> We are (yet again) considering Midas - for a system that has to run
> immediately following major disaster (especially earthquake) where
> networks, telecoms and power are unlikely to be immediately available.
>
> The attraction of Midas this time is that independent laptops could be used
> for dataentry and then reconciled with main database when server and
> network restored later, than to the clientdataset structure.
>
> Question though. We would really like data to flow into major tables that
> make use of unique ids generated by triggers from Oracle sequences. How
> do you cope with this kind of data model in a Midas environment. Several
> laptops could be use - id collision looks likely. I can think of some
> complicated solutions but wondered if there were elegant ones since this
> must be a common scenario (use of server-supplied unique ids).
The Server is the master database.
Before a laptop can be used it must be initialised by the server. This
produces a unique Node ID representing that laptop. That laptop then initialises
it's own sequence at 1.
In the field it's own sequence is used to generate new records.
At merge time the NodeID and SequenceID are used to avoid collisions.
This is guarenteed unique and provides a tracking tool for Node lifetime
and usage. Ideally for performance you'd reconcile node/node-sequence
IDs to a new server ID at merge-time (a table used only for backtracking to
which laptop generated a record).
--
Aaron@home
---------------------------------------------------------------------------
New Zealand Delphi Users group - Delphi List - [EMAIL PROTECTED]
Website: http://www.delphi.org.nz
