02:17:05moa:anybody aware of work analysing fee-paying transactions?
02:20:15gmaxwell:andytoshi: COOL re extension, usually what you want to do with tests is to flip one at a time, then two at a time. Sometimes multiple flips cancel each other out. Any way you could take a counter parameter from the command line? e.g. mutation_class_a=NNN where NNN is an index mod Nmutable_points_marked_with_class_a for which one to mutate?
02:20:45gmaxwell:that it was so easy to do even that ... making me want to use rust for more things.
02:31:42andytoshi:gmaxwell: pretty sure yes, tho idk how much expressivity i can use with command-line args ... i wasted a lot of time today reading through the libsyntax code and being ignored on IRC trying to add new unit tests (specifically, how to i get to the containing module of the function i just mutated so that i can search it for unit tests?? turns out you can't, so i had to restructure my code a little bit
02:31:44andytoshi:so it looks for a marker on the module itself)
02:32:27petertodd:are there debian/ubuntu packages for rust yet?
02:34:02andytoshi:petertodd: i don't think so
02:34:11andytoshi:aside from the binaries you can download at rust-lang.org
02:34:34petertodd:andytoshi: that's kinda my litmus test for "is this stable enough to bother?" :P
02:34:56petertodd:I read the debian discussion about the bootstrapping issue
02:36:34petertodd:(btw, by package, I mean third-party packages too, so long as they're reasonably trustworthy/well-maintained)
02:36:44gmaxwell:it's mostly a boostrapping issue that it's not packaged I think.
02:37:05petertodd:yeah, I'm getting that impression too
02:37:36rusty:Yes, but bootstrapping is kinda important.
02:37:40gmaxwell:I don't think anyone has rebuilt it from the original ocaml stuff from scratch. I found the rust team to be kind of insensitive to some of these concerns (and you may note that their package management thing enshrines a bunch of known-security-horrible practices from the ruby world)
02:38:24petertodd:gmaxwell: I saw someone's description of that process - literally dozens of intermediate steps
02:38:28gmaxwell:(There is at least a sense of 'this should probably be better' but not really appriciating how bad it actually is.)
02:38:37andytoshi:that is my impression also
02:39:00andytoshi:tho part of it is that they are extremely busy and it's easy to ignore stuff that isn't so user-visible; this may improve as things stabilize
02:39:04gmaxwell:petertodd: no one has mapped the shortest path I think, would be pretty computationally expensive to do so.
02:40:00gmaxwell:andytoshi: I talked to someone who worked directly on the package manager thing .. so it does some cool things like identifying things by hash, but there is no facility for digital signatures; so you end up with your security reducing to "trust random stuff on github/etc., but at least snapshot the hashes once you've trusted it the first time"
02:40:33andytoshi::/ that is very unfortunate
02:41:01petertodd:well, on the birght side once a sig system is slapped on that won't be so bad
02:41:05gmaxwell:looked like it would be easy to fix in the future.
02:41:06petertodd:halfway there
02:41:22petertodd:reminds me: turns out ripple labs doesn't sign any of their code
02:41:46gmaxwell:the URL support seemed pretty generic so it seemed to me that you could define a URL that encoded a pubkey(hash), and also required the things it fetched from it to be signed.
02:42:25gmaxwell:though I worry that since the ecosystem is starting without it, it'll be hard to fix it (or at least they'll be something deep in your dependency path which isn't authenticated at all)
02:43:02petertodd:gmaxwell: it's also telling peopl enew to the space that doing that is acceptable
02:43:56gmaxwell:at the same time, just dropping it on people might make it get used in a way that has only pretextual security.
02:44:06gmaxwell:e.g. github signing things for you. :P
02:44:39petertodd:nothing wrong with that so long as it's all the same key :P
02:45:03gmaxwell:there is also some politics around it that made them really scope creep resistant. (mozilla paid an outside developer to work on it and there were some delivery delays/challenges)
02:45:53petertodd:gmaxwell: might have helped had they included a minimal but explicit "null sig" mechanism that had no security, but forced the hooks to be there
02:46:03gmaxwell:petertodd: well I mean you putting your package signing key online on github.. would not be more secure than access to github and/or https.
02:46:39petertodd:gmaxwell: exactly! if you do that and github is honest and it's all with the same key, that's not *that* bad
02:48:10gmaxwell:Oh I see, use a github key. yea, makes sense, at least you know where you stand.
03:00:40gmaxwell:andytoshi: whers your mutation testing extension for rust?
03:00:42jeremyrubin:petertodd: reminds me, I downloaded a fresh copy of tor today and the signatures were broken.
03:02:21gmaxwell:I started downloading signatures on everything a while back and constantly find problems. E.g. for a long time openssh was signed with a key that couldn't easily be found anywhere.[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D[D (uh, or openssl?
03:02:28andytoshi:gmaxwell: on my system, and broken right now..
03:02:55gmaxwell:andytoshi: I was telling someone who works on servo about it (keegan) and he was quite interested.
03:03:05andytoshi:gmaxwell: cool!
03:03:22andytoshi:unfortunately i was an idiot and do not have the working code from earlier today
03:04:36andytoshi:in an hour or two it should be working again, i'm mired in implementation details of the syntax parsing phase
03:08:25andytoshi:..actually it would be very easy to get the code back to the way it was; i can do this if your friend is interested ... but it doesn't create new unit tests, just the mutated functions, and i need to keep working on the (currently broken) restructured version to add that ability
03:08:40gmaxwell:[meta] re the ontopicness stuff; sometimes things are more ontopic than it seems. E.g. the above stuff about the rust mutation testing has at least personal interest to several people here because its about the state of the art in software testing (and spawns out of andytoshi's wrapper for libsecp256k1).
03:08:52gmaxwell:andytoshi: well don't rush yourself; he'll still be interested later.
03:10:01gmaxwell:[meta] sometimes the traffic here of not really ontopic stuff (or really irritatingly ignorant stuff) gets ... irritating. I have no real solution. Some people have tried to make closed access versions of the channel (on several occasions, in fact); but I see a lot of value from public discussion so I don't join them.
03:10:35dgenr8:jeremyrubin: did you download the bad-sig tor over tor? I had to try 4 times recently to get a non-corrupted update of tbb over tor.
03:11:43gmaxwell:[meta] if people wanted I wouldn't be adverse to having a _short_ quiz to get voice or something like that; though I think it would really cut down the availablity of the channel. Really I'd probably prefer to spend administrative effort in making a digest or an index or something, rather than trying to moderate extensively.
03:16:32jeremyrubin:dgenr8: You're right I think.
03:17:48gmaxwell:we've had a few folks in here report bad certificates (seemingly) due to malicious tor exits. Ouch if they're hitting tbb transfers.
03:23:26jeremyrubin:Hmm I'm a little concerned my system may be compromised now because I did open the maybe-compromised tor bundle.
03:26:22gmaxwell:jeremyrubin: if its any consoluation it's probably good to treat your system as compromised every once in a while and reset yourself.
03:28:27jeremyrubin:gmaxwell: was thinking, "well I'll just check the binary to see if it matches a good one", but realized a good exploit would, once opened, put the correct binary in place. Your point is well taken...
03:29:42jeremyrubin:And a general reminder that sleep deprivation, and thereby inattentiveness, is probably the biggest security threat of all
03:29:59andytoshi:gmaxwell: https://github.com/apoelstra/halfsleep still does not create unit tests, but it works (and im past the problem that was stumping me :)) you can run `cargo test` on the project in the examples directory and you will see that the unit tests which use the mutated function fail
03:30:03kanzure:gmaxwell: i would prefer muting or kicking people who are incapable of /joining #bitcoin and continuing discussion there. a quiz is unnecessary.
03:30:52kanzure:nobody was suggesting that rust mutation stuff is off-topic......... what is clearly offtopic is stuff like "bitcoin needs to be reset; tell me again what source code is?"
03:33:17jeremyrubin:kanzure: I think the best policy is probably to send them a private message. I sent Relos a pm and discussed for few back and forths.
03:38:15andytoshi:btw some relevant context to this discussion is that kanzure has spent several years locating and wooing domain experts from many domains and bringing them to ##hplusroadmap, where he works to make sure the channel does not become a cesspool and they all leave
03:38:59andytoshi:tho i think the situation on -wizards is a bit different
03:50:15kanzure:yes my method in there is to scream at everyone until they fall over dead; i'm quite certain that method does not apply here.
04:05:52isis:isis has left #bitcoin-wizards
04:06:47lovecruft:jeremyrubin: hey, if you can give me any specifics on the exit node your were using, or you're willing to give me a copy of either the bad signature or the binary, i'm willing to personally assess the exit node's behaviour and BadExit it if necessary, and i and other would be happy to analyse the binary/sig
04:07:49jeremyrubin:lovecruft: I can't do that. Any record I have of such things was stored on my computer, therefore is potentially compromised as well.
04:08:37lovecruft:sure, but post-compromise analysis is far better than zero analysis
04:09:40lovecruft:anyway, i build and help develop tor browser, and i'd be horrified if any of our users were getting owned
04:11:07jeremyrubin:lovecruft: https://www.dropbox.com/s/x6s3xjgvavf764p/TorBrowser-4.0.8-osx32_en-US.dmg?dl=0
04:13:40jeremyrubin:gpg: Signature made Thu Apr 9 13:42:04 2015 EDT using RSA key ID D40814E0
04:13:40jeremyrubin:gpg: BAD signature from "Tor Browser Developers (signing key) "
04:13:51jeremyrubin:lovecraft: gpg error was along lines of the above
04:13:57lovecruft:lovecruft is now known as isis
04:15:26jeremyrubin:lovecruft: ^^^ I then googled for the error, when I read an article about tor devs messing up key signing and absent mindedly no big deal'ed it.
04:16:19jeremyrubin:I'm still not certain there was a real compromise, and it seems unlikely, but unless I can convince myself otherwise I'm in for a wipe.
04:18:06isis:hmm… i'm not sure about gnupg error, it sounds as if perhaps something happened to the download (could be a bad exit, or just a bad/incomplete download)
04:19:18isis:your binary matches the one on the build machines…
04:20:07gmaxwell:jeremyrubin: have a copy of the signature?
04:20:51jeremyrubin:isis: incomplete download actually seems most likely
04:23:53jeremyrubin:gmaxwell: https://www.dropbox.com/s/brjujsrwlmqn9aw/TorBrowser-4.0.8-osx32_en-US.dmg.asc?dl=0
04:30:34phantomcircuit:jeremyrubin, can you not link to the potentially compromised .dmg ?
04:31:03jeremyrubin:phantomcircuit: I did https://www.dropbox.com/s/x6s3xjgvavf764p/TorBrowser-4.0.8-osx32_en-US.dmg?dl=0
04:31:52isis:jeremyrubin: everything looks ok on that front. the only problems left that i can think of would be your gpgtools install, or perhaps some other gnupg problem, e.g. an outdated copy of the key.
04:37:14jeremyrubin:isis: Yeah I'm not overly concerned at this point. My downloading of the new version was over https, signatures checked out later (pointing to the partial dl being the issue), and binaries match. Although gmaxwell has a point with a periodic wipe never hurting. Thanks for looking into this :)
04:37:57jeremyrubin:isis: although in theory you could be the attacker trying to trick me over IRC ;)
04:39:03jeremyrubin:isis: I locally verified all of the above as well so I jest
04:40:12isis:quite true. all the more reason for you to become an anonymous gitian builder for bitcoind and tor browser, so that you can catch any developer hijinks or coersion.
04:42:51jeremyrubin:I'm down -- I'll set one up soonish (added to todo list)
06:50:32jinglebe_:jinglebe_ is now known as jinglebellz
07:07:59rusty:rusty has left #bitcoin-wizards
08:05:19rajaniemi.freenode.net:topic is: This channel is not about short-term Bitcoin development | http://bitcoin.ninja/ | This channel is logged. | For logs and more information, visit http://bitcoin.ninja
08:05:19rajaniemi.freenode.net:Users on #bitcoin-wizards: andy-logbot Burrito Crowley2k gielbier damethos Guyver2 hashtag_ dEBRUYNE zooko` p15 cbeams koshii devrandom hktud0 priidu MoALTz arubi luny b_lumenkraft unlord_ RoboTeddy Starduster DrWat jgarzik isis TheSeven Dr-G2 zwischenzug p15x jeremyrubin nessence afk11 d1ggy Firescar96 cluckj PaulCapestany bsm117532 Guest60260 antgreen smooth sofdg2 bosma sneak jtimon SubCreative Madars jhogan42 shesek sparetire sadoshi [d__d] hulkhogan42o wiz
08:05:19rajaniemi.freenode.net:Users on #bitcoin-wizards: leakypat tromp_ justanotheruser Pan0ram1x kefkius throughnothing_ amiller_ mm_0 harrigan_ sparetire_ rustyn gmaxwell ak maaku waxwing Xzibit17 adams_ forrestv Transisto richardus adlai Fistful_of_coins LeMiner PRab airbreather go1111111 prodatalab cornus_ammonis kanzure spinza maraoz morcos sdaftuar ajweiss Tjopper andytoshi helo Iriez cdecker copumpkin bedeho Guest96151 face Cory Kwelstr nsh nuke_ GAit Relos Krellan cfields_ wumpus mappum
08:05:19rajaniemi.freenode.net:Users on #bitcoin-wizards: btc___ jbenet NeatBasisW dasource btcdrak CodeShark dardasaba yorick aakselrod ebfull mkarrer_ Tiraspol tjader luigi1111w melvster binaryatrocity bliljerk101 grandmaster jonasschnelli merlincorey [ace] eric a5m0 nephyrin null_radix crescendo Sqt mikolalysenko sturles GreenIsMyPepper harrow vonzipper berndj Zouppen manan19 comboy realcr jaromil catlasshrugged_ Apocalyptic cryptowest_ runeks__ Graet Eliel veox indolering K1773R Keefe petertodd
08:05:19rajaniemi.freenode.net:Users on #bitcoin-wizards: jcorgan larraboj ryan-c jessepollak gribble tromp mr_burdell gnusha d9b4bef9 starsoccer dgenr8 BrainOverfl0w MRL-Relay azariah @ChanServ brand0 davout CryptOprah Oizopower Taek epscy Alanius nickler gwillen kinlo coryfields_ Muis catcow lnovy sl01 null michagogo STRML so phedny warptangent pigeons espes__ afdudley stonecoldpat roasbeef_ heath BananaLotus guruvan dansmith_btc cursive Meeh fluffypony optimator livegnik s1w AdrianG dignork
08:05:19rajaniemi.freenode.net:Users on #bitcoin-wizards: gavinandresen yoleaux nanotube luigi1111 lechuga_ EasyAt phantomcircuit Logicwax kyuupichan poggy Anduck wizkid057 fenn hguux___ mariorz TD-Linux warren BlueMatt midnightmagic huseby otoburb platinuum kumavis artifexd yrashk Luke-Jr HM dc17523be3 lmacken Hunger- SwedFTP weex
08:11:24cdecker:cdecker has left #bitcoin-wizards
11:48:29bsm117532:I've been thinking about making a *fundamentally* merge-mined chain. e.g. rather than mine against a single Merkle root on a single chain, make a list of several of them and mine it. That way the mined block would be acceptable on multiple chains, but this grows as O(N) in the number of chains you want to merge-mine.
11:48:59bsm117532:So I had the idea to mine against a Bloom filter of these Merkle roots from the different trees instead.
11:51:25Luke-Jr:bsm117532: the problems now are that 1) bloom filters have false positives, 2) you lose unique values
11:51:47bsm117532:I ran some numbers and found that...
11:52:12bsm117532:Bloom filters do have false positives, but you can adjust it by making the bloom filter bigger.
11:52:15Luke-Jr:with 1), someone can "mine" for a "collision"; with 2), the miner can hide multiple conflicting blocks in the bloom filter
11:52:52bsm117532:So I ran some numbers so as to make that statistically unlikely.
11:53:31bsm117532:What I seem to be finding is that by the time you get the error rate low enough, your Bloom filter is basically the same size or larger as the original list of hashes that it's storing.
11:53:45Luke-Jr:I think for that you really need statistical impracticality, not merely unlikely :p
11:53:58Luke-Jr:yeah, that's what I'd expect ;)
11:54:30Luke-Jr:bsm117532: is this part of a longer explanation and I should just let you finish? ;)
11:54:39bsm117532:That's basically it.
11:54:51bsm117532:I thought it was mildly interesting, so I thought I'd share.
11:55:14Luke-Jr:well, I've never tried it, but yeah, somewhat interesting to hear firsthand
11:56:47bsm117532:I was looking at e.g a Bloom filter with 512 bits, an error rate of 1e-32, but it only stores 3-256 bit hashes
11:58:00bsm117532:That's the "optimal" filter with those parameters.
12:55:08zooko`:zooko` is now known as zooko
13:42:44amiller_:amiller_ is now known as amiller
14:18:08giel_:giel_ is now known as gielbier
14:21:16c0rw1n_:c0rw1n_ is now known as c0rw1n
15:17:43fluffypony:pigeons: so what license is it under, then?
15:17:58pigeons:all rights reserved
15:18:05fluffypony:* fluffypony twitches
15:18:55pigeons:i love the RL response to the question
15:19:58phantomcircuit:pigeons, they lobbied themselves into a corner trying to lobby against bitcoin
15:24:26pigeons:looks like it should just refer to that html download page, but their response is weird
15:24:59fluffypony:if they remove it from the page text then that seems preemptive for removing it from the actual source?
15:27:00pigeons:the person pointing it out to made it seem they were relicensing the daemon source, but that doesnt appear to be the case, sorry
15:32:54andytoshi:if it isn't the case they sholud say that on the commit..
15:38:31pigeons:yes its weird
18:06:13PRab_:PRab_ is now known as PRab
18:46:38SrPx:Hello, please be aware a specific community is planning an attempt to "troll" this channel.
18:47:34phantomcircuit:SrPx, lots of them try that virtually everyday
18:47:51SrPx:Okay. I'm just reporting.
18:50:32jinglebellz:jinglebellz is now known as whale
19:08:42zooko`:zooko` is now known as zooko
19:59:34arubi_:arubi_ is now known as arubi
20:52:53SrPx:SrPx has left #bitcoin-wizards
21:43:42joshgn:joshgn has left #bitcoin-wizards
23:07:41zz_lnovy:zz_lnovy is now known as lnovy
23:39:02warren_2:warren_2 is now known as warren
23:39:04s1w-:s1w- is now known as s1w
23:39:28HM2:HM2 is now known as Guest4827
23:39:29_mm_0:_mm_0 is now known as mm_0
23:39:33s1w:s1w is now known as Guest58032