Chris Jones <cjones@rightnowtech.com> writes:
> So, assuming these numbers are reasonable, and assuming that Moore's
> Law applies to PG usage, we could have trouble with 64-bit OIDs in
> maybe 40 years.
Interesting point. However, once we bite the bullet of not tying
OID == int32, it should be relatively painless to equate OID to any
compiler-supported integer type. So we should only need to fix
this problem once. If your projection is accurate, no doubt C2020
will consider "long long long int" to be a required type ;-)
regards, tom lane