I'm curious how this holds in a scenario where block propagation is constant because of inverted bloom filters?
If block solution propagation grows slower than log
Q, where
Q is the size of the block, then the fee market becomes
unhealthy and miners are incentivized to produced arbitrarily large blocks. I show in my paper that this is not physically possible due to the Shannon-Hartley Theorem.
This discussion relates to my main point of contention with Greg Maxwell. In email correspondence, he suggested that it would be possible to have infinite coding gain for block solutions announcement. (I.e., he suggested that block solutions can truly be communicated in an amount of time that does
not depend, even asymptotically, on the amount of transactions included in the block. He said this is possible if what went into a block was already agreed upon far ahead of time [so that zero information needed to be communicated when the block was solved]). I think the onus is on him to rigorously show:
(a) Under what assumptions/requirements such a communication scheme is physically possible.
(b) That such a configuration is not equivalent to a single entity
1 controlling >50% of the hash power.
(c) That the network moving into such a configuration is plausible.
Nevertheless, I've agreed with him to discuss this point of contention when I make the other corrections to my paper.
1For example, if--in order to achieve such a configuration with infinite coding gain--miners can no longer choose how to structure their blocks according to their own volition, then I would classify those miners as slaves rather than as peers, and the network as already centralized.
