Andrew Dunstan wrote:
>
> Or we could divide the positive number space in two, by starting at 2^14
> (attnums are int2). Then a simple bitmask test would work to distinguish
> them.
Perhaps divide-by-four, then it would be possible to have calculated
columns (as mentioned recently on one of the lists). In particular, that
would let you have FK constraints with a constant as part of the key.
-- Richard Huxton Archonet Ltd