On 04/21/2014 11:40 AM, Ashley Holman wrote:
On Mon, Apr 21, 2014 at 1:36 PM, Peter Todd <pete@petertodd.org> wrote:
That is mistaken: you can't mine on top of just a block header, leaving small miners disadvantaged as they are earning no profit while they wait for the information to validate the block and update their UTXO sets. This results in the same problem as before, as the large pools who mine most blocks can validate either instantly - the self-mine case - or more quickly than the smaller miners.

 
Under the headers first scenario, wouldn't the full block still reach everyone in the same time as it would under the current rules?  So the small miner loses nothing in terms of updating their UTXO set, but gains an early "heads up" alert that a new block is coming.  This allows them spend the propagation time working more productively on an empty block in the new chain rather than wasting time on an orphan.  It's true that it doesn't solve the problem of larger pools vs smaller pools, but if it doesn't make it any worse then headers-first propagation would be a net benefit to the network since it removes the incentive to make small blocks.


I think the most important part is that nodes can reliably decide on "first received", regardless of how they subsequently act on it.  I believe it would be fine for a node to receive a header and continue mining the old block, or a subsequently-verified competing block, until it has the necessary pieces to fully verify the first header received.  If that block data doesn't come, then it will be naturally ignored.  But if multiple blocks come at once, even if a competing block "verifies" first, the node would still switch to considering the first header received as the best block when it later receives proof it is valid (which may only be a couple seconds).

In other words, the node will always consider the header-received time as the primary ordering criteria, but will not mine on anything until it has full proof of validity, even if that is out of order.  This means that new blocks "effectively" propagate at the speed of 80 bytes, which limits certain kinds of block-injection/racing attacks.