Milan Babuskov wrote:

>Todd Brasseur wrote:
>
>
>>>>That the replication process would get new primary keys and insert the
>>>>records using the new primary keys rather than the existing ones. This
>>>>would allow us to leave our application and database virtually the same.
>>>>
>>>>
>>Why would this approach "lose the ability to check for errors"?
>>
>>
>
>If you records have one value for PK in one database, and completely different
>PK value for same record in other database, how would you keep track?
>
>If somewhere the process breaks, how would you check which records are
>replicated successfully and which are not?
>
>
>

What I was planning was to add a temporary column to store the new PK on
the server. I look on client and find IMRID 200. When Replicating I
see then next available IMPRID on the server is 200000, so I write
200000 to the temporary column on the client, then I replicate to the
server replacing the 200 with 200000 before doing the inserts.

Todd


[Non-text portions of this message have been removed]