--- Log opened Tue Sep 17 00:00:46 2013
00:04 < amiller> the sad thing is i'd like to actually support pooled mining
00:04 < amiller> like if people's motivation is to lower their variance, there's nothing bad about that
00:04 < amiller> especially if they have their own hashpower
00:04 < amiller> it actually supports decentralization to support something like that
00:05 < warren> what's wrong with p2pool's approach?
00:05 < warren> p2pool implementation has scalability problems and payouts are too often in too  small dust, but that's a current implementation issue.
00:07 < amiller> well p2pool's approach is based on the same technique that makes hosted mining feasible/attractive
00:07 < amiller> (despite the fact that no one does it yet)
00:08 < warren> I mean, if users were more concerned about the risks of mining centralization, they would use p2pool-like approaches, there could be multiple of them.
00:08 < warren> p2pool needs to be a lot more efficient than it is now.  We hope to throw a few thousand dollars into its development.
00:08 < amiller> well see the thing is the risks of mining centralization aren't felt by individual users acting in self interset
00:09 < amiller> it's kind of like a social cost
00:09 < warren> amiller: p2pool miners can earn more than centralized pool mining
00:09 < amiller> warren, i am not talking about centralized pool mining
00:09 < amiller> i'm talking about hosted mining
00:09 < amiller> where you rent cpu power from a miner warehouse somewhere in the cool fjords of sweden
00:10 < amiller> where the hydroelectric power is cheapest
00:11 < jgarzik> Alydian is doing that
00:11 < jgarzik> $0.5 million for a petahash or three
00:11 < amiller> ah, thanks jgarzik
00:11 < jgarzik> though not necessarily in sweden
00:11 < jgarzik> knc and a couple others are doing hosted mining
00:11 < amiller> are there threads panicking about this
00:12 < jgarzik> and well over a year ago, "Vladimir" on the forums sold hashes in this manner.  you paid for a certain amount of hashes (GPU at the time).
00:12 < jgarzik> nope, it's already been explored
00:14 < amiller> already been explored? what conclusion did they come to? (i'm searching for such threads)
01:17 < gmaxwell> nanotube: amiller's plan to foil cloud mining is like julian assange's plan to use leaks to undermine secrecy. :P
01:18 < gmaxwell> I don't think anyone has explored foiling it through clever techno-economic hacks.
01:19 < gmaxwell> (nor do I think amiller's ideas would ever go anywhere, but they may someday turn useful should bitcoin fail to centralization)
01:19 < gmaxwell> (so that the $next_thing, in 100 years when people will finally trust a next-thing, won't have the same flaw)
01:21  * amiller can wait
01:21  * nanotube also plans on being around in 100years.
01:22 < nanotube> assuming we don't have a major cataclysm, seems within the realm of possibility
01:23 < gmaxwell> amiller: I can defeat your approach. :(
01:24 < gmaxwell> I have some independant hardware maker build my hardware with an odometer, and the hardware gets audited by people with electron microscopes (at random, which I can afford because I'm mega cloud)
01:25 < nanotube> we just need to make bitcoin asic coffeemakers and spaceheaters.
01:26 < nanotube> and have them default-set to mine solo.
01:26 < gmaxwell> yea, I've argued that before: for low level waste heat decentralization is actually more cost effective... but deploymens seem to suggest that I'm wrong.
01:26 < gmaxwell> er deployments.
01:26 < nanotube> once we have millions of these out there, no need to worry about it.
01:26 < nanotube> there are deployments?
01:27 < gmaxwell> alternatively, I just run my cloud business such that I pay the average expected payout regardless of the actual payout, and I hire trained assassins to patrol my datacenter to catch theiving techs.
01:27 < gmaxwell> nanotube: there are a number of big online highly centeralized deployments, e.g. asicminer and the 200TH mine that most of the bitfury parts went to.
01:28 < nanotube> gmaxwell: well yes, but there are no deployments of relatively cheap consumer hardware that mines automagically with no user intervention.
01:28 < gmaxwell> cointerra's original business plan was that, but the club to the head that they need to sell stuff was strong enough, but I don't know if they were just delayed or really deflected, see: http://cointerra.com/about/ "Our mission is to become a reliable and trusted node for transaction clearing on a stable and flourishing Bitcoin network."
01:28 < gmaxwell> no no right.
01:29 < gmaxwell> I'm saying that my theory that decenteralized is more efficient than centeralized because the waste heat is more productively disposed of may be wrong.
01:29 < gmaxwell> because I'm seeing lots of centeralized deployments and there is no bitcoin coffeewarmer.
01:29 < nanotube> hmm
01:30 < gmaxwell> I dunno why it's wrong, I certantly lived it in VA. with substantially free power in part of the year because mining completely replaced heating costs.
01:30 < gmaxwell> (realistically the heatpump was probably 2x more power efficient, still... half price power is good)
01:31 < nanotube> maybe because nobody's gonna buy a 3000-dollar spaceheater. :P
01:31 < nanotube> the bfl jalapenos could have been it... but bfl fscked up, as we all know.
01:32 < gmaxwell> well the actual cost of building these things is ... not that high. I titter a bit at the forum people "why would they sell them when they could mine!"  "because you morons will pay a kings randsom for the hardware!"
01:32 < nanotube> hehe
01:44 < Luke-Jr> lol
01:45 < petertodd> Why mine when you can sell the hardware and make debt payments now?
01:46 < Luke-Jr> petertodd: and make a nice profit until you actually ship!
01:46 < petertodd> heh
01:46 < Luke-Jr> bah! Qt 5 requires Perl 5.16
01:46 < petertodd> <shudder>
01:46 < Luke-Jr> not sure I want to upgrade to testing perl
01:46 < petertodd> awful, horrible language
01:47 < Luke-Jr> Perl is lovely.
01:47 < Luke-Jr> I think I prefer to stick to stable versions though
01:48 < Luke-Jr> OH! That's how I can get the election by a landslide!
01:48 < Luke-Jr> "I know Perl. =_="
01:48 < petertodd> I don't vote for the mentally ill.
01:48 < Luke-Jr> :P
01:48 < petertodd> well, at least *that* kind of mentally ill...
01:49 < Luke-Jr> Perl is the kind of thing where you hate it until you're familiar enough with it. :P
01:50 < petertodd> yeah, I got familiar with it then went to art school...
01:51 < Luke-Jr> I wrote an emulator in Perl once! :P
01:52 < petertodd> heh, of what? line noise?
01:52 < Luke-Jr> it was one of my toy MIPS emulators I think
01:52 < petertodd> I hope you ported perl to it
01:52 < Luke-Jr> :D
01:53 < warren> I don't know who to vote for.
01:53 < petertodd> I wonder what's the longest chain of emulators ever emulated?
01:53 < warren> There's no Clinton on the ballot.
01:53 < petertodd> I was hoping to vote for the other lizard.
01:54 < phantomcircuit> petertodd, well someone wrote a Z80 emulator for a Z80 and then ran it on x86
01:55 < petertodd> phantomcircuit: I was more thinking Arthur Ganson's "Machine with Concrete" - https://www.youtube.com/watch?v=5q-BH-tvxEg
03:15 < petertodd> Random number generator: https://www.youtube.com/watch?v=a6aicIcQJvc
03:15 < petertodd> and sublime work of work
03:16 < petertodd> ganson is a genius
15:53 < gmaxwell> amiller: am I correct in beleving that just having basic pairing operators (gt* gt/ g1^ g1+ gt= and loads of g1 types) is all we'd need to verify pinocchio in script?
16:05 < amiller> gmaxwell, yes definitely.
16:07 < amiller> gmaxwell, i think it would be easy to implement using PBC
16:08 < amiller> pinocchio requires a few specific twist curve
16:09 < amiller> they have two curves basically
16:12 < gmaxwell> amiller: In the SCIP they mention they have selected a curve with a particular efficient endomorphism, I assumed this was just distortion map optimization and would already be in pbc.
16:12 < gmaxwell> (I guess its a requirement that the curve and its quadratic twist have the same embedding degree?)
16:14 < gmaxwell> In any case, I was just musing on what the minimal cryptographic extensions to script were to achieve the widest increase in applications.
16:14 < sipa> OP_X86
16:15 < Luke-Jr> P2SH-for-SCIP would be useful
16:15 < amiller> i don't actually know any details about how pairing based crypto works, i only understand it at the bilinear map layer
16:19 < amiller> i may end up trying to learn it in a hurry and implement the pinocchio verifier myself :/
16:19 < amiller> of course for efficiency it's always hard to find the right abstraction
16:24 < amiller> https://crypto.stanford.edu/pbc/manual/ch08s08.html this are the BN curves y^2 = x^3 + b i think pinocchio uses
16:26 < gmaxwell> ah, okay, yea, I would have assumed it was though out of the ones in PBC. I still don't exactly understand how the pairing operation isn't slow as @#$@# for k=12 but apparently its not.
16:46 < amiller> the pinocchio guy said a similar thing once, that they picked a specific curve and used a lot of curve-specific implementation optimizations
16:46 < amiller> but maybe it's just this distortion map thing you're mentioning
19:44 < gmaxwell> So— perhaps this was obvious, but I realized that a sensible way to go about establishing the usefulness and correctness of a new scripting system for bitcoin is to implement it, and embed it in a harness that uses it as the controlling criteria in a signing oracle.
19:45 < gmaxwell> e.g. you take your script, hash it, compute a new public key from the oracle's well known public key. Then do things where you want the oracle to sign with that key... then go present the oracle your script and when it accepts it signs for you.
19:46 < gmaxwell> so then you could make any new application for your new bitcoin script opcodes you want, with the limitation that you depend on a trusted oracle.
19:46 < gmaxwell> But if the usefulness of the improved script is established then thats the on-ramp to making it part of the distributed system proper.
19:56 < amiller> that's a neat idea.
19:57 < amiller> that would work e.g. for zerocoin
20:04 < phantomcircuit> this is driving me insane
20:05 < phantomcircuit> i cant get the block header that cpuminer is finding from the info stratum provides
20:05 < phantomcircuit> even after adding a bunch of debugging stuff to cpuminer it's still not matching
20:56 < gmaxwell> amiller: it would also be very simple to implement.
--- Log closed Wed Sep 18 00:00:51 2013