[bitcoin-dev] Guiding transaction fees towards a more censorship resistant outcome

[Ruben Somsen <rsomsen@gmail.com>]

When a user creates a transaction with a fee attached, they are
incentivizing miners to add this transaction to the blockchain. The
task is usually not very specific -- as long as it ends up in a valid
chain with the most Proof-of-Work, miners get paid. The payment is an
incentive for miners to act in the way that users desire.

To the user, there’s an individual benefit: their transaction gets
added. To the network, there’s a shared benefit: all fees add to the
security of other transactions in the chain. Miners can choose to
ignore the incentives, but they would be leaving money on the table
(and eventually get replaced by more competitive miners).

Transactions from Bitcoin Core are slightly more specific about what
they ask miners to do. Every transaction is only valid at a block
height that is one higher than the last block. This incentivizes
miners to build on top of the last block, instead of going back and
reorganizing the blockchain. This is especially important in a future
scenario where the fee reward is larger than the block reward.

BIP 115* by Luke-jr is even more specific. It enables users to create
transactions which are only valid if they are mined on top of a
specific block. While originally designed as a form of replay
protection, it actually serves as a deterrent for miners to reorganize
the blockchain. If they orphan a block, it will invalidate
transactions that demanded the inclusion of the orphaned block. This
increases the cost of intentionally reorganizing the blockchain.

Coinjoin**, the act of combining payments of multiple users into a
single transaction, can be seen as yet another method for users to be
more specific. The fate of their payments are now intertwined with
that of others. If miners wish to censor a single payment, they have
to reject the entire transaction, and the associated fee amount.
Techniques like mimblewimble simplify this process, by making coinjoin
non-interactive.

This brings us to a theoretical scenario where:

- every block contains only a single coinjoin transaction
- the validity of this transaction depends on the inclusion of a
specific previous block
- the block reward is negligible compared to transaction fees

In this scenario, if miners wish to undo a specific transaction that
happened two blocks ago, they would have to create three empty blocks
(receiving negligible compensation) in order to overtake the longest
chain. And even then, users can still refuse to let their new
transactions be mined on top of the empty blocks, disincentivizing
such behavior completely.

While not easy to achieve in practice (e.g. resolving natural forks
becomes more complicated), it demonstrates that users can become more
empowered than they are today, benefitting censorship resistance***.
It is this line of thinking that I wish to convey. Perhaps it may
inspire further ideas in this direction.

-- Ruben Somsen


* BIP 115: https://github.com/bitcoin/bips/blob/master/bip-0115.mediawiki

** Coinjoin: https://bitcointalk.org/index.php?topic=279249.0

*** Risk sharing principle:
https://github.com/libbitcoin/libbitcoin/wiki/Risk-Sharing-Principle