|
Let me say that I find a DAG amazing. I also find the math in Iota's paper is creative and very well done. They did everything technically to a high class level (I am not commenting on the IPO!). It is hard to not respect what they have done technically and the programming is I assume up to CfB's high standards (but I haven't looked at the code).
The problem for me is I need to be sure the security is provable.
I am thinking here that something like the Principle of Least Power applies. It says that if your programming language is too general, then you can't extract semantics from it easily. I want to improve the definition of that principle, to be more precise. Too much generality means unbounded complexity.
A DAG is more powerful in the sense that it doesn't hard-code the payment and acceptance model and allows unlimited partitions. But the cost of this generality is that the semantics of the Consistency can't be unequivocally provably extracted.
I applaud the research, and even I can stand aside for the proponents to get their return-on-investment. It furthers ideas about crypto and it even influenced my design. But personally I would not have expended my effort on something that I have shown (for those who agree with my upthread arguments) chooses the tradeoff of less Consistency within the CAP theorem and PLP. I respect the work, but I see an egregious flaw which makes it unsuitable for my goals.
We all know that CAP is a tradeoff. Iota/DAG makes a tradeoff off unequivocally provable Consistency of the security (which is important with money), but they may be able to make the argument this depends on what the community of users does in terms of the selection algorithm widely employed. They may be able to make a convincing argument that this is the same as the community voting with its PoW shares to route away from any malfeasance in my design. I will of course be able to make some counter arguments (and I expect my arguments to be more mathematically convincing because unbounded is infinitely worse than bounded).
EDIT: for n00bs, the issues discussed upthread about a DAG are fundamental. Iota will not be able to change the fundamental trade offs mentioned above no matter what they do. So pleeeassseee don't make the excuse that "they can improve it later". We get what we get when it comes to fundamentals.
Edit#2: Iota is shipping now or soon. Often availability is more important than the most idealistic vaporware.
Edit#3: Or the simpler argument Iota has been making is that users don't care. But will the Consistency issue bite users eventually? I heard something about Iota using some centralized policy in the launch phase so I guess users will be protected from unknowns in Consistency during this phase?
|
|
|
|
Even if a tree was a relevant discussion in a decentralized block chain thread (which I am fairly certain it is not relevant), this is not an adequate excuse for introducing incongruent data structures/models when discussing a DAG, because it is just a way you attempt to save face for loading up the thread with irrelevant noise.
It's no attempt to save face, I have no problem admitting when I'm wrong. I made the mistake of assuming the way Iota and DAGs in general worked was the same as for trees, when it is not. Thanks. I hope you understand now that it is impossible to have security across multiple partitions without the analogous probabilistic model of a DAG. This is why I can assure that without a single LCR, any design will suffer the same fate. The CAP theorem is holding just as I expect it to for a DAG, because the Consistency is now unprovable in the general case. A DAG tolerates Partitions and maintains Access, but sacrifices the unequivocal proof of Consistency. It can be proven it is Consistent (within certain computational scenarios) with a particular probabilistic model, but it can't be proven that model is enforceable. Remember I was early on in this discussion on Iota emphasizing the distinction between the word "MUST" and "incentive". I edited the following: First you need to grasp that unlike in Satoshi's design where we have one single longest chain rule as acceptance truth, in a DAG there are many competing chains vying to be truth and no globally enforced protocol rule to select between them. The DAG's truth is the payer's model of how payers select which transactions to reference in their transactions and a second acceptance model of how we calculate the probability that the first model will select a conflicting double-spend. Thus the security is never based on any particular deterministic chain in the graph, but rather on the models that the payers and payees are using to model the DAG.
|
|
|
|
Because the "selection algorithm" (c.f. my prior posts on my meaning of that term) attempts to make this too costly for the attacker. Now you start to understand why it is a probabilistic security and not a deterministic one. Are you starting to understand now why your upthread posts were so irrelevant?
I admit that my examples and ordering diagrams are suitable only for a tree and not a DAG. A node in DAG can have multiple parents, so there is no one consistent ordering, however, a tree is a different matter, and since we are not solely discussing Iota here, they are still pertinent. Apologies I saw this just now after composing my prior posts. Even if a tree was a relevant discussion in a decentralized block chain thread (which I am fairly certain it is not relevant), this is not an adequate excuse for introducing incongruent data structures/models when discussing a DAG, because it is just a way you attempt to save face for loading up the thread with irrelevant noise. I urged you upthread to go take some time to think first. My PM box is available to you to ask me questions. This thread is supposed to be something that readers could digest, not some 1000 page crap of noise that no one can possibly digest. |
|
|
|
No, all heads that don't lead to conflicting ledger will merge. In case if it's not obvious, the number of heads is equal to the number of transactions. In ABCDEF sequence someone can still ignore "F" and pick "E" as a head. From the global point of view all heads will exist, "survives" doesn't make sense just like "orphaning".
Understood. So, what stops people from inserting transactions into past history by referencing really old transactions in order to change historical ordering? Damn it monsterer. If you are too lazy to understand and you continue to ignore what I am trying to teach you, then you are really pissing me off. You are rapidly losing the respect I had for you. You can only see things one way and you can't change your perspective to the probabilistic model. This is a sign of a lower IQ. You already know that there is no way to insert a transaction into an existing DAG chain. And yes a given selection rule can choose to reference an old transaction, but that creates a new tip (a new chain). But asking it the way you did shows that you are still thinking about irrelevant structure and not thinking about the model of the security as it is HOLISTICALLY probabilistic not locally deterministic to any given chain structure. The only deterministic rule is that transactions may not reference tips that would put a double-spend into the merged chains. You need to invoke math here, not just analyzing a chain structure. I believe this is beyond your comprehension. |
|
|
|
There are probably several different ways to try to explain how the security of a DAG works. For example, Come-from-Beyond characterized it as "two DAGs", one for payers and another for acceptance. My understanding is it is not that there are actually two data structures, but rather that the way the DAG is interpreted mathematically is analogous to two models (which actually depend on each other in some sense so perhaps let's say a unified model overall). First you need to grasp that unlike in Satoshi's design where we have one single longest chain rule as acceptance truth, in a DAG there are many competing chains vying to be truth and no globally enforced protocol rule to select between them. The DAG's truth is the payer's model of how payers select which transactions to reference in their transactions and a second acceptance model of how we calculate the probability that the first model will select a conflicting double-spend. Thus the security is never based on any particular deterministic chain in the graph, but rather on the models that the payers and payees are using to model the DAG. Until you understand that, you understand nothing about a DAG. Whether that is more powerful or less secure than Satoshi's design is what CfB and I are debating and discussing. My point has been that each payer and payee can choose a different model so the game theory of the security is unfathomable. The white paper shows that under a given model, certain attacks are too costly. But the white paper can't show what happens under all game theories and thus can't present any holistic model of attack cost. So, I guess in the best case, only one chain head survives this attack?
No, all heads that don't lead to conflicting ledger will merge. In case if it's not obvious, the number of heads is equal to the number of transactions. In ABCDEF sequence someone can still ignore "F" and pick "E" as a head. From the global point of view all heads will exist, "survives" doesn't make sense just like "orphaning". I find it very hard to explain the idea of Tangle, but once you get this idea you will see that if you send a double-spending right after the legit transaction you increase security of the tangle, not attack it. This is because you fight not only against the system but also against other attackers. "Enemy of my enemy is my friend" perfectly works here. Piecemeal details like this will just cause the thread to become noisy because they won't help the reader gain understanding of the real way a DAG security works. Discussing the deterministic structure of the DAG is misleading discussion. We need to first explain the above as I did. |
|
|
|
If this is the case, why can't an attacker simply place a double spend into every single chain head, and force all subsequent transactions to be invalided?
Wouldn't that bring the entire tangle to permanent halt since there would be no valid heads left to extend?
If one of the transactions is a double-spend then another is legit, right? So, I guess in the best case, only one chain head survives this attack? No necessarily! You still don't understand! You are thinking about the model entirely wrong. Wait let me compose a post for you. Please wait. Stop posting (stop spamming the thread). And wait. |
|
|
|
Sticking to classical definition - yes, a single double-spend invalidates all transactions referencing it. Once this happens the players will clone the invalidated transactions, for a particular transaction it will be the sender if he hasn't received the purchased service yet, or the recipient (because he doesn't want to lose money). But again, all this happens within the adaptation period.
If this is the case, why can't an attacker simply place a double spend into every single chain head, and force all subsequent transactions to be invalided? Wouldn't that bring the entire tangle to permanent halt since there would be no valid heads left to extend? Because the "selection algorithm" (c.f. my prior posts on my meaning of that term) attempts to make this too costly for the attacker. Now you start to understand why it is a probabilistic security and not a deterministic one. Are you starting to understand now why your upthread posts were so irrelevant? P.S. I am trying to teach (and hand slap) you that you can't go analyzing designs piecemeal based on analogies to other designs such as what you did with Fuserleer. That annoyed me incredibly that you were spamming the thread with irrelevant analysis because designs can only be analyzed holistically. Again you were attempting to analyze a DAG piecemeal based on your understanding of a Satoshi block chain, which is entirely erroneous. Ditto your analysis of Raiblocks was analogously flawed. |
|
|
|
"Orphaned" indeed is used in the whitepaper, but if you re-read my reply you'll see that my answer doesn't contradict it.
You are good at obfuscation and word-smithing. You know damn well that he doesn't understand that a DAG is probabilistic and not deterministic. And you are not trying to help him understand. Instead encouraging his myopic posts that assume a DAG has some deterministic structure. There is nothing childish about my responses. Purely facts. |
|
|
|
@Come-from-Beyond in Iota, do branches ever get orphaned? The whitepaper isn't clear on this.
All valid transactions (even conflicting) exist in the tangle and are not removed from it. "Orphaning" makes no sense in this model. That's what I thought. So, for clarity's sake, if there is a double spend on a branch, and valid transactions reference the double spend, are the subsequent valid transactions rendered invalid? Please stop and wait for my eludication. You are asking questions that don't make any sense. Your graphs are irrelevant. The security model is probabilitistic, not deterministic! You can't seem to get that into your hard head. If you didn't study computer science, please understand that you do not then understand the distinction between a probabilistic and deterministic algorithm. Wait I need to eat breakfast first. I went directly from bed to posting. Need to eat (being that I am still partially human in spite of my Cyborg implants). |
|
|
|
@Come-from-Beyond in Iota, do branches ever get orphaned? The whitepaper isn't clear on this.
All valid transactions (even conflicting) exist in the tangle and are not removed from it. "Orphaning" makes no sense in this model. Yet the word 'orphan' is in the Iota/Tangle white paper and he is obscuring from you what is your misunderstanding monsterer. Wait I will be replying to your posts monsterer because you continue to make the same egregious error because you still have not yet comprehended a DAG at all. The attacker's idea is to make the nodes reference the parasite chain, so that the "good" tangle would become orphaned. Read-only eh?  |
|
|
|
Reading my newest thread in Altcoin Discussion reveals that none of the other designs improve upon Bitcoin in terms of the decentralization, permissionless attribute. And I strongly expect that designs without a block chain (Iota and eMunie) will be proven fundamentally flawed. Open-transactions anyone?  Hey n00b, this white paper is nonsense because no where does it mention "double spend". My comment included the requirement for decentralization. The entire OpenTransactions concept is flawed. I had long ago critiqued it. It is crying shame to see someone waste so much effort implementing something that was not even well researched. Sigh. Thanks, got more? Still, the notary model, with multi-sig voting pools or other mechanisms mentioned in the paper, is a demonstration to what you said is not possible. Last I checked, even Monetas was doing a great job to make in happen IRL. Are you retarded? I wrote in the prior post that it isn't not a counter example about the decentralized, permissionless attribute which is the attribute I am writing about in this thread. Of course anyone can make a centralized payment system. Please stop posting noise here. Bye. Your level of technical understand is entirely inadequate to even comprehend what you are writing about. Please it is noisy and painful to watch you demonstrate in public that you are an utter fool and hiding behind a newbie sock puppet account. |
|
|
|
I skipped deep understanding of it and just noted it was for defeating lie-in-wait attack chains
Let me furnish you with a deep understanding of this monte carlo sampling technique; for the uninitiated, monte carlo sampling is a technique which takes a very complex data set, and samples it at random locations in order to build a probabilistic model for the truth of the underlying system, which may itself be intractable to process directly. An example, here is a monte carlo rendering of a scene at an increasing number of samples:  The idea being that this sampling process is orders of magnitude faster than processing the full data set (in this image, it would be fully raytracing the entire scene, with multiple light bounces / radiosity etc). In Iota, this technique is simply used to find the tips with the most cumulative weight (as I have been saying all along) but without having to process the entire set of transactions all the way back to genesis from every tip, which would be too slow. This is their game theoretically optimal choice as participants who want timely confirmation of their transactions. TLDR; Iota uses monte carlo sampling to find the tips with the most cumulative weight in order to decide where to extend the DAG. That is wonderful technology. And I am sure it made you guys feel proud (which probably blinded you to the higher level generalized flaw). But often in life, higher level semantics roost. In this case, we can't prove that all actors in the system are going to be incentivized to follow any specific selection algorithm for choosing which transactions to reference in their new transactions. The game theories are unfathomable. Again if you can show a comprehensive math of all the game theories as Satoshi has done, then you will be on par with the trust in Satoshi's design. Until then, you've got some fancy math without a fundamentally sound home. Iota can I am sure get many n00bs to think you have something valuable. Perhaps that is why you don't like when I call a n00b a n00b. I will not be be actively trying to spread any information about Iota. I am interested in the technical understanding for my own work. Corporations and serious users will be concerned with the fundamental soundness of the security. |
|
|
|
Let me interject a post here for layman to see I can try to explain what CfB and I are discussing about. This won't be comprehensive, just quickly off-the-top-of-my-head.
As monsterer was incorrectly doing, someone coming from analyzing a Satoshi block chain will tend to view a DAG as a plurality of block chains that are interleaved. And they will tend to analyze the security from that incorrect perspective. Even that was my first inclination. Since I've been analyzing my own consensus designs in painstaking detail, I have grown more accustomed to thinking about these matters such that I can quickly understand that a DAG can't be modeled the same way as Satohi's design.
Rather (the math dictates that) a DAG can only be analyzed according the selection rule that payers employ to choose which (two in Iota's case) transactions they point back to from their new transactions. The Iota white paper enumerates a few possible selection algorithms, but it can't be comprehensive because it would require perhaps up to infinite pages to enumerate. And payees play a role also because they decide which probabilistic calculation to trust in terms of the probability that their payment won't be reversed by a double-spend. An added wrinkle which further explodes the complexity is that payers and payees can not be controlled and can use what ever probabilistic algorithm they want (not necessarily the one chosen for them in Iota's default client) because there is no way to enforce this. So we have to add the game theories of what payers and payees might be incentivized to do from the myriad of possibilities of the interplay of all these factors. The complexity is unfathomable. I can't even wrap my head around all those possibilities (and I doubt anyone can) and yet I have already in one day alleged a new flaw that apparently the creators were not aware of before?
In Satohi's design, the selection rule of which transactions to include is only one possibility. That is a block of transactions. And the confirmation rule is the single Poisson distribution probabilistic calculation in Satoshi's white paper. Thus we are able to analyze and prove the performance of Satoshi's PoW. With a DAG, I doubt we can ever prove anything. We can just hope it won't blow up. Sorry but that is not acceptable for money.
From this, you can see why I say no design that doesn't use a block chain will work. I can imagine all the myriad of special cases Fuserleer has been rummaging through and he still won't catch all of them, because without a LCR (single longest chain rule) the number of special cases is unbounded. And this is why I needed all the details of Fuserleer's design, because I can't rely on his own interpretation of what is important to convey in a summary of his design. I have to think for myself the flaws he hasn't found and for that I need all the details.
The distinction is precisely that in Satoshi's design we have a single reference point which is the LCR, which is our clock. And these block less designs we have no reference point, so the possibilities are unbounded in potential complexity.
It is just math.
Unfortunatelly, your math doesn't contain math notation. It's more philosophy than math. I can't imagine how philosophy would look.  If seriously, Bitcoin forks showed that game theory can't be ignored. For instance, one of the latest forks happened because half of miners didn't have full blockchain. They were sticking to LCR but failed. What does this say to us? It says your math is flawed, doesn't it? You are confusing and conflating CATEGORICALLY separate concerns. As you may know, separation-of-concerns is a critically important paradigm in computer science. In other words, you have a category or ontological error in your response. Forking in Satoshi's design is tolerated by the protocol. The longest chain wins. Those miners who cheated on verification in order to propagate faster suffered economically by losing block rewards. The protocol security model worked as expected with the correct incentives. The only danger would be due to the Anti-fragile centralization in Satoshi's design, which is what my design attempts to fix. So the key point is that the recent Bitcoin event was not a protocol violation, because the provable security of the longest chain rule was not violated and the economics of the protocol were not gamed in any way that the single understanding of the protocol expects. Whereas, a DAG has an unprovable protocol security! Because there is no single characterization of the security or protocol. The protocol and security is unbounded in complexity. Yikes. If you can't translate the words above into a mathematical understanding, I am sure someone can formalize it for you in the future. I know though you are just being facetious and insincere when you pretend to not understand. |
|
|
|
If the attack is profitable, the attacker can buy as many botnets as are available on the market. There will probably be hacks on IoT devices too. More devices, more bots.
What does this change? The attack I described appears to be instead of an all-or-nothing race to be the longest chain competing against the entire network hashrate as considered in the attack models of your white paper, rather a fine-grained attack that can work at any percentage of the network hashrate with increasing deleterious effect. Also remember my main point is the general one, which is the unprovable nature of the security model. |
|
|
|
That was a factual statement. He is a n00b as proven by his inability to even comprehend what he was rolling his eyes about. His post was not in this thread, but it pertains to this thread. So I replied to it here. Don't be a B lister and conflate fact and ego. That eludication was upthread. I won't argue about this. I'm going into read-only mode until the quality of this thread comes to a higher level. Typical B lister behavior. Now you pout/sulk about reality. I am "blithely unconcerned that a bystander might think I’m egotistical" or that my comment is uncivil. His rolling eyes was uncivil and all I spoke are facts. Facts seems to bother you. What am I supposed to refer to him as, "expert", "concerned cordial commentator", ... none of those would be factual. Factual is he demonstrated he is a n00b. He also demonstrated that he is an uncivil asshole with the rolling eyes, but I didn't go that far to be totally uncordial. I attempted to maintain a balance between civility and a hand-slap for his Dunning-Kruger asshole-like posting (in this thread, but in the other thread where he posted, I called him an idiot which is my opinion). By responding about the single word "n00b" you make a mountain out of a molehill, you create several noise posts, and turn the thread into a noisy female bitch slapping event. I don't need to justify my behavior. I will not make an explanation like this again to try to explain to B listers why it is their weakness, not mine. |
|
|
|
Hey n00b, this white paper is nonsense because no where does it mention "double spend". The entire concept is flawed. It is crying shame to see someone waste so much effort implementing something that was not even well researched. Sigh. I suggested to make this thread self-moderated to keep the discussion civilized. It was acceptable while you were doing bold claims and bragging, but name calling is a little bit too much... That was a factual statement. He is a n00b as proven by his inability to even comprehend what he was rolling his eyes about. His post was not in this thread, but it pertains to this thread. So I replied to it here. Don't be a B lister and conflate fact and ego. That eludication was upthread. |
|
|
|
Reading my newest thread in Altcoin Discussion reveals that none of the other designs improve upon Bitcoin in terms of the decentralization, permissionless attribute. And I strongly expect that designs without a block chain (Iota and eMunie) will be proven fundamentally flawed. Open-transactions anyone? . That is payback for the :rolling eyes: |
|
|
|
Reading my newest thread in Altcoin Discussion reveals that none of the other designs improve upon Bitcoin in terms of the decentralization, permissionless attribute. And I strongly expect that designs without a block chain (Iota and eMunie) will be proven fundamentally flawed. Open-transactions anyone? Hey n00b, this white paper is nonsense because no where does it mention "double spend". My comment included the requirement for decentralization. The entire OpenTransactions concept is flawed. I had long ago critiqued it. It is crying shame to see someone waste so much effort implementing something that was not even well researched. Sigh. |
|
|
|
long running hedge funds will rarely if ever underperform S&P.. in a crazy bull market that we are in its easy for you to dollar cost average but in choppy markets hedge funds win. Hedge funds and mutual funds spread the risk and allocate more efficiently, the ones that underperform are quickly replaced by ones that don't
You being a dumb person who makes dumb posts regularly have failed to consider Anti-fragility in your statement above. I have no time to explain. This is payback for your inane comments in my recent Altcoin Discussion thread. |
|
|
|
|
Show Posts
