Bitcoins

[Mini-Geek]

Originally Posted by TheMawn

The best we could do is hand out coins for finding the first factor of an exponent. These results are impossible to forge.

The money handed out would have to reflect the bit depth. If it was one Mersenne Coin for a factor regardless of depth, we would have everything factored to 67 in about a week, which isn't all that helpful.

It could be based on GHz-days, and probably be near-impossible to forge or cheat. But it gives the wrong focus to the project. We'd have such an influx of TFers that to balance it with LL would mean severely over-TFing the numbers. So we might have, say, 100 times the current firepower, but only move forward roughly 10% faster. That's a terribly inefficient use of resources.

[TheMawn]

Agreed.

I can only think of ways to minimize the risks but nothing is foolproof. The best I can think of is a sort of super program for the GPU which includes TF, P-1 and LL all in one package, also taking care of submitting results itself (vs manual submission). Assignments are also handed out as packages, and you're given credit for completing the package (finding a factor or returning a composite result).

The program stores the residue for iteration X and iteration Y (different for each exponent, known only by the server) and these are submitted as well. Someone else can start that exponent from iteration X and see if the residue at Y is the same. If it's possible to keep X and Y secret, it isn't possible for the scummy user to bullshit those residues.

[LaurV]

Originally Posted by TheMawn

The program stores the residue for iteration X and iteration Y (different for each exponent, known only by the server) and these are submitted as well. Someone else can start that exponent from iteration X and see if the residue at Y is the same. If it's possible to keep X and Y secret, it isn't possible for the scummy user to bullshit those residues.

Not feasible. Don't confuse the few hex digits you see on your screen with a full residue, which is (for a 64M exponent) an ~8MB file (as they are computed mod Mp, which has 64 million bits, i.e. 8 million bytes). What you suggest is sending 8MB times 64, i.e. 500 MB* files for each assignment, plus the same for DC (totally 1GB!), which will need to be stored, cataloged, etc. on the server, and moreover, the server would have to do the work of ALL the users, divided by the percent of the iterations between X and Y, to "check" those residues. For example, for a 64M exponent, and 100 users reporting LLs in the same time (they are more right now!), the server would need to do 100/64=~1.5 times the work of an average user, to check all the expos. You will need a very powerful server...

*I considered you sent residues every 1M iterations. If you send them more often, the files will be much bigger. If you send them rarer, you will need more power in the server to do the recheck, because the server would need to do all the intermediary work between X and Y which he chooses, i.e. at leat 1M iteration. That is why I divided by 64. You still need to send ALL of them, i.e. send residues every Z iterations, or so, in spite of the fact the server will only need two of them. Otherwise, you can't keep the X and Y secret. Once you store them in the file (i.e. outside of the server) they are not secret anymore. Also, you can't send "partial" residues, or residues from "partial" LL tests, and you can't verify if I am sending you "real" residues (I can make two "fake" residues with 1M distance between them and send them to you telling you "this is iteration 37M, and this is iteration 38M, but in reality, I only made 1M iterations). I think you got my point.

[xilman]

[QUOTE=Mini-Geek;360841So we might have, say, 100 times the current firepower, but only move forward roughly 10% faster. That's a terribly inefficient use of resources.[/QUOTE]

So? Are you suggesting that a 10% efficiency gain is worse than no gain at all?

[f1pokerspeed]

Originally Posted by Mini-Geek

a) is irrelevant.
b) is a problem because (assuming you allow any ol' user to submit a DC) someone could easily pretend that he is both UserA and UserB, come up with a fake residue and submit both the first-time and DC. He'd get coins for at least one of those. You could mitigate this to an extent, e.g. by banning users who are discovered doing this, or verifying user's true identities before allowing them to submit any results, or refusing results that weren't first assigned by PrimeNet (no more manual assignments either), or leaving DCing to a trusted set of people, but none of these are foolproof.

Okay.... force a per-IP range account or something like that.

With regard to residues... SHA-256 hash them (or even something stronger)! Why would you need to send 8MB * number of iteration blocks per WU when you can send a small amount of bytes of data per iteration block?

[LaurV]

Originally Posted by f1pokerspeed

With regard to residues... SHA-256 hash them (or even something stronger)! Why would you need to send 8MB * number of iteration blocks per WU when you can send a small amount of bytes of data per iteration block?

I can hash any crap. My daughter pictures, or poetry. How do YOU know I did the (prime hunting related) work? Otherwise is like any other coin...

[chalsall]

Originally Posted by LaurV

I can hash any crap. My daughter pictures, or poetry. How do YOU know I did the (prime hunting related) work? Otherwise is like any other coin...

This is the question to be answered.

There must be a solution space to this problem.

(Not likely to be implemented quickly, but it's an interesting driving problem.)

[Mini-Geek]

Originally Posted by LaurV

edit 2: (I didn't visit that list linked above, for a long-long time, I took the link from the bookmark, but now I clicked it): You have to look to XPM Primecoin. If we don't move fast, someone else will do it hehe... Some comment from the thread, about XPM: "a coin which at least produces something is much better than bitcoin" hehe

This is a more practical alternative to "Mersenne Coins" because results can be verified without redoing all of the work. It works by finding Cunningham or bi-twin chains of primes. Finding the chain is difficult, but verifying that a claimed chain is valid is easy. Mersenne primes don't work because they are too rare, and Mersenne composites don't work because they are too difficult to validate.

[kracker]

Some miners have got to be insane.

https://images.weserv.nl/?url=thunde...e/pls5.jpg&fnr
https://images.weserv.nl/?url=i715.p...bf1b35.jpg&fnr

[LaurV]

Originally Posted by kracker

Some miners have got to be insane.

As I said in a former post, it pays the bills...

[kracker]

Originally Posted by LaurV

As I said in a former post, it pays the bills...

But you also said in previous post it is uneconomical to mine them.

It is already "uneconomic" to mine them. Without specialized equipment (read: ASICs) one wastes more money for electricity than he can get by selling the bitcoins he mines. Most people still mine them for fun, or because is "cool" to tell your friends that you have few bitcoins, or they are driven by the same adrenaline which makes you buy lottery tickets (yes, you still can be lucky to crack a block, with even the oldest hardware, or CPU, and there are about $15K in that block, so you won't think that you can waste the electricity one year and find nothing).

The second pics are 7950's.