I'm having trouble following this.
"But one of the central pillars of smallblockism is that having larger blocks will result in increased orphan rates as the larger blocks take longer to propagate across the network. But Matts relay network, if applied to bitcoin as a whole rather than a select cadre of approved players, could deliver a performance increase that would obviate that argument."
Doesn't the entire block need to be sent somewhere (at least from a miner to one of the nodes in the relay network)?
A block is a (header + n(tx)) The vast majority of the transactions should already be in the nodes mempool, so why do they have to be sent a second time? Obviously due to the nature of the network, not all nodes will have the exact same set of tx'x, but any missing ones can be requested. Also, the relay keeps track of what tx's have been sent.
This is one of the long time conundrums of bitcoin - the redundant resending of transactions. You get a series of 10 byte ids instead of ~250b transactions.
Ok assuming the miner only sends a condensed version of the block with pointers to the relay network, the relay network still has to broadcast the full block then to other nodes, correct?
Correct - while the relay backbone remains a separate network, then yes. (but not over the relay network - they get propagated over the vanilla p2p as far as i know) But I can imagine a case where it could be extended to a wider network.Well then as much as I don't like to agree with the small blockers, their argument is correct that orphan rates will increase since the full block needs at some point to be broadcast. Although as I pointed out to Adam, would need to be much bigger (60mb) before this is a problem with current Internet speeds.
the full block is never sent, miners accumulate TX's as they come in, one of the functions of the relay network is to relay those TX fast, but when it's time to send out a new block every miner has all the TXs in the meme pool, so this condensed version of the block ( god forbid i call it compressed ) this is all anyone really needs.
As long as miners can keep up with the TX's as they come in and by doing so keep all their mem pools in sync ( it doesn't have to be perfectly sync...) a full block never needs to be broadcast.
This method will reduce orphan rates due to slow block propagation, miners ( the smart ones) currently use the relay network for exactly that purpose, they are able to get the new block 250X faster and this gives them an edge.
edit: wait ya you're right, because this isn't a standard way of sending out block the relay network learns of the new block by receiving the block in full. and then relays the condensed version of the block. but whatever this could idea can make smallblockist argument that having larger blocks will result in increased orphan, no longer true. which is a big step in the right direction IMO.