Statement on commit registers

[Johan Nygren]

I now understand what the contention was with Jorge a year ago.

The original Ripple vision was a person-to-person system. By 2006 Ryan understood the timeout had to be "chunked". But this requires a penalty on each phase, which the simple "commit-on-hash-lock" 2-phase commit does not have (thus, by 2008, "staggered timeouts" was new idea instead, which then pivoted into commit registers by around 2010 after Bitcoin was launched?). Penalty on both phases can be "forced" into that 2-phase commit but does not fit naturally. The 3-phase commit that combines "cancel-on- hash-lock" 2-phase commit with commit-on- hash-lock is likely the ideal solution (article).

But, for Ripple+"collateral", Lightning Network or Raiden Network, commit registers seem ideal. They are perfectly atomic, whereas the person-to-person approach is not atomic in that way, it is based on incentives, a penalty (as I pointed out correctly then... as the Ripple Wiki also describes...).

I happened to step on a few toes of people here who are in this community more for Lightning Network, and I was told I "did not understand what atomic means" but it seems I was simply assuming Ripple as the lowest common denominator in the Ripple group, and not Ripple+"collateral".

I agree that commit registers are perfectly atomic. But, the person-to-person approach, chained timeouts, is not. It seems there was two different topics referred to rather than meeting half way. I was disrespectful to commit registers as being "the wrong way". It seems more accurate what happened in the history of Ripple was it started out as a person-to-person system, then a community formed that branched out also into Ripple+"collateral" systems. You in that community conceptually invented (or rediscovered) the building blocks for that, and then I show up and disrespect that "offshoot".

I acknowledge commit registers is great idea for Ripple + "collateral", but, it is not for person-to-person Ripple. My Resilience (path-based redistribution) only makes sense on trust-based system, and it does not make sense to subordinate such system under "commit registers", so I invented the 3-phase commit (HTLC switch at second phase from cancel to commit...). But I also like "blockchain" and there, I acknowledge commit registers seem the best, they are perfectly atomic which is only possible in such socially centralized approach. Jorge and others who focused on that are right in that (but, the decentralized approach, chained timeouts, is not atomic, which is why the 2006 idea of "chunked timeouts" is needed to manage the penalty risk they introduce, and turn it into the incentive to enforce the agreement...)

Hope this clears out that, and that any grudge from Jorge can possibly be resolved. It's a good thing person-to-person Ripple made progress from the back and forths (I found it thanks to that...), and it is a good thing if the Ripple+"collateral" has made progress too. Both important.

Peace, Johan