam Kolari wrote:

> Bijan,
>
> I had related questions about existing execution support for OWL-S, 
> since you mentioned two of them here. Also I am relatively new to the 
> OWL-S domain.
>
> 1. Does there exist any comparison of the below mentioned execution 
> environments. Each one of them will probably fit well in
> their own domain of interest.

There is no comparison that I know of. In terms of implementation, the 
VM is a collection of Jess rules encoding the "execution" semantics 
that CMU has produced (see their Journal of Web Semantics paper).

> 2. Do these environments support "Compensation Handling " ( from the 
> DB area ) or "Scope", "Exceptions" ( from the programming language 
> area). As per my understanding these are very important for a process 
> model to handle varying execution conditions. I am also not sure how 
> the OWL-S language (process) allows specifying such entities.

It does not at all. We started work on exceptions and had to lay it 
aside to deal with core functionality.

> All of the above is ofcourse taken from the BPEL domain and its 
> capabilities.  Note that my point here is not that a BPEL execution 
> platform is better.

WRT exceptions, it surely is :)

In general, let me add, the OWL-S process model is not primarily 
focused on being a programming language, but to support planning. There 
have been other tensions upon it over the years, of course, but if it 
is to be interestingly distinct, I think, it needs to be considered for 
it's off line simulation support more than it's online execution 
support.

> 3. Does there exist an analysis of process model capabilities against 
> - say the popular workflow patterns ( 
> http://tmitwww.tm.tue.nl/research/patterns/patterns.htm )

Not that I know of. There is work by, e.g., Rick Hull comparing OWL-S 
with his automata approach and by Michael Gruninger about axiomatizing 
OWL-S in NIST PSL.

Cheers,
Bijan Parsia

Received on Friday, 24 September 2004 00:21:25 UTC