Actually, what I tried was a reactionary system. Any update made by the end
user is stored immediately. It is this store that uses a database trigger,
and this trigger fires a procedure that then sends the updated value to the
rest of the client. Thus, nothing gets to a client that has not been
stored.
NOTE that I say Stored. In ORACLE, it is possible to transact against an
instance of the database that is stored in memory, and when the oracle
processes actually write the information to disk is actually of little
concern (oracle will guarantee that it can recover anything placed in its
memory space almost 100% of the time). Thus there is not a need for auto
updates because the only possible time an update is needed is when a user
commits a change.
HOWEVER, it is entirely possible to implement something either using
Oracle's check-pointing process or a similar timing mechanism in another
product...
Just thoughts,
Andy
>I don't suppose you can get it to save location etc.
> just before you crash ; )
actually
automatic updates of the dbms *are* a possibility
any thoughts on periodicity??
jeffs
Jeff Sonstein http://ariadne.iz.net/
http://ariadne.iz.net/~jeffs/jeffs.asc
===========================================
There are no bugs
there are just undocumented features
