If I'm reading the slides right, longer chains of transactions on a side channel do not increase risk, so it's innappropriate to imply that a more heavily used side channel introduces risks comparable to credit-card chargebacks.
Each party to the side channel has a multisig transaction signed by the other parties involved, so if one party cheats (or fails to comply for innocuous reasons), then any party can shift the entire chain of transactions to the blockchain without risk. So the only reason you don't want side channels to persist for too long is that each party has to keep around more and more information as the number of transactions over the channel increases; that's burdensome in its own right, but it also likely means that a side channel naturally decays over time, as the probability increases that one of the parties involved drops out or loses track of information. So you don't want to roll over side channels too many times, but it's not a matter of risk of trust.
This really is a very simple and elegant solution. Basically, it recognizes (1) that cryptographic verification of intermediate transactions only needs to be shared among the parties involved, rather than among the entire network, and (2) that any sequence of transactions among a sub-network of parties can be considered as a string of intermediate transactions. And (3) it requires no trust because the entire sequence of transactions can be moved onto the blockchain by any party at any time.
Sure. I simply meant that the assumption that an average person participates only in 2 chains per year and performs 3650 transactions in one chain is really far fetched.
Consider the amount of transactions you do. Basically every party becomes a single side channel which requires some activity in the actual blockchain. It would reduce the amount of transactions that are the style of "My local supermarket which I visit every other day" but not anything else.
This technique can be applied to created side channels among any connected sub-network of Bitcoin users, so if we're able to build good technology to figure out when to create a side channel (and with whom), it's likely that a large fraction of all transactions could be moved into side channels. It really does have very good potential for scaling the Bitcoin network by an order of magnitude, or two or three.
Combined with increases in the block-size limit, this technology and others are increasingly making it seem like scalability is unlikely ever to be a real problem for the Bitcoin network. (And file that under great problems to have.) Recently I've seen scalability cited as a demerit of the Bitcoin network more and more by people who don't keep up with the technology and the community developing it—and less and less by people who do. Good progress is being made.
But I agree with you that the rough assumptions the authors used to quantify this technique's potential to scale the network don't make much sense.
Each party to the side channel has a multisig transaction signed by the other parties involved, so if one party cheats (or fails to comply for innocuous reasons), then any party can shift the entire chain of transactions to the blockchain without risk. So the only reason you don't want side channels to persist for too long is that each party has to keep around more and more information as the number of transactions over the channel increases; that's burdensome in its own right, but it also likely means that a side channel naturally decays over time, as the probability increases that one of the parties involved drops out or loses track of information. So you don't want to roll over side channels too many times, but it's not a matter of risk of trust.
This really is a very simple and elegant solution. Basically, it recognizes (1) that cryptographic verification of intermediate transactions only needs to be shared among the parties involved, rather than among the entire network, and (2) that any sequence of transactions among a sub-network of parties can be considered as a string of intermediate transactions. And (3) it requires no trust because the entire sequence of transactions can be moved onto the blockchain by any party at any time.