[Cryptography] What is total world transaction volume?

[James A. Donald]

On 2/8/2017 5:16 AM, Jameson Lopp wrote:
> From some quick googling it appears that there are ~100 Billion credit &
> debit card transactions per year. But the real question is how many
> /cash/ transactions occur? I suspect (globally) it is probably far
> higher than electronic transactions.
>
> Given that blockchain-based cryptocurrencies require every validator
> node to process the entirety of the network's transaction history, in
> order to scale them up to reach mainstream level volumes, more complex
> layers of transaction processing need to be built on top of the base
> layer in order to achieve sufficient scalability.

What I have been thinking on for many years is that you do not really 
need every full node to process the entirety of the network's 
transaction history, or indeed any nodes processing the entirety of the 
network's transaction history.  Rather, your node needs to be able to 
prove that every node connected to your node by a rather short chain of 
transactions is in agreement about all transactions directly or 
indirectly affecting you.

Node C tells Node P  "I will pay you one crypto coin"

Node P says "Do you in fact have one crypto coin?"

Node C says "All my pals agree I have one crypto coin", and registered 
this consensus with the global consensus some time ago, and the global 
consensus is that I and all my pals together have quite a lot of crypto 
coins, so if my pals are lying about me, they are the ones who will be 
out by that amount"

Node P.  "OK, gimme."

Node C. "OK, notifying all my pals that I am down one crypto coin, and 
you are up one crypto coin.  And in a short while, I and all my pals 
will notify the global consensus that I and all my pals together are 
down one, and you and all your pals together are up one.

Node P "waiting for notification from all my pals that the global 
consensus agrees we are up one."

Node P "OK, I and all my pals are up one."

Node P  "and my pals agree that I am up one, transaction finalized."

This scales up a lot further than bitcoin, at the cost of failure modes 
that may be difficult for end users to understand, but it is not 
entirely obvious that it will scale up to ten thousand transactions per 
second.

The problem is grouping nodes to get locality of transactions, that 
people you transact with are generally near you in the tree.

If there are major failures of locality, we still get substantial 
savings relative to bitcoin, but "substantial" may not be enough to make 
a difference.

The proposed protocol as described is subject to a Sybil attack, in that 
you may find yourself in a subtree dominated by nodes controlled by a 
scammer.  Dealing with such attacks in a fashion likely to be 
intelligible to end users is not obvious.