Quote from: Altcoinfanatic on July 10, 2015, 10:35:55 AM

Quote from: HR on July 10, 2015, 09:49:04 AM

Quote from: Altcoinfanatic on July 10, 2015, 09:16:15 AM

Quote from: HR on July 10, 2015, 08:50:17 AM

1. (And I think that 5 times safer from a marketing standpoint is good since we're working with 5 algos - from the most simplistic theoretical vantage point, you need to gain 51% control of 5 algos instead of 1 algo, and that is 5 times safer in spite of the fact that it's not mathematically 5 times when the 5 are considered as one joint value. Correct, or not?

2. Also, the diff of all the algos rises and falls in correlation with one another: if the sha256d diff rises, the diff on the other 4 algos rise correspondingly. Correct, or not?)

Good questions! I will try to answer 2 of them (as how i understand it, or is the community not allowed to answer technical questions? If so, i apologize and will stop answering them)

1. It was always true that an attacker could attack the coin with just 1 algorithm but they would have needed at least 87% if all algorithms were weighted properly. All algorithms are not weighted properly in Digibyte and thus only 61% on SHA256D is sufficient to attack Digibyte. If these problems get sorted (all algorithms weighted properly) then you an attacker can still attack Digibyte with only 87% with just 1 algorithm. How is that 5 times safer from a marketing standpoint? It is not true that you need to gain 51% control of 5 algos instead of 1 algo. So not correct.

2. No also not correct. This is one of the flaws MaNI was talking about:
I don't want to go into too much technical details but most of the flaws revolve around the fact that 'difficulty' is a somewhat arbitrary measurement, while it can be used to meaningfully compare two blocks from the same algorithm to one another, there is no real relation between the difficulties of two different algorithms. i.e. It is not really meaningful to say that a 500 difficulty Scrypt block is worth more or less than a 500 difficulty Groestl block.
https://bitcointalk.org/index.php?topic=554412.msg11804113#msg11804113

1. You still need an average of 51% over the 5 algos. From an easy to understand point of view, why is it technically incorrect to say that DGB is 5 times safer then?

2. I'm very interested is seeing how the 5 algos are not weighted correctly.
"As each algorithm controls only 20% of the network, a 51% attacker would need to control an average of 51% of each algorithm to successfully attack the network. For example, if someone brought in enough ASICs to capture 80% of both SHA256 and Scrypt, which would be very difficult on its own, that would still only amount to 32% of the overall network. The remaining 19% would need to be carved out of the Qubit, Skein, and Groestl algorithms, and there are no ASICs in development for any one of those."

1. That is how multi-pow SHOULD work and how it was intended to be. But the formula used currently does not reflect this. Currently only 61% on SHA256D algorithm is sufficient to attack Digibyte, you dont even need 1% of the other 4 algos. After the pull request of mentalcollatz will be merged (and if it will be merged) this will change to: an attacker with 90% of the SHA256D hashrate and 33% of each of the other 4 algorithms would have insufficient hashpower to mount a 51% attack. So basicly what we know today about multi-pow, it is not true. You dont need an average of 51% over the 5 algos since 61% of SHA256D is enough. So in plain english:
currently: the attacker only need hashrate in only 1 algorithm
after pull request is merged: the attacker must have some hashrate in all 5 algorithms

2. That is how multi-pow SHOULD work and how it was intended to be. But the formula used currently does not reflect this. In the current implementation there is no real relation between the difficulties of two different algorithms. Right now nodes consider the average sha256d block to contain much more work than any block from any of the other algorithms.

Good contributions. All I can say is that this needs to be thoroughly studied and documented with hard data. If you're correct, then we will have made a major improvement. If not, we run the risk of creating a major SNAFU.

Personally, until we have clear substantiated documentation on the subject (this would be a huge milestone that would make many people look like they would have been better off reading C++ for Dummies), I will remain skeptical and adhere to the original analysis and numbers backed by so many extremely experienced and knowledgeable devs. I find it incredible to think that a 51% attact would be possible with only 61% of one algo, except in the hypothetical case I concluded my last post with (and that is, in practice, impossible): The only way I see that this could be an issue is if it were possible to get huge strings of blocks discovered by only one algo in particular while the others find nothing, and that doesn't happen with DGB.