Not sure what tangent you've wandered off on now, but the crux of the matter was that one camp clearly pressurises on-chain tx, mostly to the exclusion of all else, and the other camp pressurises off-chain tx, mostly to the exclusion of all else. Neither camp wants to consider a healthy mix between the two. Both are willing to herd and funnel users into their desired and preempted growth ideal by either providing potentially too much or potentially too little space, respectively. This is the inherent problem with a static blocksize. In the act of choosing in advance a maximum amount of space to allow, you're also deciding in advance how the growth will occur, rather than allowing demand to speak for itself. SegWit 1MB vs SegWit 2MB are both stupid answers to the problem. Make it variable.
I think you're right, that if one had to re-design bitcoin into something with a 2-tier network, with an on chain settlement layer and an off-chain payment layer, that
the settlement layer should never have a hard limit, because it is an element of trust and security. In as much as "not getting your transaction through" is annoying with transactions as a scarce inelastic resource, it is not a security issue. If you can't pay, you can't lose your ownership either. If, however, *publishing potential tranactions* are the security element, they have to have a guarantee to be able to get published. Any form of scarcity of transactions becomes then a security issue. That doesn't mean that it has to be free, but
the LN needs a guarantee that any settlement transaction is going to be included within N blocks where N is significantly lower than the "time-out" of the security. If the LN has the ambition to become much bigger than the current bitcoin network, then the on chain capacity should *have the potential* to settle it entirely in N blocks, and even have a mechanism to guarantee that settlement.
Otherwise, you have the equivalent of in total only 5 appointed judges that have to deal with all potential contract litigations, and only consider litigations when the problem is less than a week old. If the number of contracts within these 5 judges' jurisdiction is one million, then as a contract subscriber, you are at the mercy of a judge having time next week to settle your problem if you're scammed. If enough people get scammed, most scams will get through, because the judges can only settle so many cases per week, and all the others will never be treated.
Add to that the funny part that these judges are also responsible for judging parking tickets, the day that everybody gets scammed, there are also 5 million parking tickets to be settled. (spam)
No, the LN without a mechanism of guarantee of settlement on chain within N blocks is outright dangerous.
That said, my question is: why on earth would one want to re-design bitcoin and not simply do this elsewhere on a different crypto currency, made for the purpose ? After all, bitcoin was a system designed to have a single layer of settlement. Why trying to modify the design of a car so that it can fly and why not design an air plane from scratch ? People wanting to continue to use a car can do so with the system they used to ; people wanting to go to the new air plane, can switch and use air planes, no ? I don't see why cars should be transformed in air planes.
