Minor protocol change that could reduce CTLV delta from many hours to almost nothing

¹ CTLV delta is many hours regardless of payment size, is that correct? This is a severe attack vector for "reserve payment attack". It can be trivially shrunk to almost zero. To achieve this, the 2-phase commit needs to be improved and advanced to a 3-phase commit, and the penalty made gradual (X amount at timeout and then continuous timeouts rather than the full payment at timeout). To achieve this, a penalty must exist on every phase (which in 2-phase it does not, there is none on "prepare" step) as the cumulative timeout gets very long. To achieve this, the other 2-phase (finish-on-timeout) has to be added first and then the cancel-on-timeout (the one Lightning Network uses) afterwards, with an intermediary phase that switches the timeout behavior. I have implemented this in full. It is quite easy. I would be happy to discuss it openly, here for example. The hours long CTLV delta (regardless of payment size...) is the elephant in the room. It has been ignored as no better solution was known. There is now a better solution with the 3-phase commit. Edit: For complete implementation (trust-backed) see here (it demonstrates how to add 3-phase, and it is trivial to add 3-phase to Lightning Network), for a pretty good article see here and for a video presentation see here below.