This is a direct request of the many IBM tech support and Toronto Lab
people who monitor this group... and of anyone else in the world.

I would like to be able to know if anyone in the world has ever made a
heterogeneous join work. Something like this....

Select a.col1, a.col2, b.col5, b.col7
From sqlservertable a, udbtable b
Where a.colx = b.coly

The issue is that such queries are supposed to work through OLEDB
drivers. But, they do not. For example, according to IBM and based on a
support research by Microsoft, the above query would confuse Microsofts
Query optimizer and not know how to pass the remote table references
constraint to the remote table. The effect is that the Query Plan in
SQL*Server will just bring back the entire remote table and do the join
locally to Sql*Server.

The above type query also causes SQL*Server to not release memory
resources. The original problem report was about that memory leak. The
resolution proposed by Microsoft to IBM was that it doesn't work, so use
OPEN QUERY instead. Case closed.

Classic case of agreeing to declare victory in the face of abject
failure.

The key is supposedly in "using the right OLEDB driver" yet none are
explicity mentioned. Thus my request. Has anyone in the world ever
gotten a heterogeneous join to work correctly, meaning with the remote
constraints passed along as they should be.

Thanks,

Kevin Knaus

DB2 & UDB email list listserv db2-l LazyDBA home page