On 03/27/2017 10:29 AM, Tom Zander via bitcoin-dev wrote:
> For some time now the relation between block size and propagation
> speed has been decoupled. Using xthin/compact blocks miners only
> send a tiny version of a block which then causes the receiving node
> to re-create it using the memory pool. Immediately getting double
> benefits by including pre-verified transactions from the memory
> pool you avoid the old problem of having to validate them again
> when a block was mined.
> As such there is no downside to a miner creating a bigger block, as
> long as all the transactions they include are actually in the
All transactions being publicly available is not something that can be
With no opportunity cost for a miner to generate withheld
transactions, a larger miner still maintains the economic advantage of
latency as a function of block size. Fast relay works to reduce
latency in relation to the opportunity cost created by the space
constraint. IOW, the more fees a miner must give up to mine withheld
transactions, the greater the economic disadvantage of doing so. So
there is a "downside" (i.e. centralization pressure) up to the point
where the advantage gained from withholding transactions turns negative.
The rational competing miner must presume that a block is valid upon
confirming the announcement's PoW. He then has the choice of mining on
top of the (partially-visible) block, or ignoring it until it can be
fully populated. The former *eliminates fee opportunity*, since the
next block must remain free of all public fee-generating transactions
until all of the preceding block's transactions are visible. The
latter increases orphaning probability, since it implies mining on the
weak chain, which *increases total reward loss*.
One can conclude that no matter how much space is created, it will
always be filled by a rational miner, as a competitive necessity,
given the centralizing effect of latency. Making blocks big enough to
include low cost transactions nullifies the benefits of fast relay
techniques based on your above assumption, since a rational miner will
simply substitute withheld transactions.