jl2012 (OP)
Legendary
 Offline
Activity: 1792
Merit: 1097
|
 |
May 08, 2015, 04:02:10 PM |
|
Let's have some thought experiments.
m = Max Block Size
--------------------------
The very minimal size of a bitcoin block would be around 250 bytes, which will allow nothing but only a standard coinbase transaction. No one will agree to have m=250B because Bitcoin like this could only be a mechanism to timestamp a hash.
How about m=500B? That will allow one real transaction in each block, in addition to the reward. May be it would be enough to allow transactions between 2 or 3 users. But if there were only 2 or 3 users, we won't even have a Byzantine General problem. Mining is just a waste of resources.
Similarly, no one will agree with m=1k, 2k, or 10k. Bitcoin in such form may support transaction within a very small group but has no practical utility as a global payment network
Conclusion 1: there must exist a size x, when m < x, Bitcoin system would have no practical value and people would use something else. Different people may have a different value of x in mind but there should be a reasonable range. For example, Satoshi thought that x will be =< 1MB for a certain period of time, so he put the cap as 1MB.
-------------------------
On the other hand, we may have a Bitcoin network to record all transactions of all creatures in this galaxy, with m = GoogolplexB. However, only the Inter-galactic Payment Network Federation may be able to run it, exploiting the Hawking radiation from the central black hole of Andromeda Galaxy.
By reducing m, thus limiting tps, running a full node will become more affordable.
Conclusion 2: there must exist a size y, when m > y, the level of centralization will make Bitcoin nothing better (or even worse) than a centralized payment system. Again, different people may have a different value of y in mind but there should be a reasonable range. For example, Satoshi thought that y was >= 1MB, so he put the cap as 1MB.
------------------------
Conclusion 3: x will increase as the user number increase, as the number of tx thus demand of block space increases.
------------------------
Conclusion 4: y will increase as time increase, with time as a proxy of technological advance. y might be 10kB in 1989 and 1MB in 2009
------------------------
Conclusion 5: If x > y, there will be no possible value for m. Bitcoin is nothing but an academic experiment. That's why Bitcoin was not possible in 1989 but created in 2009. Satoshi is a genius but timing is equally important.
------------------------
So the question is, what is the value of x and y in the coming few years?
I think it's easier to determine y. I think no one would think $1/month is not affordable. As Gavin# suggests that $5 could run an pruning^ full nodes with 20MB block, $1 should be enough for 4MB.* . And if we allow it to be $10/month, as I think many Bitcoin enthusiast are willing to pay, y is 40MB.
However, for the sake of consensus, let's say the very minimum value of y is 4MB at this moment. Anyone disagree?
-----------------------
Determining x is more difficult. It is a function of user number, while user number is a function of transaction cost, and transaction cost is a function of maximum block size. So it is sort of a loop.
x is also a function of the availability of off-chain system capacity. If we have a good off-chain system, x would be smaller.
But not being too aggressive, if we want to be 50% of Paypal which processes 11.6 million tx/day or 134tps, that will be roughly 70tps, which means 10MB##.
I'd say 50% of Paypal is really humble. Even if we have all alternative scalability solutions implemented, we still need to have enough room for on-chain tx.
So I would say x is 10MB, at least for a few years. Anyone disagree?
----------------------
WAIT! So x > y and Bitcoin is a joke? Well, we are still far away from 50% of Paypal. Actually, the 1MB cap still works well at this moment with no one complaining confirmation delay. This suggests the current x is smaller than 1MB. In this regard, Satoshi made a good choice.
---------------------
The final questions:
Should we schedule a block size increase in 2016? My answer is yes, because we can see x is increasing and will reach 1MB soon, and we are still well below y (4MB). As the hardfork process will take many months to complete, it will be too late if we initiate the raise after x is beyond 1MB.^^
What is the new size? I will support any proposal that's between 1 to 20MB but I don't think it should be smaller than 4MB, as the y estimated above. Actually there is no point to make it between 1-4MB because the "pain" of running a full node in this range would be similar.
--------------------
#http://gavinandresen.ninja/does-more-transactions-necessarily-mean-more-centralized
^ well, you may argue that a REAL full node should not prune but it still monitors the network for any malicious miner activity can could warn all SPV clients
* Of course you can't buy a VPS with $1 but as the machine is not fully utilized you may do other things
## Don't forget 10MB is actually not enough for 70tps as blocks are created randomly.
^^ We did not have a problem when it hit the 250kB soft-limit in 2013, simply because that was a soft-limit and miner could easily response to it.
|
