[COAS-List] Re: MP&E

["Jon Farmer" <jfarmer@caredatasystems.com>]

-----Original Message-----
From: Jay K. Brinton <brinton@adelphia.net>
To: 'Jon Farmer' <jfarmer@caredatasystems.com>
Cc: 'Spangler, Rick' <richard.e.spangler@cpmx.saic.com>
Date: Thursday, July 09, 1998 8:07 PM
Subject: RE: MP&E


Hi John,

Manufacturing, Process, and Engineering (MP&E) is a cousin to Internal
Research and Development (IR&D -- pronounce it IRAD) in SAIC.

We should probably talk again by phone.  There are a few issues that come to
mind:

1)  I was hoping that the federation would be built into your Java PIDS MPI
wrapper, and that the search down (or up for that matter) could be
configurable to take place in an unattended manner.

I agree.

  The use of CORBA Naming is perfect -- I believe that is the recommendation
in the PIDS spec, no?

yes it is.

2)  If the client handled the federation, then where would correlation take
place?
It would take place on the server but be invoked (register ids) from the
client.

  I saw the client as the initiator of the lookup, and possibly also a
federation configuration client for the PIDS-based MPI, but not the
dispatcher of all PIDS invocations.

I have mixed reactions on this.  Perhaps we should have a server-side agent
that mediates the server-server coordination.  Which is in fact a PIDS
client but not an end-user interface.  I am curious how Chris feels on it.

3)  I guess doing it the way you describe has advantages in terms of
simplifying intellectual property issues -- all the work is done external to
the Care Data EMPI.  Are you proposing it this way for that reason and/or
Care Data cannot currently take on the financial/manpower burden of making
these enhancements to the EMPI?  (I can certainly understand if this is the
case.)

If we get our new hire then we can cover it easily and quickly.  If not then
it would wait until John Landers begins his scheduled SAIC-related PIDS
enhancements. this Fall.  Chris might be able to commnet on this as well.
We are now doing some locator experimentation for TeamSAIC that might affect
the design as well.

4)  Just for the sake of trying to gauge this activity as you described it,
what would you estimate (ballpark/WAG will do for now) would be the
composite hours involved using your new hire?

240 hours. (6 full-time weeks) for basic stable prootoytpe.  I suspect that
the ultimate user interface would then be attained by 240 more hours.  This
is knee-jerk.

We have scheduled the beginning of our PIDS Java client for last week of
this month.  It will do PIDS-based searches but will also do
Schema-dependent JDBC retrivals of visit history and insurances from an SQL
repository.  It will be productized, but we will then want to factor-out and
parameterize the schema dependiencies.  Maybe we should model insurances as
observations and call it COAS (half kidding).  If we transparentize the
distributed search by sticking it all into the IdentifyPerson operation on
the server (as a configurable behavior) then this client-side work of ours
will leverage the distributed search for free, but then the server-side
logic is subject to property rights.

How about creating the distributed-query-and-self-regsitering version as an
alternative implementation fo the same PIDS interfaces?  I would still be
server-side-only but we would need something like a royalty agreement for
one of us.


Look forward to discussing this further.

Regards,
Jay

-----Original Message-----
From: Jon Farmer [SMTP:jfarmer@caredatasystems.com]
Sent: Tuesday, July 07, 1998 1:11 PM
To: brintonj@saic.com
Cc: scott powers
Subject: MP&E

Is that what the project is called?

Jay,
I spoke to Scott about this project to determine both our interest and our
resources.

As for interest: for GCPR and commercial value, we already plan to implement
a facility to search upward domains in the correlating tree (up one level at
a time under user control, not transparently).  I suspect that we may find
as much interest from the marketplace in searching downward in the way you
described (search immediate descendents concurrently) and perhaps search
user-selected descendent domains.  Therefore the interest is there.

We plan to use the CORBA Naming service as the topology map.  I imagine that
we could produce the map in the user interface and then let the user ask for
upward and/or downward searches and then follw up by invoking "register" in
one or more ID Domains subject to privileges.  Think of it as a
topology-aware PIDS client.

As for resources:  Since we had not budgeted the work, we are low on
discretionery;  However, earlier in the day we spoke (Friday) we found a
candidate (for our fourth hire in the two months!).  This person has done
entry-level CORBA/Java programming - actually on a PIDS-related project.  If
we succeed in hiring this person, we would like to propose him on this
project but give him the assistance of myself for design and of John Landers
(our lead CORBA/Java programmer) for programming assistance.  We offer you a
composite rate of $65/hr.  This means we would bill you based on the
programmer's hours, but he would be guided by me and assisted by John
without additional charge.

As for intellectual property and marketing rights, I dont know where to
start.  Perhaps we can make a royalty agreement for the client, with the
stipulation that the product remains as an Applet that relies on PIDS-only.
Otherwise, the moment we couple it with anything else the benefits for one
or the other of us could be compromised.

I look forward to your response.  I will let you know when we confirm the
new hire.

Jon

<< File: ATT00000.html >>