[Reaction-tg] New Member Introduction...

[Asaf Adi]

Hello,

I would like to welcome David to the TG and to thank him for the comments.
We should definitely discuss them on the next call.

Regards,

Dr. Asaf Adi
IBM Research Lab in Haifa
Haifa University, Mount Carmel, Haifa 31905
e-mail: adi at il.ibm.com, Tel: +972 4 8296455, Fax: +972 4 8296116



                                                                                                                                         
                      David W Ash                                                                                                        
                      <dash at juno.com>            To:       reaction-tg at ruleml.org                                                        
                      Sent by:                   cc:                                                                                     
                      reaction-tg-bounces        Subject:  [Reaction-tg] New Member Introduction...                                      
                      @ruleml.org                                                                                                        
                                                                                                                                         
                                                                                                                                         
                      29/01/04 21:50                                                                                                     
                                                                                                                                         




Hello all,

I wanted to quickly introduce myself because I am a new member of the
Reaction Rules TG.  My name is David Ash--I've worked pretty extensively
on rule-based and artificial intelligence systems in the past.  Most
recently,
I've worked a lot on real time trading systems both in Chicago and New
York in options, fixed income, and other derivatives markets.  I believe
there
are very strong potential applications for RuleML, and in particular
reaction
rules, in trading and compliance systems.

I also spent nearly four years at Brightware/Mindbox, working on
ART*Enterprise consulting projects, and co-authored Planning for Real Time
Event Response Management in 2000.

I'm definitely looking forward to participating in this forum--I wasn't
able to
participate in the telecon this Tuesday because of a telecon snafu but hope
to do so in the future.

In the meantime, Asaf asked me to look over the white paper and give my
thoughts from the point of view of someone who hasn't been involved in the
process so far and thus might be able to give a fresh perspective.  A few
thoughts initially come to mind:

--The white paper does a good job of defining what a reaction rule is.  It
might, however, be good to define what type of rule is not a reaction rule.
It seems to me that, potentially, any rule whatsoever can be viewed as a
response to some kind of event, and is therefore a reaction rule.  Is a
reaction
rule, therefore, characterized more by how we express the rule--a rule
expressed in terms of events is a reaction rule, and a rule expressed in
terms
of preconditions is not a reaction rule?  Is it the contention of the group
that
pretty much any rule-based system can be expressed in terms of reaction
rules?  That would probably be my own belief--if it is the belief of the
group
and the white paper, it should probably be so stated in the white paper.

--It would be good--and I mention this primarily because it is my own
background--to include some kind of a model of time, probably in
conjunction with the event model.  Certain kinds of events require response
faster than other events, and this needs to be modeled.  Certain kinds of
events are disastrous if you don't respond quickly.  Other kinds of events
require a quick response if you are going to respond at all, but if you let
slip
by the results are not catastrophic.  Certain kinds of events allow more
time
but do require response eventually.  Certain kinds of events have a notion
of
diminishing returns where the faster you respond, the better.

--In conjunction with the event and time models, someone at some point is
probably going to need to look at designing large scale search
algorithms--probably
along the same lines as the major search engines use--to allow large
organizations
on an enterprise wide basis to respond in real time to events using
reaction
rules.  You have a very large, distributed knowledge base, events happening
all
the time, and varying needs for response.  I'm not sure if the classic Rete
algorithm scales up to this type of situation.  Looking at what the major
search
engines are doing, in terms of a distributed knowledge base, and then
combining
with classic Rete techniques, might be the way to go.  If everything
described
in the white paper is ultimately implemented, you have an enterprise wide
rules
engine which is a very different problem technically from what rules
engines
have traditionally been applied to.

These are my thoughts right now.  I'll let you know if I have anything to
add.
I'm open to starting work on some of my suggestions myself, and am open to
discussions either in email or in the telecon.  Looking forward to next
Tuesday.

David Ash, Ph.D.
Real Time Agents Inc.
446 N. Wells St. #214
Chicago, IL 60610
 _______________________________________________
Reaction-tg mailing list
Reaction-tg at ruleml.org
http://mail.ruleml.org/mailman/listinfo/reaction-tg