New Mystery about Satoshi

[johnyj]

I like the lab scenario, it's highly likely he ran lots of machines to simulate a small sized P2P network

[Carlton Banks]

I like the lab scenario, it's highly likely he ran lots of machines to simulate a small sized P2P network

I believe this also, he would've had to have been almost too insightful with his various "goldilocks" values and also with the innovative distributed database in order to be happy that 0.1 worked well enough, and for none of these values (50 BTC blocks, 2016 block difficulty adjustments, 10 minute average block solutions) to run into trouble in real world use. This mystery of the clue to the construction of the genesis block only highlights this further IMO: Satoshi clearly had a range of self developed software tools that he used to test and develop the system in a networked setting, but only released the source for 0.1 client and the White Paper to the public.

This would have been very much a long term project for Satoshi, whether he was tackling it full time or in spare time. He devoted quite some resources to it. He was also highly motivated to seeding the network and concept amongst the public (or at least amongst interested computer programmers). The (still supposed) effort to produce the Genesis Block he wished for via trial and error also indicates strong motivation.

[Mike Hearn]

There were long gaps in 2009 when no blocks were mined at all despite min difficulty. I doubt Satoshi would have been running Bitcoin on public computers, given his preference for privacy.

Most likely there was just a bug that got fixed at some point.

[jackjack]

There were long gaps in 2009 when no blocks were mined at all despite min difficulty. I doubt Satoshi would have been running Bitcoin on public computers, given his preference for privacy.

Indeed

Maybe he owned 6 computers, each of them running ~10 instances of an own designed mining software.
Look how each block of 10 LSB bytes has roughly the same value.
To be clearer:

• LSB=0-9: ~450
• LSB=10-19: ~0
• LSB=20-29: ~550
• LSB=30-39: ~450
• LSB=40-49: ~330
• LSB=50-59: ~300

[Sergio_Demian_Lerner]

There were long gaps in 2009 when no blocks were mined at all despite min difficulty. I doubt Satoshi would have been running Bitcoin on public computers, given his preference for privacy.

Indeed

Maybe he owned 6 computers, each of them running ~10 instances of an own designed mining software.
Look how each block of 10 LSB bytes has roughly the same value.
To be clearer:

• LSB=0-9: ~450
• LSB=10-19: ~0
• LSB=20-29: ~550
• LSB=30-39: ~450
• LSB=40-49: ~330
• LSB=50-59: ~300

I don't see the difference (at least in the first 20K blocks). But if a divergence appears, I would say that there were 6 machines, each on them running a limited range. The owner of the machine running 10-19 decided not to be part of the Bitcoin experiment and left the group, and Satoshi were 5/6 people.
Nevertheless, as I said, I don't see any statistical meaningful difference.

[gmaxwell]

The nonce uniquing to avoid collisions seems unlikely to me. The blocks are already made unique by paying to different public keys. If nonce-index were the method used to unique a cluster of machines we would have expected to see pubkey reuse, but we do not. It could have both used the LSB and the pubkey, but I don't know why it would.  It may have also been rolling the nonce early in order to test the extranonce rolling.

I'd suggest looking at the actual code and see what it does… If the initial release has the same behavior "mystery solved".

[Mitchell]

I don't understand anything of what is said here. Anyone care to explain?

[jackjack]

There were long gaps in 2009 when no blocks were mined at all despite min difficulty. I doubt Satoshi would have been running Bitcoin on public computers, given his preference for privacy.

Indeed

Maybe he owned 6 computers, each of them running ~10 instances of an own designed mining software.
Look how each block of 10 LSB bytes has roughly the same value.
To be clearer:

• LSB=0-9: ~450
• LSB=10-19: ~0
• LSB=20-29: ~550
• LSB=30-39: ~450
• LSB=40-49: ~330
• LSB=50-59: ~300

I don't see the difference (at least in the first 20K blocks). But if a divergence appears, I would say that there were 6 machines, each on them running a limited range. The owner of the machine running 10-19 decided not to be part of the Bitcoin experiment and left the group, and Satoshi were 5/6 people.
Nevertheless, as I said, I don't see any statistical meaningful difference.

What do you mean "difference"? I'm talking about this:

There are the 6 values I've described
Looks like 59 was not used but if the theory is correct it means that the last machine was also the least powerful (even with only 9 miners on it) so he/they may have decided not to run as many miners as on the others machines

[Mitchell]

I don't understand anything of what is said here. Anyone care to explain?

He is looking for patterns in the way early blocks mined to try to determine if they were mined by Satoshi.  This is to get an estimate of how many blocks Satoshi mined and how many Bitcoins he (they) has (have) stashed away.

I see! That is really interesting then!

[jdbtracker]

what if he was using bi-endian machines? A single segment of servers? I can see Itanium processors doing this.

[Sergio_Demian_Lerner]

What do you mean "difference"? I'm talking about this:

There are the 6 values I've described
Looks like 59 was not used but if the theory is correct it means that the last machine was also the least powerful (even with only 9 miners on it) so he/they may have decided not to run as many miners as on the others machines

Oh yes, that graph in the post included up to block 36K. I was analyzing the graph up to block 20K and I didn't notice the difference.

Now there are two choices:

1. there were originally 6 "machines", whatever those machines were. One of the machines broke or was unavailable.

2. there were 58 machines, divided in 6 labs. Each lab had a certain kind of computer, and models were slightly different with +-50% performance.

I like more the second since 59 is not divisible by 6.

Good catch jackjack !

[gmaxwell]

The original nonce update code:

Code:

            if ((++tmp.block.nNonce & 0x3ffff) == 0)
            {
                CheckForShutdown(3);
                if (tmp.block.nNonce == 0)
                    break;
                if (pindexPrev != pindexBest)
                    break;
                if (nTransactionsUpdated != nTransactionsUpdatedLast && GetTime() - nStart > 60)
                    break;
                if (!fGenerateBitcoins)
                    break;
                tmp.block.nTime = pblock->nTime = max(pindexPrev->GetMedianTimePast()+1, GetAdjustedTime());
            }

An interesting characteristic here is that it only checked for new pindexPrev every 2^18 hashes performed locally. This may have resulted in work with higher least-significant-byte nonces ending up orphaned more often.

MakeNewKey() was called at the start of the mining process outer loop, and then the key only saved if a block was found.  On this basis, I'd say that anyone using their nonce for deduplication was confused or using very different bitcoin block creation (not just mining) software. Though it would have been very easy to change the starting value and increment.

[Ente]

Normally I don't bother with speculation about Satoshi.
But this analysis and writeup is beautiful!
Thank you for the inspiration!

Ente

[deepceleron]

As the extranonces on the Satoshi miner incremented much faster than released code, it is logical to say that not all of the nonce bitspace was searched before resetting the loop and incrementing the extranonce. This is just a small mining code tweak, just significant in that the miner can be identified as not using the "official" release client due to this. Occam's Razor. You can get the same effect by just changing the maximum nonce value at which the loop exits and starts again with a +1 extranonce, as demonstrated in the gmaxwell code above.

You don't need a huge mining farm to make the ~2-4 MHash/s of the baseline miner throughout 2009's difficulty 1. This could even be a prototype multithreading miner, give each CPU thread it's own unique two or three bits of the nonce. The only strange thing is the "gap" of missing nonces, but this may also be from mistaken assumptions leading the methodology of analyzing the bit use.



Figure 1.

[Rampion]

Fascinating research as usual, Sergio.

Going back to your first research (the one that proved that Satoshi is holding to aprox. 1M Bitcoins that never moved) I think somebody should write a script that immediately sends a warning if some of those coins are transferred… That would be probably the best time to dump everything and run for cover!

[Ente]

Fascinating research as usual, Sergio.

Going back to your first research (the one that proved that Satoshi is holding to aprox. 1M Bitcoins that never moved) I think somebody should write a script that immediately sends a warning if some of those coins are transferred… That would be probably the best time to dump everything and run for cover!

Poor Satoshi..
Reminds me of Midas - the greek guy which would turn everything to gold what he touches.. eventually starving!

Also, put that "1bitcoineaterdonotsend" address on that list ;-)

Ente

[Sergio_Demian_Lerner]

I think deepceleron was right, as this image proves it:



This is the same image as above, but Satoshi time is corrected by the 59/256 factor.
Now Satoshi computer looks like standard computer, but is it?

More analysis is on my last post:  http://bitslog.wordpress.com/2013/09/04/satoshi-machine-one-mystery-is-solved-and-another-opens/

Best regards, Sergio.

[vandeam]

any way to translate this to english ?  

---Edited---

Quote from: HELP.org on September 03, 2013, 06:57:47 PM
Quote from: bitcoininformation on September 03, 2013, 06:48:18 PM
I don't understand anything of what is said here. Anyone care to explain?

He is looking for patterns in the way early blocks mined to try to determine if they were mined by Satoshi.  This is to get an estimate of how many blocks Satoshi mined and how many Bitcoins he (they) has (have) stashed away.
I see! That is really interesting then!


Thanks.

[Atheros]

So this solves the mystery of Satoshi's miner being 4 times faster, and we can tell he had only one computer mining. Which, as far as I understand it, makes the current question: Why was Satoshi's miner skipping byte values 10 through 18 and then resetting after 58?

[jdbtracker]

What if it was overclocked? would that affect the outcome of a computation. I find that overclocking even when things look stable... are not quite right.