I followed every one of these threads and it is really good stuff.
I've done some research to realize I need to do more.
I see that most of you are in the client/server world where you can ask
the client various questions about the query.
In my case, I am a standalone server embedded in a device that has
communication in one direction only. Application to device.
The device can only return status but the device can only return status
when the application asks for it.
The device must support nested transactions.
I gave the following example:
BEGIN parent;
insert into t values ('a');
BEGIN child;
insert into t values ('b');
insert into t values ('c');
insert into t values ('d');
ROLLBACK child;
insert into t values ('e');
COMMIT parent;
Someone asked why I just didn't do the following:
BEGIN parent;
insert into t values ('a');
insert into t values ('e');
COMMIT parent;
The reason for this is that the rollback was caused by a system error
and not a program decision.
For example:
My device may have enough disk space to store records 'b' and 'c' but
not 'd'. So this transaction must be rolled back.
Also, since I have enough space to store 'b' and 'c' I definitely have
enough space to store 'e' since 'b' and 'c' were rolled back.
This is a requirement for my device and may not be a real world
situation for any other system.
Ray
-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------
