Re: Oracle Streams and Uniqueness

>Well I don't immediately see why a statement at the source such as
>
>delete from t1 where c1 = '42';
>
>absolutely requires each row in t1 to be unique in order for the delete to
>be successfully replicated to a remote target (this is the nature of our
>problem transactions). I would have hoped that in a solution that trumpets
>itself with the description "sql apply" this would turn into
>
>delete from t1 where c1 = 42;
>
>at the remote side, rather than 300,000 executions of
>
>delete from t1 where some_unique_key = :1 .
>
>:(

Because redologs contain operations on rowpieces rather then actual sql... So your delete would result in delete operations on 300,000 rowpieces or more if there's chaining involved (and of course the operations on undo). How would you translate that back into sql ? 300,000 unique delete statements ;-)

K.

--
http://www.freelists.org/webpage/oracle-l