And where's the original poster at?? We're about ready to close this thread out .

There's one thing you forgot. Don't trust the execution plan. Use the profiler and see the actual execution stats before making a decision. I've already see 50/50...

Hmm, I meant the single query with a covering index (where both columns are in ONE index).

Have you tried with a covering index (1 index with both columns) and checking the derived table version?

exec sp_rename 'tablename.colname', 'newcolname', 'COLUMN'

That's the simple way... the problem is that if you have a table with 100 text columns and a few million rows, it'll take hours/day to finish this. While...

AS I said... that kind of table should have the username as the pk.

How about the actual key of the table??

DDL = the create script (create table...)

Use EM to generate those, I don't mind about the extra columns.

I like to be accurate .

No it's faster (assuming it is) because both search condition can use a single index seek, instead of possibly a scan. So I'd definitly check out the union version.

It's also a matter of missed opportunities .

It's a good thing that the "o" and the "a" are far apart on the keyboard .

Sorry but we need more info to answer that. Can you post the tables definition (ddl + keys/relations) so that we can built it up?

Might I point out that you'll want to also copy the primary key to the second table, most likely making it the clustered index, also I'd add the relation with...