[sdm7g@elvis.med.virginia.edu: comp.agents? (was: Dylan Agents)]

Jeff Pan (pan@eit.COM)
Mon, 29 Aug 1994 10:55:17 +0200


Welcome back. We have the opportunity to explore IA/agnet vs
CORBA under NIIIP (since one tool NIIIP will use is a CORBA-compatible
OO system). On the message you forward, I would have some reservation
on using Obliq: this product from Iona is one of the best (I talked to
Iona's president every year in ObjectWorld), but I am not sure what's
the commercial future for a platform-indepedent implementation of
CORBA (I saw companies with similar market position failed - Hyperdesk
for one). Oh, yes, one more problem for Iona is that it's based in
Ireland - they need a very good marketing rep at US to have a shot.
This is a pure business talk.

At this time, the primary CORBA tool is IBM's SOM/DSOM (and
the second candidate, DEC's COM, or Common Object Model, which has an
interesting feature of interoperating with Microsoft's OLE). Art
Goldschmidt is very interested in exploring the use of IA in
implementing some part of NIIIP infrastructure. I will pursue this
further with him (in fact, very soon - since he will be visiting me
tomorrow: he is here attending a tutorial workshop at Taligent).

One thing people appreciated IA for is that it is a "working
code" today (even if we can argue whether it is the best "product"
code at this time - I wish I can push for a better implementation
faster) - examples are the SmartProcurement demo for NIST; Sriram at
Lockeed, NIIIP, and Alan Weber on Sematech's distributed CIM
Application Architecure. I had an even stronger appreciation last week
when I was writing the Proposals and completing the demo. Seeing is
believing, and a working code means that people can start
computational experiments NOW (rather than "simluating" on "paper").
Having IA speaking HTTP is a master stroke! Further, I pointed out in
our proposals that some extensions of the current IA protocols will
make it even more powerful, including extending it to SHTTP (for
secured communication, and may be access control with digital
signature); MBONE for very efficient multi-party message distribution.
Safe-TCL is possibility if it really become commercially "popular" (I
would hate to spend time to include another "research prototype" - we
learned enough lessons :-).

Let's talk further when you can find time.