Well, in iota you cannot say anything like "cumulative weight of X is enough for a tx to be considered confirmed", since the time enters the story. For the system to be secure on its own (i.e., without additional checkpoints or smth), the ability of the attacker to issue tx's must be much less than the "natural" flow of tx's in the system. So, what you have in iota is rather statements like "if a tx got cumulative weight X by time T, then the probability that this tx can be annihilated is at most p, provided that the potential attacker's hashing power is at most h and the natural tx's flow in the system is at least m", see examples in the bottom of page 14 in the paper. In that "6 blocks confirmation rule" of Bitcoin there are some implicit assumptions as well, by the way.
Please translate that general math to some concrete examples for us, so we can compare likely confirmation delays.
Okay so qualitatively we just need to wait for some chain of subsequent TXs to reference the DAG node for our TX, and as long as the attacker's capacity to generate TXs is sufficiently low (perhaps similar to Bitcoin at less than 25 - 33% for selfish mining), then the double-spend probability will also be practically very low for tangles consensus. Indeed Bitcoin has similar probabilistic risk of double-spend:
https://bitcoil.co.il/Doublespend.pdf
Three observations:
- Confirmation is not instant, as it can be in other designs such as Lightning Networks (and also my design). Yet it will also be much faster then Bitcoin's block period. You are probably estimating at a few seconds at most assuming attacker's power is sufficiently low?
- The attacker's power is given by his ability to incur TX fees relative to the total TX fees being paid for TXs. This is qualitatively different situation than PoW block chains, because in a tangle every payer has to incur the cost of security by paying higher TX fees (which I assume be burned so the coin is deflationary to 0 supply, unless you burn PoW hashes instead which is what I would advise). In Bitcoin all users pay the cost of mining security as well though either through debasement or TX fees, so it seems about the same. A potential advantage for a tangle is if you burn PoW hashes as I suggest, then every user is mining, unless they delegate this hash to a server. However appears neither tangle nor traditional PoW block chain can be immune to a 51% attack, whereas in my block chain scaling design I do claim this immunity (sorry to plug my work here but is necessary in order to point out relative strengths and weaknesses of all designs available to the crypto community).
- No way to prevent double-spends in partitions of the network (no fault tolerance to network partitioning). I claim to solve this in my design and I think perhaps Lightning Network is also.
So overall I think this DAG stuff is an improvement over traditional PoW block chains in general, not just for IoT. But I do think I may have a superior design, but I am still analyzing to see what attributes the DAG might have that are superior. The elimination of the blocks and the aliasing error of chain reorganizations perhaps, but seems there are analogous issues in the DAG.
Please do not take my post as desiring to rain on your thread. I won't belabor my points. I am just trying to see where for example we might even collaborate if at all. Looking with an open mind. Cheers.
P.S. to the mathematician, kudos on working out all that math. I haven't made a decision to wrap my head around it yet. So pressed for time.
