<<long story on>>

In a past engagement around 7 to 8 years ago at the company I was working
for we implemented a client/server scheme that still is in production today.
We wanted to lookup records in a "real-time" basis but did not want to copy
the database from the remote system to our current HP3000.

Our "old" method was to gather records that were posted during the day, send
them to the "other" system for verification, returned back with their
success/failure, and then updated our system records.  This event was
scheduled twice a day and the users just had to wait for verification before
proceeding further.  The HP3000 system was a purchasing system and the
information that required to be checked was accounting information that
allowed the purchase or disallowed the purchase because of closed accounts.
The old accounting system was hosted on an IBM 3090 that was located on the
east coast.  The DBMS was not IMS but I believe IDMS (which looked a lot
like IMAGE in its structures).  The data passed back and forth was
accomplished using NRJE and IBM JCL.  The system was written in the 60s/70s
and was the main accounting system.

The goal was to provide a some what real time access to the accounting
information so that POs could be processed quicker.  A few months prior I
was involved in a project that utilized the Open Client & Open Server
software (callable C subroutines) available from SYBASE.

My first task was to make the HP3000 be a Open Server and could accept SQL
requests from any SYBASE Client on the network, process and retrieve the
data and pass it back to the client.  This was the time when 3-tier was the
hype.  The project proved that a client could access data from any system
(Unix, HP3000, IBM-IMS, DEC, etc.) and present it to the client without the
client knowing where the data came from or cared.  The open server sat on
top of ALLBASE TurboConnect (now IMAGE/SQL), passed the SQL to ALLBASE which
did the work, read the returned rows and returned them back to the
requesting client.  The process worked very well and quickly.  We were able
from start to finish put this environment in place and functional in 6
weeks.

For the purchasing system we went the opposite direction and made the HP3000
be a client and requested information from a server which happened to be the
old IDMS accounting system.  Using the Open Server software for the IBM
mainframes and the Open Client software for the HP3000 we were able to
create a client environment for our purchasing system as well as another
HP3000 application on another system.

When the process was all said and done the overall links involved included
the basic links of the HP3000 (located in one building), a SUN server
(located in another building a few miles away), connection to a CICS process
on the IBM (located in Florida), and finally a server program (written in
COBOL) on the IBM that accessed the old accounting data (also in Florida).
When the process was running we would receive a response of 1 to 2 seconds
between the start of the request and the receipt of the data.  The initial
logon time would be 7 to 10 seconds but once the connection was made the 1
to 2 seconds was constant from that point forward.  Not bad for the number
of systems and distances traveled for data retrievals.

The HP3000 application was written using Speedware using a process written
in C and implementing message files.  The use of the C routine and message
files allowed for multiple processes written in multiple languages to
utilize this process.  The C routine would execute a procedure stored on the
SUN server passing several fields and receiving other data fields from the
remote server process.

Using the Open Client and Open Server combinations our local HP3000 system
does not care where and what the data is stored on.  The system could be
moved to another platform (which is being done today) and our system would
not care.  We continue to send the requests to the SUN server process which
will be changed to point to the new system.

Sybase also had tools to access Oracle and Ingress databases which could
make the access seamless.  The use of these tools was a predecessor to ODBC
and I don't know what the overall status is at this time.  I also don't know
if Sybase still supports the Open Client and Open Server software for the
HP3000 systems.  I haven't been involved in this area for a while.  I also
don't know what other DBMS' they support or support ODBC.  I know there's
probably a cost for these but the functionality is there.

Just a little background/history info before ODBC became the "norm".

<<long story off>>

Tom Renz