"rlee0001" <robeddielee@xxxxxxxxxxx> writes: > ... I know, for example, that by default PostgreSQL assigns every record a > small unique identifier called an OID. Well, actually, that hasn't been the default for some time, and even if you turn it on it's not guaranteed unique without additional steps, and seeing that it's only 32bits wide it's not usable as a unique ID for very large tables. "unique" and "small" are more or less contradictory in this context. > ... it seems like this is the sort of thing that even a fully SQL-compliant > DBMS could do internally to compensate for the performance issues with > using large natural keys in relationships. What performance issues are those, exactly? I have seen no data that proves that large key size is a bottleneck for reasonable schema designs in Postgres. Maybe that just means that we have more fundamental problems to fix :-( ... but there's no point in investing a lot of effort on an aspect that isn't the next bottleneck. regards, tom lane
