On the subject of 'cluster unique dentifiers':

Take this as I am thinking aloud.

First question: Is tehre a limit on the size of the identifier, if not:

Why don't you let the 'cluster unique dentifiers' be composed of 2
UUIDs (i.e. 128 + 128 bit number)?

Which means, the first UUID half represents the cluster/node/CPU
(whatever) ID, and the second UUID half is the record ID.

Naturally, this is assuming UUIDs can be made unique enough (which I
believe they can be).

Or is this such a wild/unworkable idea?

Cheers,
Adem