The blockdumps would look the same, but the hex for character data would of course be different. As to performance, there would of course be a significant performance overhead for a two-byte database character set, but I suspect that there would be no difference between one-byte character sets. I have no solid information on the matter. I am merely guessing based on my understanding of the way that the NLS functionality is layered into Oracle.

We are planning to develope an application which supports multiple languages. So I am planning to create a database with UTF-8 as database charset. But i am wondering how this is going to impact the data storage inside oracle.

In other words... if i dump the datablock of UTF-8 database and US7ASCII database both will be same or different. If same how the conversion takes place?

I am not able to find more intenrals references in either metalink or mailist archives. ANy ideas about the performcen implications of NLS vs Non_NLS databases.