I already indicated bit-coin addresses and part of signed transactions being random, what I was talking about the address occurring multiple times in block chain.
This would require some sort of local index of addresses. This would be bloat. Further, how big would you make the index field size? 4 bytes? This was once considered enough for packet-switched networks.. and looked what happened there!
Satoshi intended for addresses to be only be used once. Future address generation could far exceed addresses used in other protocols (ie IPv6 = 16 bytes).
Mr Smith is correct. Great minds have already thought of such things for us already. In fact, somewhere along the way someone already figured a way to compress public keys to half their original size. No harm in us mere mortals trying to understand however.