This does need to elucidated unequivocally. I don't think it is my role to impact Iota's launch. Come-from-Beyond demonstrates a very high S/N ratio, he has brought important ideas to the community which have even aided my design, and who knows I might even want to work with him. Let him get a return on his investment. I have said enough for the time being. I can say more at a future date.
I don't think a general discussion about why you see blocks as essential will be seen as an attack on Iota. I'd also like to know for my own learning process what ambiguities in particular you are referring to with double spends in a generic system with multiple chains of POW.
I want to write that white paper.
I want to write that white paper.
That's good, but it's a shame you can't talk about the ambiguity; I think it's possible you may have overlooked something in your analysis. There again, I may be plainly wrong, but I am happy to be proved wrong.
For us to be sure that all chains would never end up creating a double-spend (i.e. that they watch each other in real-time) would require signing all transactions
synchronously (one after the other, so the entire world has a simultaneity requirement to stand in a global queue to send a transaction), which is of course impractical.
So there will be windows of propagation wherein double-spends can be inserted in various places in the DAG that invalidate good transactions.
The ambiguity is which of the double-spends is the valid one. You proposed a rule that selects the double-spend with the greatest succeeding cumulative proof-of-work, but it is possible this is also ambiguous. Can you really not graph scenarios to confirm this to yourself? I suspect your myopia is because you didn't do a holistic analysis. Incorporate the first paragraph above and then suddenly it becomes clear that the longest chain rule is ambiguous within the propagation window of asynchronous activity. Now do you start to see why it is impossible to resolve ambiguity without a block chain.
Fuserleer pay attention.
In that case can't you drop both of the transactions ? If it's truly impossible to form consensus on which of 2 conflicting tx is supposed to be apended/included then drop them both.
But maybe I'm thinking a little to simple here and am missing the bigger picture.
Don't forget a double-spend might not be identified until well after many derivative transactions depend on that spend. So you also have to kill all those innocent transactions too, which is death to a coin. You will think we can just wait for some window of time to be sure but...
We've already identified that there can't be just one chain, because we need asynchronous transactions. But these tangles will be diverse (if they are generated decentralized) and we can't ever know at any point in time how many chains there are in the next moment in time. Thus we have no way to define the bounds of the window in which to declare that any transaction has been confirmed. Without blocks, we have no way to count time. The Iota paper goes into complex math to try to argue that we can calculate a probability. It seems to me this will only work well by centralizing the servers not just in the start but ongoing. The precise eludication of the math and rational is going to take more effort, but I think you can see conceptually the issue.
In every design I contemplated, it always came back to the fact that without a single longest chain, we can't get consensus without centralization. This is will be no different for Iota, because otherwise it could violate the CAP theorem. CfB has stated that he doesn't agree the CAP theorem can be applied, because he has some narrow definitions and conceptualization of the CAP theorem.
Specifically you can't have these plurality of Partitions for asynchrony and also keep both Consistency and Access. Thus we must lose Partition tolerance or Consistency.
It is going to require some more effort to elucidate this, yet I am confident the CAP theorem will not allow a DAG to work.
I expect Fuserleer has employed very complex designs to convince himself he has worked around the fundamentals of CAP, so that is why I need to see all the details of his design to drill down. It will require holistic analysis.
I hope this is instructive to monsterer as to why you can't analyze a design piecemeal as he was trying to do with Fuserleer. It is pointless.
Drilling down to how to demonstrate my expectation for Iota is going to require some intensive study of their math and deep thought. I had not prioritized that now, because I don't think that is urgent for me for my goals. I am quite confident a DAG can not work without centralization. I am confident it creates a Tragedy of the Commons around Consistency, which require servers to unify on a policy which mimicks a block chain.
My apologies to CfB for stating this without having an unequivocal proof. He should expect this, since he knows the math and system is complex. Peer review analysis will require time. Many of us have other priorities.