Re: Recognising one's own messages on an anonymous broadcast channel?

mikael.brannstrom@xxxxxxxxx writes:
Clients create client-unique numbers, say 64-bit. The client is given
some private key, say 64-bit. The result is a 128-bit number. 128 bits
are enough to guarantee that collisions will not happen (simple
permutation).

Yeah, something like that can probably work, but now you've got a
uniqueness management problem. There's a similar situation with
ethernet interfaces--every ethernet device is supposed to have a
unique 48-bit address, and the uniqueness is supposedly assured by
giving a unique prefix to every manufacturer and then having them
issue id's serially within their prefix. But the scheme fails,
manufacturers sometimes use the wrong prefix or use the same serial
number twice by accident (e.g. two different production facilities end
up using overlapping numbers). You end up with higher practical
assurance of uniqueness by assigning 128-bit numbers randomly than by
using a "guaranteed" unique numbering system.
.