--- Log opened Thu Jun 27 00:00:21 2013
--- Log closed Thu Jun 27 03:30:08 2013
--- Log opened Thu Jun 27 03:30:26 2013
--- Log closed Thu Jun 27 17:55:35 2013
--- Log opened Thu Jun 27 17:56:18 2013
22:51  * jgarzik sends a draft of the decentralized identity sacrifice protocol off to petertodd 
23:20  * jgarzik also tries to figure out some semi-decentralized method of conducting an ebay auction, where buyers bid ever-increasing amounts with proven funds
23:28 < petertodd> #2 sounds really similar to the fee auction process you know...
23:28 < petertodd> but we don't have the ability (yet) to lock a txout in any way which makes a in-Bitcoin port tough
23:29 < Luke-Jr> sounds easy enough?
23:30 < Luke-Jr> the seller can just publish an output which is used as an input to bids
23:30 < Luke-Jr> bidders sign transactions consuming it
23:30 < Luke-Jr> seller only signs the one winner
23:31 < petertodd> you mean a multisig output?
23:33 < petertodd> ah, I see, the output is to ensure only one tx, IE bid, can go through, simple enough
23:35 < Luke-Jr> yep
23:35 < gmaxwell> on the subject of random crypto protocols, I came up with one so that a movie renting place could rent you a single movie while learning nothing about which of the movies they loan is the one you picked.
23:36 < gmaxwell> (I came up with it on the spot when I was explaining to someone how funny crypto things like you make trustless protocols for things that model relationships people want to have— and they suggested keeping your movie preferences private as an example)
23:37 < Luke-Jr> haha, that's the opposite direction rental businesses want to go I think :P
23:38 < gmaxwell> (You encrypt all the movies and give them to them; you also encrypt all the movie keys with homomorphic encryption and give them all the E(Renter_key,K)s. They pick the movie they want and compute E(rentee_key,E(Renter_key,K)) for the movie they want and then ask you to decrypt it.)
23:38 < gmaxwell> (and you only decrypt one key for them)
23:39 < realazthat> you can still eavesdrop which they download though
23:39 < petertodd> jgarzik: re decen identity: you seem like you have a protocol that doesn't create a strong proof after the fact that the sacrifice was genuine
23:39 < realazthat> unless they download all of them
23:39 < petertodd> jgarzik: it's almost but not quite a proper announce-commit
23:40 < realazthat> erm
23:40 < realazthat> I was thinking of a similar scheme
23:40 < realazthat> nvm
23:41 < realazthat> in ur case you stated they get them all
23:41 < gmaxwell> realazthat: yes, they get them all. I do not know of a way to do a _efficient_ oblivious database where the reader and writer are different parties.
23:42 < petertodd> Luke-Jr: nah, gmaxwell's protocol still works for that: just give people the *option* of using this fancy feature, if they do, they're almost certainely renting Enemy of the State
23:42 < petertodd> realazthat: use a DHT
23:42 < gmaxwell> Luke-Jr: well part of my point is that one of the reasons businesses go the route of watching everything is that it's easiest to do that, and "impossible" for a superior business partner to prove that they aren't.
23:43 < Luke-Jr> gmaxwell: it's also useful information
23:43 < gmaxwell> Except with cryptographic protocols a superior business partner can actually prove that they're not, and perhaps benefit from their superiority.
23:43 < Luke-Jr> gmaxwell: I often wish someone did some analysis of anime preferences, and recommended me ones I'm likely to enjoy
23:43 < gmaxwell> (Privacy is like the worst lemon market that there ever was)
23:43 < realazthat> lol
23:44 < gmaxwell> Luke-Jr: sure. I use movielens for things like that. But there is no need to force you into analysis which: doesn't benefit you, which loses more than it strictly needs to etc...
23:44 < gmaxwell> (http://movielens.umn.edu/login)
23:45 < Luke-Jr> gmaxwell: without enough sample data, it won't work
23:45 < gmaxwell> e.g. nothing would stop you from also submitting your movie preferences to another party— perhaps behind a pseudonym—  even automatically. So then no one learns more than they need to— and the party learning movie preferences is actually providing you with a useful service.
23:46 < gmaxwell> vs the renter doing it, and which they may just be selling the data to someone who wants to make a list of people with varrious politics in order to oppress or what have you.
23:56 < Luke-Jr> gmaxwell: I can't even find 5 movies I've seen -.-
23:56 < gmaxwell> hah
23:57 < gmaxwell> on movie lens?
23:57 < gmaxwell> "You've rated 371 movies." .. and there are a bunch I've seen but haven't rated because I don't remember them well enough to give them a rating.
23:58 < petertodd> needs a new rating: "Didn't rate; probably sucks"
23:58 < gmaxwell> I emailed them and pointed out that in the netflix data that saw vs not-saw actually had most of the predictive power.
23:59 < gmaxwell> (and suggested they add a "dunno; but I saw it")
23:59 < gmaxwell> but they didn't respond. :(
23:59 < gmaxwell> A "I made a conscious decision not to watch this" would be interesting too, I expect.
23:59 < petertodd> heh, maybe that's already mostly what their algorithm actually is...
--- Log closed Fri Jun 28 00:00:31 2013