> So, during DB running time, there would not be same transaction ids ?
> I thought, that to uniquely correctly identify current sql

dependency we

> have to use composite key : current_connection + current_transaction.
> Adomas

If transaction ids repeated themselves, then garbage collection would
be in a lot of trouble, as would the whole multi version record thing.
Every transaction gets the next ID.

But you cant really use it for anything with long term significance,
as a backup-restore cycle resets it back to 0. Still it is helpful
information from time to time.

Adam