That violates relational systems and also not a good idea...How are you going to administrate 1,000's of customer tables?? Huh? Doing this will also cause normalization problems meaning that your users could have duplicate contact information across different "on-the-fly" customer tables that are created.

Again, scrap the randomid approach and use the above schema I provided and it will address your goals of this project.

The randomid approach you are attempting to execute will limit the power of customers to view their order history and also they will be assigned a new "randomid" each time they login, which could pose problems in the ordering process, meaning that they will have to input all their contact information again...which is cumbersome and not user-friendly.

========================================
Buh Bye!

Cheers,
Me