The copy command would probably NOT be the way to go.

With the copy command - you'll be connected to database1 and it will then open another connection to database2.

It will then SELECT from database1 and retrieve the data to the sqlplus client and then array INSERT that data from the sqlplus client to database2. Thus, the data tends to travel a lot further.

with a database link - you just pull the data from database1 into database2 - no pull to a client, push to a server.

The copy command has limited datatype support - it doesn't support much of any of the new types added since about 8.0.

I would not really recommend it as the first approach, the only advantage it might have would be that it can copy a LONG type if you know the maximum length of your longest long.

Others might say it also has an advantage in that it can commit ever N records. I firstly disagree that that would be an advantage (create table as select can skip undo and redo - there is no real need to commit every now and again). And if you really thought it was, it is nothing that dbms_parallel_execution in 11.2 could not give you as far as breaking an insert up and committing the bits just as easily (if not more so)