Subject: | |
From: | |
Reply To: | |
Date: | Tue, 23 Dec 1997 08:33:30 -0800 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Evan writes after Ewart:
>>> "Rudderow, Evan" <[log in to unmask]>
12/19/97 08:14am >>>
Ewart North <[log in to unmask]> asks,
<snip>
>My user is now saying "why can't we do this on a transaction by
>transaction basis?" I think what he means is to closely link the two
>systems so that the SOP (9k) checks codes which exist in Image on the
3k
>so that mis-postings can be eliminated earlier in the processing cycle.
>
>My questions-
>Is this scenario possible? If so what do I need on each machine to
>accomplish this?
>What are the consequences (if any) for response on either or both
>machines?
Definitely possible. The way we've done it is by using Berkeley sockets
to pass a message from the HP9K to a listener process on the HP3K; the
HP3K process executes some application code, performs all of the
TurboIMAGE database access and sends a message back to the HP9K
with the
completion status.
Response time is great -- especially because all of the database access
are local to the HP3000 (i.e.; they don't flow across the network).
-------------------------------------------------------------------------------------
What Evan writes probably works fine, however I suspect that a generic
solution that doesn't require user written listeners and senders is what is
wanted, probably similar to the way ODBC works. I don't know diddely
about it, but isn't the Remote Procedure Call part of DCE supposed to
address this issue. Perhaps those in the know could elaborate on this.
Mike Berkowitz
Guess? Inc.
|
|
|