The reason I ask is because I got curious over whether people have used insecure passwords to create brain wallets. For this purpose I've created 1.5 GB worth of addresses in that 20 byte raw format, out of a word list made by John the Ripper.
So, that same list in ASCII will be 3GB ... BFD, if you'll pardon my french.
When the block chain is loaded into memory, the operation will mostly be memory-bound. Loading an additional 3 GB into memory is not an option.
I've just ordered more RAM, to bring my machine up to 6 GB though. Although it hurt a bit to see that the same amount of DDR3 RAM was half the price of the DDR2 RAM my machine supports
So the reason is that even though there are only ~300 megs of different addresses in the block chain, there are (2^160)*20 bytes of possible addresses using the raw 20 byte format. Having the addresses lined up on 20 byte-boundaries also increases performance, I imagine, since I don't have to do any string parsing on the data.
IMO, the time it'll take to parse the ASCII will be dwarfed by
the time taken to chew through the blockchain while hammering
in a completely cache-incoherent manner at that 1.5GB hashtable
of restricts you just forced the parser to load and work with.
I'm currently working with 50 MB chunks (~2.6 million addresses) at a time, because of the aforementioned memory-restriction. How long do you think blockparser would take to parse that amount of addresses? Armory's Python tools does it in 120 seconds (not including the time it takes to load the block chain into memory).
Also ... I'm hoping you're planning to Do The Right Thing (tm) with
that attempt of yours. Nuff said.
Of course. I'm not a thief. Just curious.
) of all the bitcoins being traded in the past couple of years, how many of them (atomically) haven't exchanged hands at all in the past 2 years? Since all coins should be able to be traced back to the moment they were mined, it should be fairly trivial to see how many times and the dates when they were transacted. No?
