https://bitcointalksearch.org/topic/ann-bitmsg-a-proof-of-sacrifice-distributed-messaging-layer-over-bitcoin-308483
It supports multiple-recipient AES encryption (using RSA for keys), for example.
1. Create 16 addresses each one corresponds to a hex value a-f 0-9
2. Broadcast a transaction to each address that corresponds to the the message, including incremental value to keep order.
54657374206d657373616765 would send multi to
address5 value .0001
address4 value .0002
address6 value .0003
address5 value .0004
address7 value .0005
address3 value .0006
address7 value .0007
address4 value .0008
address2 value .0009
address0 value .0010
address6 value .0011
addressd value .0012
Wow, an entire output to designate half of a byte. Have you done the math for this? A bitcoin address is 20 bytes, which means you'd need 40 outputs alone just to transmit an address this way.
The approach is interesting, nonetheless. Why not just make 65536 addresses (for 0x0000-0xffff)? There's no reason why you couldn't have a database this large. It'd reduce the amount of outputs by a factor of four. Also, saving coins (from my solution) shouldn't be that much of a concern, since technically 0-satoshi outputs are valid outputs and hopefully will eventually be supported with prunable outputs.
You can. Dunno if Bitcoin-Qt will do it though.