Currently the most compact way (proof-size) to sacrifice Bitcoins that
does not involve making them unspendable is to create a anyone-can-spend
output as the last txout in the coinbase of a block:

scriptPubKey: <data> OP_TRUE

The proof is then the SHA256 midstate, the txout, and the merkle path to
the block header. However this mechanism needs miner support, and it is
not possible to pay for such a sacrifice securely, or create an
assurance contract to create one.

A anyone-can-spend in a regular txout is another option, but there is no
way to prevent a miner from including a transaction spending that txout
in the same block. Once that happens, there is no way to prove the miner
didn't create both, thus invalidating the sacrifice. The announce-commit
protocol solves that problem, but at the cost of a much larger proof,
especially if multiple parties want to get together to pay the cost of
the sacrifice. (the proof must include the entire tx used to make the
sacrifice)

However if we add a rule where txouts ending in OP_TRUE are unspendable
for 100 blocks, similar to coinbases, we fix these problems. The rule
can be done as a soft-fork with 95% support in the same way the
blockheight rule was implemented. Along with that change
anyone-can-spend outputs should be make IsStandard() so they will be
relayed.

The alternative is sacrifices to unspendable outputs, which is very
undesirable compared to sending the money to miners to further
strengthen the security of the network.

We should always make it easy for people to write code that does what is
best for Bitcoin.

-- 
'peter'[:-1]@petertodd.org
00000000000000ce3427502ee6a254fed27e1cd21a656a335cd2ada79b7b5293