Pipes or Alerts? How to do a batch server for a client/server system.

we've got a bunch of forms applications (2.3, 3, and 4) that we are gradually migrating to Forms 4.0 (and up) on a client platform such as the Macintosh. One app is already on the Mac.

some of our apps use the HOST packaged procedure. generally, it's used for one of three things: 1) printing things on networked printers 2) initiating batch jobs 3) sending email (we don't use Oracle Mail). this is all being done on VMS. as we move toward the Mac, we'll need a different approach.

i'm thinking of doing a simple server that runs on the host looking for requests from the clients. using ORACLE 7, i see two obvious ways to approach this:

1. i could have the clients insert tasks requests into a table and use the DBMS_ALERT package to communicate to the server there is something to do.
2. i could use the DBMS_PIPE package to send the task requests.

the main difference between these is that 1 is commit based and 2 is not.

advantages of method 1 are that the client can proceed even if the server is down and a record of the task will be created. another (very slight) advantage is that i could make this work in ORACLE 6, by having the server poll the table. the disadvantage is that the client needs to commit. note that HOST does not require a commit. i can imagine that it might be a pain to require a commit in order to print something.

using method 2 i don't need to commit, but i think the client will hang
(or die) if the server is not up at the time of the request. also, if the
client wants to print something that isn't committed to a table, i'm not sure it's practical to send the entire thing that needs to be printed
(which might be big) through a pipe.

has anybody out there actually done this? any advice? are any of my assumptions wrong?

email replies and i'll post a summary. thanks. Received on Sun Aug 14 1994 - 05:36:58 CEST