How does blind merge mining (BMM) work?

The idea is that Bitcoin miners are blind to the sidechain/spacechain that is being merge-mined.

In normal merge-mining Bitcoin miners that choose to participate must run a node on the separate blockchain (Namecoin, Rootstock, for example) and actually build the block for these chains, then include it somewhere in their Bitcoin block. See https://bitcoin.stackexchange.com/a/35870/1026 for a more detailed explanation.

In blind merge-mining constructs, such as Spacechains, a new role is added between the parallel chain that is being merge-mined and Bitcoin: that of BMM miners.

The BMM miners collect fees and build the blocks on the parallel chain and submit their hash for inclusion on Bitcoin in what we can call a BMM transaction. They pay Bitcoin miners a fee in BTC. Bitcoin miners can just include these hashes and collect the fees without caring about their meaning, without having to run a node on this parallel chain.

It’s expected that competition between multiple BMM miners will drive up the BTC fee paid in the BMM transaction up to basically the same amount of fees they collect on the parallel chain.

The hardest part of implementing such a protocol is ensuring that Bitcoin miners don’t include more than one BMM hash for the same parallel chain in the same Bitcoin block. That is bad because:

1. It can be messy for the parallel chain — it would have to have tricky and confusing rules in order to determine what block is valid, for example.
2. It’s wasteful — it will use Bitcoin block space for no purpose.
3. It undermines the goal of turning fees collected in the parallel chain into BTC paid directly to Bitcoin miners — since there is no atomicity guarantees between paying the BMM transaction fee and actually having a valid block accepted in the parallel chain, there is no incentive for competing BMM miners to bid the BTC fee up to the amount received on the parallel chain.

There are multiple ways of achieving these atomicity guarantees, such as the Ruben Somsen covenants BMM/Spacechain proposal and the BIP-301/Drivechain proposal.