(oops sorry greg- replied to you by mistake)

That address he gives is 77 characters/bytes (same thing). What I'm asking is how can it be so small.

That's an alternative design for multisig addresses that would put a byte giving the type of transaction and the 20-byte hashes of each of the public keys involved. They would not have been redeemed using CHECKMULTISIG, but would use DUP HASH160 CHECKSIG and the arithmetic or logical opcodes to create the "m of n" condition.

Nobody really liked that solution because it means a new 'type' of bitcoin address every time we want a new transaction type and long addresses.

Its only advantage is it didn't use CHECKMULTISIG, so there were no problems with maximum-sigops-per-block.