This is quite an interesting point. In data engineering, compression is a key goal -- yet random identifiers are often used.
Part of the reason is, presumably, that over a distributed system it's hard to coordinate choosing a deterministic globally unique identifier without a performance cost.
Does any one have any thoughts on implementing a globally unique identifier in a compressible/deterministic way?
My initial guess is choose some low-collision probability random part, and then glue on timestamp?
Part of the reason is, presumably, that over a distributed system it's hard to coordinate choosing a deterministic globally unique identifier without a performance cost.
Does any one have any thoughts on implementing a globally unique identifier in a compressible/deterministic way?
My initial guess is choose some low-collision probability random part, and then glue on timestamp?