so if I understand correctly Peter's "landmark paper" rise to the top was.."short like leprechaun".
http://lists.linuxfoundation.org/pipermail/bitcoin-dev/2015-August/009916.htmla valiant effort I guess. getting some more peer-review probably would've been of better judgment?
ps. I'm not against you "crowd-sourcing" your grammar but I thought the whole exercise on reddit made your thread unreadable.
I don't see how proportion of hash rate has anything to do with orphan rate. Each broadcast block has the same chance of being orphaned as the next broadcast block, regardless of your proportion of hashrate, with network speed and connectivity being the only major variable. He needs to show some peterr-style analysis.
It is simple. You never orphan your own blocks. So if you have hypothetically a 99% hash rate you will have an orphan rate that is <1% regardless of propagation time. Someone else on the same network may have an orphan rate that is much greater than 1% given high propagation time.
The numbers work out differently with a more realistic hash rate share (say 15%) but the principle is the same. The higher your share the more of an advantage you get from the prevailing orphan rate being high.
Ok that makes sense
I think the "you" in the phrase "you never orphan your own blocks" needs to be thought about more carefully, but the point is valid that a block's propagation impedance [my (gamma x C)^-1 in the paper] is much less when the information is communicated across a miner's own hardware network compared to when the information is communicated to other miners. The paper spoke to this [albeit less than I should have in hindsight and I didn't explicitly talk about intra-miner communication] in the Conclusion and in End Note 13.
I was actually working on an "Appendix B" to formalize my definition for tau(Q) by considering these details more rigorously, but the math become too complex and so I felt that such an analysis deserved a follow-up paper instead.
In any case, my suspicion is:
(1) As long as
information regarding the transactions in a solved block needs to be communicated between miners (and even within a miner's own network), the Shannon-Hartley limit will apply, orphaning cost will be non-zero, and the fee market will remain healthy.
(2) We will be able to show that any attempts to create a "mining cartel" that prevents outsiders from accessing the same "fast relay networks" as the cartel members will fail, as individual members will improve their profitability by "cheating" by providing access to non-members. [A point Erdogan mentioned earlier in this thread.]
(3) The results of the paper will hold "as long as honest nodes collectively control more CPU power than any cooperating group of attacker nodes." [Satoshi White Paper]
Peter,
I read through about half of your paper and it outlines the issues very well. What it appears to me is how things will be moving forward there will be no block size limit and the network will be self-regulating as there is increased risk in creating larger blocks for more fees (i.e. pay off) in the form of that block being orphaned.
What I am still on the side lines about is whether there will be a successful attack on creating a very large block like say 1 TB just be happenstance.
Of course putting an upper limit on the block size could help I personally would like to see bitcoin stand on its own two feet with no limit restrictions on block sizes and let the free market of miners and attackers decide on how big the blocks should be going forward.
I'll have to read the rest of your paper but so far it is very interesting how you have outline the major issues and put them into visual representations for all to see and understand.
