> or carefully structure your dblink joins so they can perform efficiently,
> possibly using temp tables as a sort of materialized view.
According to the documents unless you are writing procedural code with
cursors when you touch the dblink view it will pull the entire
table/recordset over.
>
> thats really all that the fancier database engines do behind the scenes...
> and even then, distributed joins can be painful.
I am not sure what they do but I have done this kind of thing in SQL
server without any problems and with almost no performance penalty if
the two databases were on the same instance.