1 2012-11-08 02:52:19 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  2 2012-11-08 02:52:27 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  3 2012-11-08 02:52:28 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  4 2012-11-08 02:52:29 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  5 2012-11-08 02:52:30 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  6 2012-11-08 02:52:31 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  7 2012-11-08 02:52:32 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  8 2012-11-08 02:52:34 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
  9 2012-11-08 02:52:35 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 10 2012-11-08 02:52:46 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 11 2012-11-08 02:52:47 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 12 2012-11-08 02:52:48 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 13 2012-11-08 02:52:49 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 14 2012-11-08 02:52:50 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 15 2012-11-08 02:52:51 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 16 2012-11-08 02:52:52 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 17 2012-11-08 02:52:53 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 18 2012-11-08 02:52:54 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 19 2012-11-08 02:52:55 <kreal> [MOFO] why?
 20 2012-11-08 02:53:01 <Luke-Jr> kreal: /ignore it please
 21 2012-11-08 02:53:06 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 22 2012-11-08 02:53:07 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 23 2012-11-08 02:53:08 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 24 2012-11-08 02:53:09 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 25 2012-11-08 02:53:10 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 26 2012-11-08 02:53:11 <kreal> just wondering the reason.
 27 2012-11-08 02:53:11 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 28 2012-11-08 02:53:12 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 29 2012-11-08 02:53:13 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 30 2012-11-08 02:53:14 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 31 2012-11-08 02:53:26 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 32 2012-11-08 02:53:28 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 33 2012-11-08 02:53:29 <kreal> or maybe his daddy hit him?
 34 2012-11-08 02:53:32 <[MOFO]> do do do do
 35 2012-11-08 02:53:32 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 36 2012-11-08 02:53:33 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 37 2012-11-08 02:53:34 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 38 2012-11-08 02:53:35 <[MOFO]> Fuck the Jew for me! Fuck the Jew for me! You can do for me! http://www.youtube.com/watch?v=VuSei2pDDlY
 39 2012-11-08 02:55:20 <kreal> is there a .wb topdomain?
 40 2012-11-08 02:59:16 <xIsalty> no
 41 2012-11-08 03:00:03 <kreal> thank. no I didn't find it.
 42 2012-11-08 03:00:06 <kreal> could be nice though.
 43 2012-11-08 03:03:46 <kreal> also I would like .bit to be a real topdomain :)
 44 2012-11-08 03:07:51 <Hasimir> you could use .biz and pronounce the z more like ts
 45 2012-11-08 03:08:12 <kreal> hehe yeah
 46 2012-11-08 03:28:08 <BlueMatt> nvm
 47 2012-11-08 08:19:59 <Guest52889> how can I setup coinbasercmd ocmmands?
 48 2012-11-08 08:47:11 <Guest52889> what does b'\\xFA\\xBF\\xB5\\xDA mean
 49 2012-11-08 08:47:33 <Diablo-D3> you sunk my battleship!
 50 2012-11-08 08:48:35 <kinlo> it would help if you said you were trying to set up eloipool
 51 2012-11-08 08:48:41 <kinlo> at least - I guess you are
 52 2012-11-08 08:48:46 <kinlo> I don't really know
 53 2012-11-08 08:50:55 <Guest52889> kinlo: yes, I'm trying to set up eloipool
 54 2012-11-08 08:51:00 <Guest52889> It's the message header for bitcoin
 55 2012-11-08 08:51:05 <Guest52889> (Testnet).
 56 2012-11-08 08:51:17 <kinlo> see, if we need to guess, we probably wont help you
 57 2012-11-08 08:51:23 <kinlo> you gotta learn to ask smart questions
 58 2012-11-08 08:51:49 <Guest52889> sorry, I was just thinking if it would look familiar to someone who done it before
 59 2012-11-08 08:52:34 <kinlo> this is the bitcoin development channel, not the eloipool development channel
 60 2012-11-08 08:53:10 <kinlo> one would assume you'd be discussing a bitcoin client, and unless specified the reference client
 61 2012-11-08 08:53:31 <Guest52889> oy
 62 2012-11-08 08:53:50 <Guest52889> it's* actually part of the default bitcoin client (the message headers), but yeah
 63 2012-11-08 08:54:00 <kinlo> not in python format :)
 64 2012-11-08 08:54:05 <kinlo> in any case
 65 2012-11-08 08:54:12 <kinlo> you already answered your own question
 66 2012-11-08 08:55:15 <sipa> Guest52889: it's not the message header for either mainnet or testnet3
 67 2012-11-08 08:55:27 <sipa> maybe it's the message header for an older instance of testnet
 68 2012-11-08 08:55:58 <kinlo> sipa: it's older testnet, it's from the sample config luke provides
 69 2012-11-08 08:56:05 <Guest52889> Yeah it was a year ago
 70 2012-11-08 08:56:08 <kinlo> but he wrote his sample config when only testnet2 existed
 71 2012-11-08 08:56:09 <Guest52889> https://github.com/gavinandresen/Bitcoin-protocol-test-harness/blob/master/BitcoinClient.py
 72 2012-11-08 08:56:30 <kinlo> or not
 73 2012-11-08 08:56:31 <kinlo> see
 74 2012-11-08 08:56:48 <kinlo> that's why you need to learn to ask questions, you were not talking about eloipool
 75 2012-11-08 08:56:51 <kinlo> I guessed wrong
 76 2012-11-08 11:31:46 <sipa> ;;bc,nethash
 77 2012-11-08 11:31:47 <gribble> 25321.544560350707
 78 2012-11-08 12:51:09 <etotheipi_> I've got a fun idea, tell me if this is crazy
 79 2012-11-08 12:52:21 <etotheipi_> choose a string "This is secret but not really", and encrypt that into the header information of the wallet file with the encryption passphrase that is used for all the other sensitive data
 80 2012-11-08 12:53:01 <etotheipi_> then if a user forgets their passphrase, they can post that encrypted string on the forums with a bounty and information they know about their passphrase
 81 2012-11-08 12:53:30 <etotheipi_> then users who are motivated to help crack it, can do so without having the full wallet, and can prove to the owner that they have succeeded
 82 2012-11-08 12:54:05 <_dr> except if some guys can crack it a motivated attacker can, too
 83 2012-11-08 12:54:25 <etotheipi_> none of this changes the already-strong encryption on the wallets...
 84 2012-11-08 12:54:44 <etotheipi_> all it does is it adds a mechanism by which others can try to brute force the passphrase without having the full wallet
 85 2012-11-08 12:54:59 <drizztbsd> why?
 86 2012-11-08 12:55:13 <drizztbsd> universal question: why :P
 87 2012-11-08 12:55:18 <etotheipi_> tons of people on the forums post that they forgot their passphrase
 88 2012-11-08 12:55:38 <etotheipi_> their only solution is to post their encrypted wallet and information about their passphrase
 89 2012-11-08 12:55:43 <_dr> they should just upload their wallets and provide an address
 90 2012-11-08 12:55:46 <wumpus> you're basically adding a known-plaintext.. doesn't the current wallet already have some known plaintext areas that could be used for that?
 91 2012-11-08 12:55:51 <_dr> and pray that someone will split 50/50, haha
 92 2012-11-08 12:55:54 <etotheipi_> and hope that the person who gets it feels generous enough to give some back
 93 2012-11-08 12:56:14 <drizztbsd> it's better to remember passphrase or to usa paper wallet
 94 2012-11-08 12:56:15 <drizztbsd> :P
 95 2012-11-08 12:56:21 <etotheipi_> drizz, of course
 96 2012-11-08 12:56:53 <_dr> or you could implement security questions :) 'what's the name of your dog?'
 97 2012-11-08 12:57:08 <wumpus> for example you could just post an encrypted, unused private key
 98 2012-11-08 12:57:10 <etotheipi_> but it doesn't stop people who from forgetting their passphrase and not having a paper backup, anyway
 99 2012-11-08 12:57:20 <wumpus> and the public key for that
100 2012-11-08 12:57:24 <etotheipi_> wumpus: that's true
101 2012-11-08 12:57:29 <etotheipi_> well, almost
102 2012-11-08 12:57:38 <etotheipi_> actually no
103 2012-11-08 12:57:50 <wumpus> I'm sure it has some header bytes that could be used to verify you have the right decryptio nkey
104 2012-11-08 12:57:53 <etotheipi_> because the private key is not obvious when it's found
105 2012-11-08 12:58:13 <etotheipi_> you have to add to each guess of the passphrase, an expensive ECDSA operation to get the public key
106 2012-11-08 12:58:17 <sipa> wumpus: we only encrypt private key data
107 2012-11-08 12:58:18 <wumpus> as the current bitcoin client can also check whether you entered the right passphrase and won't just use invalid decoded keys
108 2012-11-08 12:59:48 <etotheipi_> I'm going to be making a new wallet format for Armory soon... I'm thinking of adding this as a feature
109 2012-11-08 12:59:59 <etotheipi_> I have had 3-4 requests in the past month for help recovering passphrases
110 2012-11-08 13:01:27 <gmaxwell> etotheipi_: we don't have an extraction tool but they could post their master key and some keypool key for cracking.
111 2012-11-08 13:01:56 <gmaxwell> Having a seperate passphrase doesn't make sense for your purpose, I think.. since an attacker with the local data could just brute force the weaker one. Might as well just have the weaker one, enh?
112 2012-11-08 13:02:14 <etotheipi_> gmaxwell: it's not a weaker passphrase
113 2012-11-08 13:02:37 <wumpus> hm it'd make it a lot cheaper to bruteforce the passphrase
114 2012-11-08 13:02:38 <etotheipi_> I'm saying it's a string known only to the owner of the wallet that is obvious once the correct passphrase is guessed
115 2012-11-08 13:03:01 <wumpus> currently it indeed does a ECDSA operation to check whehter the entered passphrase was valid
116 2012-11-08 13:03:12 <sipa> wumpus: which is actually not necessary
117 2012-11-08 13:03:29 <sipa> as the private keys are encoded using AES-CBC, which pads the keys from 32 to 48 bytes
118 2012-11-08 13:03:33 <wumpus> it isn't? how would it check otherwise if you only encrypt private keys as you said?
119 2012-11-08 13:03:36 <wumpus> hmm right
120 2012-11-08 13:03:44 <wumpus> so just check the padding.. that's too bad
121 2012-11-08 13:03:54 <sipa> yes, that's a bug, essentiallyu
122 2012-11-08 13:04:00 <etotheipi_> if the difference between being actually secure and not secure is adding an extra operation like that to each guess, you probably did it wrong
123 2012-11-08 13:04:16 <sipa> right, it doesn't break the security model at all
124 2012-11-08 13:04:36 <sipa> but the initial assumption was that it would require an ECDSA multiplication to verify
125 2012-11-08 13:05:03 <etotheipi_> sipa: and I like the intent of that... from the perspective of deterring attackers
126 2012-11-08 13:05:21 <etotheipi_> but I wonder if it's worth accommodating the all-too-common case of people forgetting their own passphrase and not having any kind of backup
127 2012-11-08 13:05:38 <wumpus> I don't think bruteforcing is the right way for that
128 2012-11-08 13:06:06 <etotheipi_> what other choice is there?
129 2012-11-08 13:06:24 <etotheipi_> (it wouldn't be strictly brute force, the owner of the wallet would post any information they know about the passphrase)
130 2012-11-08 13:06:42 <sipa> have the client mail an pubkey-encrypted-to-alert-key version of the plain wallet key to gavin
131 2012-11-08 13:06:50 <sipa> problem solved for us!
132 2012-11-08 13:06:56 <etotheipi_> the way Armory wallets are set right now...  the computer on which the wallet was created, would take about 500 years to guess a 6-char passphrase with no information
133 2012-11-08 13:07:06 <wumpus> some kind of key escrow, but the details are indeed difficult
134 2012-11-08 13:07:10 <etotheipi_> (900 years max)
135 2012-11-08 13:08:49 <wumpus> maybe some N out of M scheme with people that you somehow trust
136 2012-11-08 13:09:30 <gmaxwell> yea, escrow is nicer than bruteforce BUT  ... it must be optional in order to not be evil... but when you actually realize you needed it.. it was too late.
137 2012-11-08 13:09:45 <SomeoneWeird> yep
138 2012-11-08 13:10:28 <gmaxwell> OTOH, if it's just locally stored.. perhaps it could be on by default?  The risk then is that if the trustees are evil _AND_ they get a copy of your file.
139 2012-11-08 13:10:37 <wumpus> but bruteforcing also gives opposite incentives, you want your key to be hard to bruteforce but what if you want to recover it?
140 2012-11-08 13:11:02 <gmaxwell> bruteforcing also funds creating infrastructure we'd rather not exist or at least be as expensive as possible. :(
141 2012-11-08 13:11:18 <etotheipi_> wumpus: my guess is ... if you created the passphrase yourself, and you yourself "forgot" it... you are likely to know enough about it to make this operation feasible
142 2012-11-08 13:12:19 <etotheipi_> gmaxwell: if that infrastructure is going to exist, isn't it better for it to just happen, and for users to adjust to the fact it exists?  (by picking stronger passphrases, etc)
143 2012-11-08 13:12:44 <etotheipi_> (not that users will ever pick stronger passphrases...)
144 2012-11-08 13:13:13 <wumpus> sure but you'd want some technical specification for "information you know about your passphrase" I guess, otherwise people will have to write custom software every time to crack a passphrase :p
145 2012-11-08 13:13:16 <da2ce7> etotheipi_: or have a ???Encrypt my passphrase to my buddies??? option... where you you put the EC public keys of say 4 of your buddies, and set (need 3/4 to agree)...
146 2012-11-08 13:13:27 <gmaxwell> etotheipi_: It's also about the cost of it. encouraging people to post keys for bruteforcing makes it cheaper.
147 2012-11-08 13:13:56 <gmaxwell> da2ce7: you wont do that because you don't believe you're at forgetting risk until its too late. :(
148 2012-11-08 13:13:58 <drizztbsd> you can use SSS (Shamir)
149 2012-11-08 13:14:05 <drizztbsd> and split your passphrase to many pieces
150 2012-11-08 13:14:08 <wumpus> da2ce7: yes I like that idea
151 2012-11-08 13:14:17 <drizztbsd> and you can send one of them to your friends
152 2012-11-08 13:14:27 <da2ce7> gmaxwell: you may do it the _second_ time...
153 2012-11-08 13:14:31 <etotheipi_> da2ce7: but that only accommodates a use-case that very few people could even exercise
154 2012-11-08 13:14:33 <drizztbsd> when you lost your passphrase you can ask him the pieces and recomplete the passphrase :P
155 2012-11-08 13:14:49 <etotheipi_> a lot of people are creating offline wallets, and don't have lots of Bitcoin buddies
156 2012-11-08 13:15:10 <gmaxwell> etotheipi_: so why not have a default escrow set and a checkbox to replace it or deactivate it?
157 2012-11-08 13:15:15 <etotheipi_> and if they were going to go through that effort, it would be safer just to create a paper backup
158 2012-11-08 13:15:27 <wumpus> well if you have an offline wallet in a physically safe place, you could just give the full key to someone
159 2012-11-08 13:15:40 <da2ce7> etotheipi_: but your buddies could well be your 'other wallets'  eg... your phone, everyday wallet, and your brothers wallet.
160 2012-11-08 13:15:57 <etotheipi_> gmaxwell: I don't understand...
161 2012-11-08 13:16:23 <helo> isn't it better to just recommend best practices? i.e. paper
162 2012-11-08 13:16:54 <etotheipi_> helo: there is a reason I'm reminders all over the place in Armory, encouraging the user to make paper backups
163 2012-11-08 13:17:09 <etotheipi_> and they can't be encrypted, so they are guaranteed to have a plan B if something happens
164 2012-11-08 13:17:12 <helo> won't people rely on this, and then nobody will be able to bruteforce their passphrase?
165 2012-11-08 13:17:21 <etotheipi_> but that doesn't stop people from not taking those extra steps
166 2012-11-08 13:17:47 <helo> if they think there is some protection even if they do not, i think fewer people will do it
167 2012-11-08 13:17:52 <gmaxwell> etotheipi_: when you crate a wallet a dialog would say   [Add data the following people to cooperate to help you recover your wallet if you forget your passpharse [X] ]
168 2012-11-08 13:18:18 <gmaxwell> etotheipi_: then store in the wallet a copy of the derrived key (not the passphrase) which has been split and encrypted for those parties.
169 2012-11-08 13:18:35 <gmaxwell> and have some option to replace the people with your friends if you have any, or turn that off entirely.
170 2012-11-08 13:19:13 <gmaxwell> And a text field to indicate how you're going to identify yourself to those people.
171 2012-11-08 13:20:36 <kjj_> does anyone know how named pipes work in Windows?  I think I just about have a pipe notification system working in linux, but that is super easy because it acts exactly like a file
172 2012-11-08 13:21:10 <gmaxwell> then if you forget it you send those people a recovery blob... and do what you said you do to idenfy yourself. And they decrypt it and send it back.
173 2012-11-08 13:21:55 <da2ce7> well I'm off... hope you guys have a great meeting! :)  1am here.
174 2012-11-08 13:22:18 <gmaxwell> ideally they'd be able to pay the escrow parties for this service.. dunno how to do it if their funds are tied up.
175 2012-11-08 13:22:49 <da2ce7> prepay.
176 2012-11-08 13:23:56 <gmaxwell> yuck. I'd rather there be no contact with the escrow parties for security.. and prepay means almost no one will do it.
177 2012-11-08 13:24:42 <gmaxwell> "Sorry, you didn't have the foresight to protect your passphrase" doesn't really help the problem etotheipi_ has seen in practice.
178 2012-11-08 13:28:38 <etotheipi_> gmaxwell: the theory is cool, but I can't imagine how many people would actually use it, even if they had foresight
179 2012-11-08 13:29:05 <etotheipi_> or how many people even has that many technically-inclined friends to make it possible
180 2012-11-08 13:30:26 <gmaxwell> etotheipi_: I was suggesting there be a default list (e.g. well known trusted community members, interested attornies, etc); that users that care could optionally override.
181 2012-11-08 13:30:36 <etotheipi_> ooh, i see
182 2012-11-08 13:30:54 <gmaxwell> so the only thing required of them would be typing in some stuff that they plan to do to prove themselves.
183 2012-11-08 13:32:01 <etotheipi_> part of the problem is that people are so used to the safety net of websites that can help them recover their passwords
184 2012-11-08 13:32:59 <etotheipi_> they don't realize how important *this password* is, above all others
185 2012-11-08 13:33:32 <gmaxwell> Yep. I fully agree. You should have seen the idiotic argument in #bitcoin 24 hours ago I got myself into.
186 2012-11-08 13:34:29 <gmaxwell> someone was suggesting using a really secure wallet passphrase and never writing it down. I recommended writing it down on paper and kept seperate from the computer, potentially split and was accued of being security incompetent.
187 2012-11-08 13:34:52 <etotheipi_> heh
188 2012-11-08 13:35:35 <etotheipi_> I am slightly bothered by the folks who don't understand the massive difference between online and offline security
189 2012-11-08 13:35:52 <gmaxwell> I think a default escrow behavior would be fine... the one real problem is that to be useful you have to provide something to prove yourself to the trustees if they aren't your friends... otherwise a theif could do it too.
190 2012-11-08 13:36:04 <wumpus> writing it down is pretty good idea, as most of the threat in bitcoin is from computer attacks... how many burglars will understand :p
191 2012-11-08 13:36:12 <etotheipi_> wumpus: exactly
192 2012-11-08 13:36:23 <etotheipi_> people assume that if there is any physical evidence anywhere, they will be compromised
193 2012-11-08 13:36:37 <wumpus> unless you're really high profile in the community and people know you have a lot of bitcoins ofc... but that's more the exception than the rule
194 2012-11-08 13:36:43 <etotheipi_> even though 99%+ of issues will arise digitally, not physically
195 2012-11-08 13:37:20 <etotheipi_> I want to write a blog post (if I had a blog), about how you're actually using a brainwallet if you don't create a paper backup or write it down or keep a copy in a safe-deposit box, etc
196 2012-11-08 13:37:33 <gmaxwell> wumpus: I agree. Plus you could hide the written copy pretty effectively??? a theif looking for it would be unlikely to find it.
197 2012-11-08 13:37:55 <gmaxwell> etotheipi_: even when you tell people about that they just dont reason well about their risk of forgetting... You don't remember most of the things you've forgotten at all! :P
198 2012-11-08 13:37:59 <wumpus> right, though it's easy to forget where you put something that you hid effectively but never go looking for :-)
199 2012-11-08 13:38:25 <etotheipi_> wumpus: sure... but you will find it eventually
200 2012-11-08 13:38:40 <etotheipi_> if it's the difference between never seeing your $20k again... you'll find it :)
201 2012-11-08 13:38:45 <wumpus> true
202 2012-11-08 13:39:07 <gmaxwell> well, not for sure, but its pretty likely.
203 2012-11-08 13:39:18 <etotheipi_> although you don't want to end up like the firefighter in "Rescue Me" who hid his $20k in cash somewhere
204 2012-11-08 13:39:29 <etotheipi_> and his wife thought that the container was trash and threw it out
205 2012-11-08 13:39:30 <wumpus> anyway it's still quite more robust and easier to find than storign it digitally somewhere
206 2012-11-08 13:39:31 <gmaxwell> e.g. you hide it in a book and forget it.. then give the book away...
207 2012-11-08 13:40:04 <gmaxwell> Right. It's not for sure but its a good improvement.
208 2012-11-08 13:40:16 <gmaxwell> But people won't do that because they think the odds of needing it are ~0.
209 2012-11-08 13:40:22 <wumpus> yes
210 2012-11-08 13:40:38 <etotheipi_> gmaxwell: yet they believe the odds are 90%+ that if it exists, an attacker will find it
211 2012-11-08 13:40:47 <wumpus> reminds me that the bitcoin gui should really have a way to print the key
212 2012-11-08 13:40:59 <etotheipi_> (plug Armory, here)
213 2012-11-08 13:41:06 <etotheipi_> :)
214 2012-11-08 13:41:15 <gmaxwell> Yea, armory has done good there.
215 2012-11-08 13:41:26 <wumpus> though last time it ended up in some paranoid discussion where people were afraid that their printer stored it... :-)
216 2012-11-08 13:41:28 <gmaxwell> etotheipi_: is the paper backup unencrypted?
217 2012-11-08 13:41:34 <etotheipi_> gmaxwell, yes
218 2012-11-08 13:41:34 <gmaxwell> wumpus: well??? it's a risk!
219 2012-11-08 13:41:46 <gmaxwell> wumpus: my printers can print the last job from memory.
220 2012-11-08 13:41:52 <etotheipi_> gmaxwell: I'm conflicted on allowing encrypted paper backups
221 2012-11-08 13:42:00 <etotheipi_> due to my previous comment about brainwallets
222 2012-11-08 13:42:18 <gmaxwell> etotheipi_: agreed. I'd rather it support splitting instead of encryption. e.g. print three pages, and you need two of them.
223 2012-11-08 13:42:43 <etotheipi_> some would argue that I shouldn't be protecting users from themselves (let them encrypt and forget it if they are that irresponsible), but I just don't want to deal with it when they do
224 2012-11-08 13:42:45 <wumpus> so writing it down the old fashioned way is the only way that is secure really
225 2012-11-08 13:43:09 <wumpus> also less recognizable
226 2012-11-08 13:43:10 <etotheipi_> gmaxwell: doing a split-paper backup is an excellent idea
227 2012-11-08 13:44:22 <gmaxwell> I dont want to protect people from _personal_ stupidity, but I strongly believe software should protect people from inherit _human_ stupidity.
228 2012-11-08 13:44:42 <etotheipi_> gah... I should go to work...
229 2012-11-08 13:44:50 <etotheipi_> I'd love to sit here and chat about this stuff all day...
230 2012-11-08 13:44:54 <gmaxwell> (well, personal stupidty I have sympathy for being often stupid myself; but the universe will just invent better idiots)
231 2012-11-08 13:44:59 <wumpus> I'm not that sure about splitting, it sounds nice in theory, but if you hide something in multiple places so much can go wrong
232 2012-11-08 13:45:20 <etotheipi_> wumpus: I like the idea as having it as an option
233 2012-11-08 13:45:28 <gmaxwell> wumpus: then put them in one place. I don't think that humans are too likely to underestimate that risk.
234 2012-11-08 13:45:32 <etotheipi_> most users will print a single backup as they do already with Armory
235 2012-11-08 13:45:59 <wumpus> oh most people are very optimistic about their own memory
236 2012-11-08 13:46:01 <etotheipi_> but those that understand the risk (and are in "Expert usermode" in Armory), can have the option to use split-paper
237 2012-11-08 13:46:20 <wumpus> it's so easy to remember lots of hiding places... for a day :)
238 2012-11-08 13:46:47 <etotheipi_> well, if they do 3-of-5 or something... they don't have to find all of them...
239 2012-11-08 13:47:07 <wumpus> hehe
240 2012-11-08 13:47:09 <etotheipi_> although the more you create and the lower threshold you have ,the more likely an attack of opportunity could arise from someone finding pieces
241 2012-11-08 13:47:46 <etotheipi_> but I still think 99%+ of the risk is online, and thus a single paper backup (1-of-1) meets the needs of the vast majority of users
242 2012-11-08 13:47:57 <gmaxwell> wumpus: right but at least you have the 'you can just tear the house apart'   it took me two days of solid searching to find my 2002 tax returns. ... they were in a folder... at the bottom of a box of clothes from two moves ago.
243 2012-11-08 13:48:19 <wumpus> but yea the splitting is nice for experts or people that have a lot of coins, sure
244 2012-11-08 13:48:47 <etotheipi_> I've enjoyed having a split interface in Armory and separating "Expert" features from regular users
245 2012-11-08 13:49:01 <etotheipi_> because I keep coming up with ideas that *I* want implemented, but would just confuse beginners
246 2012-11-08 13:49:15 <wumpus> fun would be to put 5 bitcoins in every private key, print them out separately and hide them all over the place
247 2012-11-08 13:50:25 <etotheipi_> alright unlike everyone else here, *I* have to go to work...
248 2012-11-08 13:50:28 <etotheipi_> :-/
249 2012-11-08 13:50:43 <etotheipi_> (which, unfortunately, has nothing to do with Bitcoin)
250 2012-11-08 14:29:44 <kjj_> gavinandresen: got a sec?
251 2012-11-08 14:29:50 <gavinandresen> nope
252 2012-11-08 14:30:03 <drizztbsd> lol
253 2012-11-08 14:30:45 <gavinandresen> (I'm in a Foundation board meeting in another window....)
254 2012-11-08 14:31:01 <kjj_> no problem.  that's why I asked
255 2012-11-08 15:23:52 <gavinandresen> kjj_: ok, now I have a sec
256 2012-11-08 15:36:08 <sipa> gavinandresen: meeting in 1.5h, right?
257 2012-11-08 15:36:16 <gavinandresen> Right!
258 2012-11-08 15:36:22 <gavinandresen> Yay meetings!
259 2012-11-08 15:36:46 <gavinandresen> GIVE ME AN "M"!  GIVE ME AN "E"!  GIVE ME (ok, I'll shut up now)
260 2012-11-08 15:37:06 <sipa> /me tings gavinandresen needs some valium
261 2012-11-08 15:37:23 <kjj_> gavinandresen: I was looking at the named pipe thing
262 2012-11-08 15:37:52 <sipa> i think named pipes are very hard to do in a cross-platform way
263 2012-11-08 15:38:03 <gavinandresen> sipa: I'm ok with it not working on Windows.
264 2012-11-08 15:38:04 <kjj_> that was the conclusion that I was coming to
265 2012-11-08 15:38:12 <gavinandresen> named pipes work nicely on Mac and LInux
266 2012-11-08 15:38:38 <sipa> obviously we need a built-in scripting language to deal with these events
267 2012-11-08 15:38:47 <kjj_> on unix systems, it appears to be very, very simple to make a stripped down copy of the OutputDebugStringF function
268 2012-11-08 15:38:52 <sipa> how about x86?
269 2012-11-08 15:39:16 <gavinandresen> I'm lost.  how about what x86?
270 2012-11-08 15:39:38 <sipa> using x86 as a scripting language for handling events
271 2012-11-08 15:39:44 <sipa> i hear it is very flexible
272 2012-11-08 15:39:52 <gavinandresen> kjj: https://groups.google.com/forum/?fromgroups=#!topic/boost-list/rOZy1xxS-rw  looks relevant
273 2012-11-08 15:39:56 <kjj_> if it is up to the user to create it in advance with mkfifo, then the function just needs to fopen it on the first pass, and write the rest
274 2012-11-08 15:40:51 <gavinandresen> kjj_: yes, assuming the pipe never backs up and blocks.  Which might be a good assumption (or can you fwrite and have it fail if it blocks?)
275 2012-11-08 15:40:52 <kjj_> it appears that named pipes on windows appear in a different namespace than regular files
276 2012-11-08 15:40:58 <sipa> well you could just say "upon event, write to fd N", and you can connect thay fd yourself to a pipe, a file, a network socket, ...
277 2012-11-08 15:41:28 <gavinandresen> sipa: still need code to open that fd
278 2012-11-08 15:41:41 <sipa> gavinandresen: bash; done
279 2012-11-08 15:42:05 <gavinandresen> sipa: ? huh ?  can you start a process with a fd open already for it?
280 2012-11-08 15:42:12 <gavinandresen> (other than stdout/stderr)
281 2012-11-08 15:42:16 <sipa> sure
282 2012-11-08 15:42:24 <sipa> blabla 3>bla
283 2012-11-08 15:42:37 <sipa> will have fd 3 opened for writing to bla
284 2012-11-08 15:42:40 <gavinandresen> I did not know that.
285 2012-11-08 15:42:56 <sipa> but you probably want it built in in the program
286 2012-11-08 15:43:05 <sipa> to reopen a connectiin if it dies, eg
287 2012-11-08 15:43:22 <gavinandresen> mmm.  and will the fd survive the -daemon fork?  (suppose it would...)
288 2012-11-08 15:43:41 <sipa> yechild threads inherit the fd table by default
289 2012-11-08 15:43:56 <sipa> *child processes
290 2012-11-08 15:44:46 <gavinandresen> sipa: re: re-opening connections, etc:  I'd like all of that code to be external to the bitcoind process, which is why just writing to a fd or named pipe appeals to me
291 2012-11-08 15:45:00 <kjj_> er, I *think* that the write will fail if the buffer is full
292 2012-11-08 15:46:11 <gavinandresen> no, I think fwrite will block if the output buffer is full
293 2012-11-08 15:46:53 <sipa> by default, it will block
294 2012-11-08 15:47:00 <sipa> but you can make the fd non-blocking
295 2012-11-08 15:47:17 <sipa> in which case you can use select or poll to query for place in the write buffer
296 2012-11-08 15:48:43 <gavinandresen> meh... if it blocks, maybe just a debug.log ERROR and stop reporting new transactions until it unblocks, and rely on users to write transaction handling code that is fast enough?
297 2012-11-08 15:49:33 <kjj_> hmm.  that means a check before the write to figure out if it will block, right?
298 2012-11-08 15:50:15 <kjj_> been a while...
299 2012-11-08 15:50:25 <gavinandresen> I dunno, I've never actually written any non-blocking IO stuff in C/C++
300 2012-11-08 15:51:19 <kjj_> hmm.  I think that if I just fcntl to make the FD non-blocking, it'll just error through silently.
301 2012-11-08 15:52:09 <kinlo> basicly you write to a nonblocking file as is
302 2012-11-08 15:52:41 <kinlo> and you look at the return value, it will write less bytes if the fd would block instead, or return E_TRYAGAIN or something  like that
303 2012-11-08 15:53:20 <kinlo> it's been a while, but basicly you just need to remember what was successfull and what not, and retry writing to it as soon as select/poll/epoll/whatever tells you it is ready to be written to
304 2012-11-08 15:54:02 <sipa> or you have a separate thread to write to the pipe, and use an internal buffer with a separate critical section
305 2012-11-08 15:54:15 <gavinandresen> sipa: do you have a strong preference for:  bitcoind -txnotify=/path/to/fifo   versus   bitcoind -txnotify=3 3> /path/to/fifo ? I don't really care, although I think the most common case will be a fifo instead of writing directly to a socket....
306 2012-11-08 15:55:17 <sipa> gavinandresen: you can always use -txnotify=/proc/self/fd/3 3>/path/to/fifo :)
307 2012-11-08 15:55:24 <sipa> with the first syntax
308 2012-11-08 15:55:37 <kinlo> I prefer filenames too
309 2012-11-08 15:55:40 <sipa> at least on linux
310 2012-11-08 15:55:41 <kjj_> gavinandresen: I was actually thinking just -notifypipe=BOOL
311 2012-11-08 15:56:03 <kjj_> if true, it opens a named pipe in the data dir, where everything else already is
312 2012-11-08 15:56:16 <gavinandresen> sipa: nifty, I'm learning all sorts of tricks today
313 2012-11-08 15:56:39 <gavinandresen> kjj_: NACK, path to named pipe should be given as a command-line arg
314 2012-11-08 15:57:13 <kjj_> heh.  that was my first plan, but when I started looking, I noticed how easy it would be to just make it boolean.  :)
315 2012-11-08 16:02:31 <gavinandresen> kinlo: okey doke.  I was assuming a stdio fwrite() so you get buffering for free.
316 2012-11-08 16:02:33 <kinlo> gavinandresen: and it's common to do it that way, the application might be able to prioritize other tasks or stop generating data to be sent or something...
317 2012-11-08 16:03:46 <kinlo> btw, there is a clear distinction you can make here: a failed write versus a write that has not been done because it would block
318 2012-11-08 16:03:53 <kinlo> you will need to handle failed writes as normal
319 2012-11-08 16:04:12 <kinlo> you do get different error messages
320 2012-11-08 16:04:52 <gavinandresen> I'm saying in this case "failed" and "would block" should be treated the same: write error to debug.log and go on about your business.
321 2012-11-08 16:05:41 <sipa> gavinandresen: pipes have a kernel-level 4 KiB buffer too, iirc
322 2012-11-08 16:06:18 <kinlo> look at irc: regular irc server has a buffer of let's say 16Kb.  if you as a client do not process your data fast enough so you exceed that buffer of 16kb, you are disconnected with the error message Excess Flood
323 2012-11-08 16:06:25 <kinlo> it makes sense to do it that way
324 2012-11-08 16:07:57 <kjj_> is "a" the proper mode for a pipe instead of "w" ?
325 2012-11-08 16:13:10 <kjj_> lol.  seems that even fopen() will block on a pipe until the other end is ready
326 2012-11-08 16:14:01 <kinlo> possibly, that's probably why all network code uses the lower level calls to bind/connect/open instead
327 2012-11-08 16:14:52 <kjj_> that seems to suggest that we need a new thread, and a wrapper function that can exit if the fd isn't open yet
328 2012-11-08 16:15:28 <kinlo> or just use the lower level cases instead of those that start with f
329 2012-11-08 16:18:34 <kjj_> heh.  with open() and O_WRONLY, it can either succeed if the reader is already connected, block if you don't specify O_NONBLOCK, or fail.  I guess that can work.
330 2012-11-08 16:19:13 <kjj_> so, we can just fall through on failure
331 2012-11-08 16:25:41 <kjj_> hmm.  open() and poll() actually ended up being pretty easy.  now to see if it compiles and works...
332 2012-11-08 16:29:11 <kjj_> by the way, I'm using a single pipe for this, with different notification messages depending on the context.  As in "Block <hash>" or "Wallet transaction <hash>"
333 2012-11-08 16:30:22 <kjj_> if we imagine future expansion of the sorts of things people might want to use it for, maybe JSON objects would be a better idea
334 2012-11-08 16:52:43 <midnightmagic> wow. might have just lost a hot mining wallet worth 167 bitcoins.
335 2012-11-08 16:52:57 <midnightmagic> machine rebooted, come up "unknown filesystem" "grub rescue>"
336 2012-11-08 16:56:03 <midnightmagic> i guess that means the hdd at least isn't failing to respond.
337 2012-11-08 16:56:49 <gavinandresen> bitcoind/Bitcoin-Qt needs a better default wallet backup mechanism....
338 2012-11-08 16:57:49 <midnightmagic> i have been pretty diligent about wallet backups, but in this particular case I'd recently wanted to flush out all non-compressed keys in the wallet and so I ran through the whole old set of them, but for that particular wallet I'd neglected backing it up subsequent to that. :(
339 2012-11-08 16:58:55 <midnightmagic> and of course i have my stupid p2pool mod which runs through a randomized set of fresh addresses rather than mining just to one address.
340 2012-11-08 16:59:21 <gavinandresen> sounds like something I'd do.  we need a better default, idiot-proof (because we're all idiots sometimes) backup mechanism....
341 2012-11-08 16:59:32 <midnightmagic> thank you for not mocking me. :)
342 2012-11-08 16:59:36 <gmaxwell> midnightmagic: it's probably recoverable. (not that it matters for your backup comments)
343 2012-11-08 17:00:03 <sipa> *DING*
344 2012-11-08 17:00:35 <gavinandresen> Official Meeting Time ?
345 2012-11-08 17:00:39 <midnightmagic> yeah, i haven't given up yet. I have a nice little hard drive recovery station i've set up for friends and I suspect i'll be able to recover it. just.. not remotely.
346 2012-11-08 17:00:50 <gavinandresen> Somebody read the minutes of the last meeting....  (KIDDING)
347 2012-11-08 17:00:52 <midnightmagic> 
ACTION pipes down.

348 2012-11-08 17:01:28 <sipa> 
ACTION pokes aroun

349 2012-11-08 17:01:55 <gmaxwell> Minutes of the last meeting:
350 2012-11-08 17:02:52 <gavinandresen> So... BIP process.  I think we need somebody to take over the BIP editor role from genjix
351 2012-11-08 17:03:08 <sipa> agree
352 2012-11-08 17:03:27 <gavinandresen> Self-assigning BIP numbers is fine with me, as long as there is somebody to weed out REALLY badly thought out submissions
353 2012-11-08 17:04:02 <gmaxwell> I'm willing. I assume all it will involve is advising number conflicts and deflecting less obviously fine stuff to the lists for discussion.
354 2012-11-08 17:04:28 <gavinandresen> Great, I think you'd make a great BIP editor, gmaxwell
355 2012-11-08 17:04:57 <sipa> yes - no need to do heavy editing like judging whether something is good enough or not - just weed out obviously wrong/meaningless things
356 2012-11-08 17:05:07 <gmaxwell> (Also, I'll only have about half attention for this meeting today: I'm currently at the IETF meeting)
357 2012-11-08 17:05:37 <gavinandresen> Do we need a github.com/BIPS repo as the canonical place for BIPS, or is the wiki good enough?
358 2012-11-08 17:05:42 <gmaxwell> Right. More advisory than actually editorial. Let the list convince people their ideas are crazy or approve the crazy stuff for BIPs.
359 2012-11-08 17:06:15 <sipa> there used to be genjix/bips.git on github, but it's outdated and i don't think anyone considers it to be the canonical repository anymore
360 2012-11-08 17:06:22 <gavinandresen> err, github.com/bitcoin/BIPS ....
361 2012-11-08 17:06:24 <Luke-Jr> Might make sense to be more strict that stuff get posted to the list, at least as a summary, before getting a BIP #
362 2012-11-08 17:07:27 <sipa> people will browse the wiki and not a github repository
363 2012-11-08 17:07:28 <Luke-Jr> how practical is it to back up MediaWiki data?
364 2012-11-08 17:07:40 <sipa> but a wiki has the disadvantage that anyone can just create pages
365 2012-11-08 17:07:42 <gmaxwell> Luke-Jr: you can easily extract the history for single pages as xml.
366 2012-11-08 17:08:22 <midnightmagic> Who runs the host for the wiki and is it sure enough to be trusted longer-term with the BIP/development discussion data?
367 2012-11-08 17:08:24 <Luke-Jr> sipa: that is also an advantage
368 2012-11-08 17:08:32 <Luke-Jr> midnightmagic: MagicalTux
369 2012-11-08 17:08:38 <gmaxwell> sipa: that could be changed... e.g. if BIPs were moved to a BIP namespace then it could be set so only some users could create pages in that namespace.. but I dunno how much twiddling anyone wants to do with the wiki.
370 2012-11-08 17:08:42 <Luke-Jr> MagicalTux: can you set gmaxwell mod for BIP purposes?
371 2012-11-08 17:08:48 <gmaxwell> midnightmagic: make backups.
372 2012-11-08 17:09:08 <sipa> the discussion should be on the mailing anyway, imho
373 2012-11-08 17:09:37 <gavinandresen> aside:  moving the wiki from under the wing of Mt.Gox is something I want to talk about sometime in the next couple of months.
374 2012-11-08 17:09:44 <Luke-Jr> sipa: perhaps come to a quick consensus on IRC then continue on ML? ie, get the "long" part over with in realtime but leave room for further discussion if necessary
375 2012-11-08 17:10:11 <gavinandresen> sipa:  Yes, I think consensus here then proposing "here's what we think should happen" to the ML is the right thing to do.
376 2012-11-08 17:10:15 <gmaxwell> midnightmagic: ideally it would have public full site dumps like wikipedia has. (http://dumps.wikimedia.org/enwiki/20121001/)
377 2012-11-08 17:10:30 <sipa> Luke-Jr: sure
378 2012-11-08 17:11:02 <midnightmagic> gmaxwell: I agree..
379 2012-11-08 17:11:21 <gavinandresen> Ok, so to summarize: we're proposing that gmaxwell take over the light gatekeeper duties from genjix, wiki becomes canonical place for BIPS (locked down if that becomes necessary, definitely backed up)
380 2012-11-08 17:11:45 <midnightmagic> gavinandresen: Did you have anything in mind of where might be a good replacement site?
381 2012-11-08 17:12:04 <gmaxwell> We can also have edits to bips show up in here or in the commits channel.
382 2012-11-08 17:12:11 <ThomasV_> any news from genjix?
383 2012-11-08 17:12:28 <gavinandresen> midnightmagic: Foundation is giving us (core dev team) a grant for servers, the discussion will be what to put on them....
384 2012-11-08 17:12:36 <midnightmagic> very nice.
385 2012-11-08 17:12:41 <BlueMatt> aside: why does it seem that the -commits channel has been dead for some time?
386 2012-11-08 17:12:50 <sipa> haven't heard from genjix since the conference in london
387 2012-11-08 17:13:05 <gavinandresen> BlueMatt: I was told that the bot died....
388 2012-11-08 17:13:10 <midnightmagic> BlueMatt: Perhaps that was the broken CIA bot death?
389 2012-11-08 17:13:17 <gavinandresen> yeah, that's the one....
390 2012-11-08 17:13:21 <BlueMatt> hmm...ok Ill look into it later
391 2012-11-08 17:13:24 <Luke-Jr> BlueMatt: CIA is dead dead
392 2012-11-08 17:13:35 <BlueMatt> ah, ok, well then Ill fix the rss thinggy
393 2012-11-08 17:13:42 <BlueMatt> nanotube: is the rss thinggy through gribble still up?
394 2012-11-08 17:13:42 <Luke-Jr> cool
395 2012-11-08 17:14:06 <midnightmagic> lol, yea like so dead some guy (esr I think) had to reverse it again just to figure out how it worked.
396 2012-11-08 17:14:44 <gavinandresen> So: 0.8 release: consensus seems to be NOT to do a full feature freeze yet, but get bloom filters in and have no more major changes.
397 2012-11-08 17:15:08 <BlueMatt> re: bloom filters: merged sipa's stuff, working on adding more unit tests
398 2012-11-08 17:15:15 <Luke-Jr> gavinandresen: I thought it'd be nice to see the p2p networking rewrite in, but that's major and I guess could just as well be in 0.9 at the rate it's going
399 2012-11-08 17:15:42 <gavinandresen> who is doing a p2p networking rewrite? I don't think I've heard about that work....
400 2012-11-08 17:15:59 <Luke-Jr> I don't know that anyone is doing it yet, but we desperately need it.
401 2012-11-08 17:16:17 <Luke-Jr> the current code is completely synchronous
402 2012-11-08 17:16:19 <sipa> you can't plan a feature without anyone writing it
403 2012-11-08 17:17:10 <gavinandresen> yup... my priority list would still look like "scaling and wallet security/backup" and... well, not much else
404 2012-11-08 17:17:24 <Luke-Jr> scaling involves the rewrite ;)
405 2012-11-08 17:17:39 <sipa> Luke-Jr: i disagree
406 2012-11-08 17:18:03 <sipa> surely a rewrite would help, but even just moving processing of blocks to separate threads would offload the message handling thread massively
407 2012-11-08 17:18:04 <gavinandresen> me too, I haven't noticed any issues dealing with transaction load once I'm caught up to the chain
408 2012-11-08 17:18:09 <gmaxwell> There are a number of somewhat low hanging things visible now that ultraprune has exposed.
409 2012-11-08 17:18:23 <Luke-Jr> sipa: while the client downloads/uploads a block to/from a single peer, all other peers are completely ignored
410 2012-11-08 17:18:28 <sipa> i know
411 2012-11-08 17:18:32 <gmaxwell> E.g. block download (not just initial but for hosts that have been offline a few days) matter much more now.
412 2012-11-08 17:18:51 <sipa> i don't disagree that it would help, but it's far from the most pressing issue, imho
413 2012-11-08 17:19:30 <gmaxwell> Luke-Jr: this improvement is a competative advantage for miners. Lets try to expand the number of people working on bitcoind by getting someone else to hack on it.
414 2012-11-08 17:20:05 <Luke-Jr> gmaxwell: not really, as it matters on the relay nodes, not the mining nodes
415 2012-11-08 17:20:31 <sipa> my TODO list is now finishing ultraprune's leftovers, and then deterministic wallets i guess
416 2012-11-08 17:20:34 <gmaxwell> Luke-Jr: In any case. Need code first. The goal there is well defined.
417 2012-11-08 17:20:46 <Luke-Jr> yep, like I said, might as well wait for 0.9
418 2012-11-08 17:21:53 <gmaxwell> Luke-Jr: if you want a stopgap, writing a very fast block relayer that just does spv validation based on picocoin may be a useful short term thing.
419 2012-11-08 17:22:10 <BlueMatt> bitcoin backbone!
420 2012-11-08 17:22:23 <gavinandresen> So, looking at 0.8....   what can we do to encourage testing?  Obviously getting auto-upgrade working.  Nightly builds would be great.
421 2012-11-08 17:22:52 <BlueMatt> as always, nightlies are at: http://jenkins.bluematt.me/job/Bitcoin/ws/
422 2012-11-08 17:23:14 <gavinandresen> BlueMatt: awesome, when we're ready we need to publicize those more.
423 2012-11-08 17:23:25 <gmaxwell> gavinandresen: A point which is more development testing than user testing centric:  at the moment I'm thinking of putting up a ultraprune centric testing checklist on the wiki. I assume if I start such a think someone will dump more stuff into it?
424 2012-11-08 17:23:40 <nanotube> BlueMatt: i havent done anything with it, so hypothetically should still be up.... let's check
425 2012-11-08 17:23:54 <sipa> i thought about announcing "demo builds" on the forums, but want auto-upgrade working first
426 2012-11-08 17:24:13 <BlueMatt> nanotube: na, cant be...either its broken on the url or smth or its down in gribble (it should be displaying pull reqs in -commits)
427 2012-11-08 17:24:13 <gavinandresen> gmaxwell: yes, I'll dump more stuff into it.
428 2012-11-08 17:24:25 <gmaxwell> sipa: instead do builds that stop working after X weeks?
429 2012-11-08 17:24:50 <nanotube> BlueMatt: ah it seems the rss check thread is hung.
430 2012-11-08 17:25:14 <nanotube> BlueMatt: guess it'll fix itself next time the bot is restarted... not sure i wanna kill the bot just now. :)
431 2012-11-08 17:25:27 <gmaxwell> I had some problems on testnet where broken ultraprune nodes overpowered my nodes that inserted an older-ultraprune killing block. (I forked and overpowered them and got it back into the chain)
432 2012-11-08 17:25:41 <sipa> gavinandresen: any idea when foundation hardware would be available?
433 2012-11-08 17:25:47 <gavinandresen> I like the idea of nagging users to upgrade if they're running an ancient build.  I actually kind of like that idea for releases, too (nag after 1 year or something....)
434 2012-11-08 17:25:54 <BlueMatt> nanotube: heh, ok then
435 2012-11-08 17:26:37 <BlueMatt> gavinandresen: there is an outstanding auto-update pull...
436 2012-11-08 17:26:53 <gavinandresen> sipa: could happen very quickly, Foundation has the bitcoins.  I did some research into bitcoin-accepting service providers....
437 2012-11-08 17:28:48 <gmaxwell> I'd really like to get the coverage reports I've done from time to time running on a CI server. (It's easy to run them from Jenkins, I do it on other stuff; I just haven't wanted to add more work to the already slow pull tester stuff)
438 2012-11-08 17:29:45 <gavinandresen> Somebody want to volunteer to propose exactly what server to get, where ?
439 2012-11-08 17:30:05 <gavinandresen> lemme go find the budget numbers....
440 2012-11-08 17:30:09 <phantomcircuit> gavinandresen, high security server or something else
441 2012-11-08 17:30:12 <BlueMatt> gmaxwell: if you have the time, putting it in a pull for the scripts that jenkins/pull-tester runs would be much appreciated https://github.com/TheBlueMatt/test-scripts
442 2012-11-08 17:30:30 <phantomcircuit> for hosting bitcoin.org ?
443 2012-11-08 17:30:48 <sipa> CI server?
444 2012-11-08 17:30:51 <gavinandresen> phantomcircuit: server will be for the jenkins pull-testing and nightly builds, probably will be a new DNS seed
445 2012-11-08 17:30:55 <Luke-Jr> gavinandresen: What did your research come up with? I'd probably suggest we just go with the most reputable/organized provider for low-risk stuff.
446 2012-11-08 17:30:56 <BlueMatt> phantomcircuit: for various things (incl jenkins/pull-tester, among many things)
447 2012-11-08 17:31:05 <Luke-Jr> gavinandresen: or maybe diversify if there's enough options
448 2012-11-08 17:31:14 <BlueMatt> gavinandresen: seems to me getting a dedicated vm server self-managed would be the best security/cost option
449 2012-11-08 17:31:21 <gmaxwell> BlueMatt: it will further slow down the execution (factor of 2 if your IO is fast?), and creates a ton of temporary data on disk. If you're okay with that, I'll send a patch.
450 2012-11-08 17:31:48 <BlueMatt> gmaxwell: well the idea was to get the patch written and not merge it yet until it gets better hardware ;)
451 2012-11-08 17:31:51 <gavinandresen> My research into bitcoin-accepting dedicated server companies:  https://docs.google.com/spreadsheet/ccc?key=0AmMi4Whh8u5HdGxmT0ZoUXZSSDExLWZWRzFMdmt5U1E
452 2012-11-08 17:32:00 <gmaxwell> BlueMatt: ah okay. sure.
453 2012-11-08 17:32:10 <Luke-Jr> gavinandresen: that being said, we're almost certain to need a locked-down cage at some point, so it might make sense to plan for that in advance even if we don't get it now (ie, pick somewhere that has them available)
454 2012-11-08 17:32:26 <phantomcircuit> gavinandresen, what's the GLBSE note on bitvps?
455 2012-11-08 17:32:50 <gavinandresen> phantomcircuit: I think they're a GLBSE-listed company, which is a negative in my opinion.
456 2012-11-08 17:32:57 <phantomcircuit> i've heard good things about cinfu and snelservers
457 2012-11-08 17:33:03 <gavinandresen> (well, were, not are....)
458 2012-11-08 17:33:16 <Luke-Jr> gavinandresen: FWIW, you can get a discount at BitVPS via Eligius; maybe could get even a better deal by talking to rg/grubles directly
459 2012-11-08 17:33:22 <phantomcircuit> gavinandresen, im pretty sure that isn't a major part of their business
460 2012-11-08 17:33:50 <Luke-Jr> gavinandresen: 12 GB RAM doesn't sound like much these days, especially if the plan is to run VMs on it
461 2012-11-08 17:34:14 <BlueMatt> (CI building bitcoin eats tons of ram...)
462 2012-11-08 17:34:32 <sipa> CI?
463 2012-11-08 17:34:40 <gmaxwell> continuous integration
464 2012-11-08 17:34:43 <gmaxwell> What jenkins does.
465 2012-11-08 17:34:43 <sipa> oh
466 2012-11-08 17:34:53 <sipa> sure
467 2012-11-08 17:35:01 <BlueMatt> s/CI//
468 2012-11-08 17:35:08 <phantomcircuit> Luke-Jr, im betting rg can find a dedicated box with > 12GB of ram if someone is willing to pay for it
469 2012-11-08 17:35:18 <gavinandresen> renting more RAM is certainly possible.  Getting a discount would be spiffy....
470 2012-11-08 17:35:44 <gmaxwell> BlueMatt: well we need peak-compile-usage * cores ram.
471 2012-11-08 17:35:58 <gmaxwell> Since I can build it on my 2gb laptop, we shouldn't need more than 2gb*cores.
472 2012-11-08 17:36:01 <phantomcircuit> although i think you'll find that if you're running mostly identical guests ksm works fairly well
473 2012-11-08 17:36:06 <gavinandresen> Do we care if the server is US or someplace else?  Diversifying to another country appealed to me
474 2012-11-08 17:36:13 <BlueMatt> gmaxwell: * 2 (jenkins/pull-tester often run in parallel)
475 2012-11-08 17:36:13 <Diablo-D3> MOAR RAMS
476 2012-11-08 17:36:40 <gmaxwell> BlueMatt: I mean I can run a 1 core build on my laptop, and we just shouldn't have more executors than cores.
477 2012-11-08 17:37:22 <BlueMatt> gmaxwell: well, depends on if Im lazy, if its low on ram, sure I can limit it...or just be lazy and let them both eat full core count and let them keep running parallel
478 2012-11-08 17:37:24 <Diablo-D3> heh, get this
479 2012-11-08 17:37:31 <Diablo-D3> you know the G34 socket?
480 2012-11-08 17:37:54 <Diablo-D3> up to 4 sockets on a board, 128gb per board for 512 total
481 2012-11-08 17:38:17 <Diablo-D3> enough rams? :D
482 2012-11-08 17:38:52 <sipa> http://www.dfi.ch/en/servers-hosting/dedicated-server.html  <-- switzerland?
483 2012-11-08 17:39:02 <phantomcircuit> gavinandresen, the primary issue with all the dedicated server providers that accept bitcoins is that their ram pricing is well above market
484 2012-11-08 17:39:11 <BlueMatt> sipa: I believe the goal was to pay with bitcoin
485 2012-11-08 17:39:35 <sipa> ah, of course
486 2012-11-08 17:39:36 <gmaxwell> in any case, more is better. I'm just saying that 12 for 6 core isn't terrible.  We could quite easily keep 16 cores busy (by adding network and disk fuzzing test cases)... but jenkins remotes wore excellently.
487 2012-11-08 17:39:55 <luke-jr_> gmaxwell: right, but BlueMatt was talking about doing this in 1 VM on a host with more than 1
488 2012-11-08 17:40:19 <phantomcircuit> luke-jr_, iirc the build vm is temporary
489 2012-11-08 17:40:21 <BlueMatt> luke-jr_: in theory the others would consist of not much more than a dnsseed and maybe some hosting, so...maybe 1GB ram max
490 2012-11-08 17:40:26 <BlueMatt> phantomcircuit: no
491 2012-11-08 17:40:37 <phantomcircuit> no?
492 2012-11-08 17:40:40 <Diablo-D3> gmax\\oh
493 2012-11-08 17:40:41 <Diablo-D3> oh
494 2012-11-08 17:40:44 <Diablo-D3> and that G34 socket?
495 2012-11-08 17:40:52 <Diablo-D3> has those 16 core piledrivers
496 2012-11-08 17:40:54 <phantomcircuit> i thought jenkins build a temporary vm for each build so you'd get consistent results
497 2012-11-08 17:40:55 <Diablo-D3> so 64 cores :3
498 2012-11-08 17:41:01 <sipa> phantomcircuit: that's gitian
499 2012-11-08 17:41:06 <phantomcircuit> ah
500 2012-11-08 17:41:07 <phantomcircuit> right
501 2012-11-08 17:41:19 <phantomcircuit> 
ACTION goes back to fiddling with stuff

502 2012-11-08 17:41:20 <Luke-Jr> jenkins should probably use gitian at some point, but meh
503 2012-11-08 17:41:24 <gavinandresen> Foundation grant is 500 BTC which we should plan on lasting us for the next year, and needs to pay for servers, SSL certs, and code-signing cert.
504 2012-11-08 17:41:49 <BlueMatt> Luke-Jr: yes, it should...wanna implement lxc for it? https://github.com/TheBlueMatt/test-scripts
505 2012-11-08 17:42:02 <gavinandresen> Rough budget is: https://docs.google.com/a/bitcoinfoundation.org/spreadsheet/ccc?key=0AmMi4Whh8u5HdEpTejNQbm9rR0Z2clNtTVMtSGJ1bHc#gid=0
506 2012-11-08 17:42:18 <sipa> if not, just separately running gitian for dailies (not necessarily pulltests) would be nice
507 2012-11-08 17:42:34 <sipa> as gitian is probably more resource intensive than just building?
508 2012-11-08 17:42:52 <BlueMatt> no, but it needs kvm support
509 2012-11-08 17:42:57 <BlueMatt> (well, used to, now its possible)
510 2012-11-08 17:43:04 <BlueMatt> I just havent looked at the lxc stuff
511 2012-11-08 17:44:05 <BlueMatt> re: unmetered 1Gbps...I can leave servers on in the cs building with that, if necessary...
512 2012-11-08 17:44:28 <gmaxwell> BlueMatt: in your list of things on it.. we should have a dnsseed and a git mirror. (even if we don't bother to make it public)
513 2012-11-08 17:44:43 <BlueMatt> gmaxwell: agreed
514 2012-11-08 17:44:47 <phantomcircuit> gavinandresen, startssl would be cheaper for the certificate
515 2012-11-08 17:45:06 <phantomcircuit> i think
516 2012-11-08 17:45:12 <BlueMatt> gmaxwell: but...I cant see any of those + web hosting being > 1GB in total
517 2012-11-08 17:45:20 <gmaxwell> We just switched Xiph.Org to a startssl wildcard and it works well. But I assumed gavin's choice was influenced by bitcoin payments.
518 2012-11-08 17:45:27 <phantomcircuit> at the very least it would be cheaper if you ever needed another certificate
519 2012-11-08 17:45:42 <BlueMatt> do we own btc.org?
520 2012-11-08 17:45:50 <phantomcircuit> you can buy a rapidssl cert using bitcoins?
521 2012-11-08 17:45:56 <gavinandresen> BlueMatt: yes, Bitcoin Foundation owns btc.org
522 2012-11-08 17:46:00 <BlueMatt> ok
523 2012-11-08 17:46:26 <gavinandresen> certificates I assume we'll have to convert to dollars to buy
524 2012-11-08 17:46:30 <phantomcircuit> only problem with startssl is they're pretty much constantly under ddos
525 2012-11-08 17:46:44 <drizztbsd> Registrant Email:roger@memorydealers.com
526 2012-11-08 17:46:46 <drizztbsd> uh?
527 2012-11-08 17:46:46 <phantomcircuit> in that case startssl is definitely the better choice
528 2012-11-08 17:47:06 <gavinandresen> yes, roger bought it for the Foundation
529 2012-11-08 17:47:22 <phantomcircuit> 
ACTION grumbles

530 2012-11-08 17:47:34 <midnightmagic> <3 new version of bitcoind's speed of block download/sync
531 2012-11-08 17:47:40 <drizztbsd> Usually I use cacert, but Windows does not accept it by default ;(
532 2012-11-08 17:47:55 <drizztbsd> neither Android
533 2012-11-08 17:47:57 <BlueMatt> drizztbsd: nothing accepts it by default...
534 2012-11-08 17:48:02 <phantomcircuit> gavinandresen, 120 USD @ startssl will get a personal -> organization validation which would allow for unlimited issuance of certificates in the name of the bitcoin foundation
535 2012-11-08 17:48:06 <drizztbsd> archlinux and debian
536 2012-11-08 17:48:08 <drizztbsd> :)
537 2012-11-08 17:48:10 <drizztbsd> maybe ubuntu
538 2012-11-08 17:48:24 <drizztbsd> 120$/year!
539 2012-11-08 17:48:30 <midnightmagic> why bother with a centralized-issue ssl cert again?
540 2012-11-08 17:48:42 <gmaxwell> midnightmagic: because web browsers.
541 2012-11-08 17:48:51 <sipa> because reality
542 2012-11-08 17:48:52 <gavinandresen> phantomcircuit: cool, can you email me details and I'll look at doing that?
543 2012-11-08 17:48:56 <drizztbsd> do you need ssl? :P
544 2012-11-08 17:49:01 <phantomcircuit> gavinandresen, also comes with code signing certificates which could be used to sign the bitcoin executables distributed on bitcoin.org
545 2012-11-08 17:49:07 <Luke-Jr> drizztbsd: you'd prefer no security at all? :P
546 2012-11-08 17:49:23 <gavinandresen> phantomcircuit: even better, send me links. I'm all for saving money.
547 2012-11-08 17:49:26 <drizztbsd> bitcoin website does not have "reserved" data
548 2012-11-08 17:49:35 <gmaxwell> use a selfsigned cert = nasty you are being 0wned notice, which is the same one you get when there is a real MITM.  And no ssl means its very easy to substitute binaries since _no one_ checks the signatures.
549 2012-11-08 17:49:47 <Luke-Jr> drizztbsd: SSL prevents someone from sending you different data
550 2012-11-08 17:49:52 <midnightmagic> gmaxwell: No I mean what's the ssl protecting that needs the assurance of the little safe icon?
551 2012-11-08 17:50:04 <drizztbsd> oh for mitm
552 2012-11-08 17:50:04 <gmaxwell> midnightmagic: downloads.
553 2012-11-08 17:50:15 <Luke-Jr> gmaxwell: downloads from where?
554 2012-11-08 17:50:24 <drizztbsd> you can still do mitm using http://www.thoughtcrime.org/software/sslstrip/ :P
555 2012-11-08 17:50:26 <Luke-Jr> bitcoin.org doesn't (and shouldn't IMO) have downloads
556 2012-11-08 17:50:33 <gmaxwell> I assume we're going to move downloads to bitcoin.org.
557 2012-11-08 17:50:34 <midnightmagic> Luke-Jr: I guess the extra assurance so people don't have to verify signatures anymore..
558 2012-11-08 17:50:38 <phantomcircuit> gavinandresen, sent
559 2012-11-08 17:50:43 <drizztbsd> you can still use github or sourceforce (https) :)
560 2012-11-08 17:50:44 <gavinandresen> phantomcircuit: thanks.
561 2012-11-08 17:50:54 <gmaxwell> Because sourforge is just a really attractive target for exploitation.
562 2012-11-08 17:50:57 <midnightmagic> drizztbsd: That's mickey0mouse though.
563 2012-11-08 17:51:07 <phantomcircuit> drizztbsd, sslstrip doesn't work unless the user is ignoring whether it's https or not
564 2012-11-08 17:51:19 <phantomcircuit> and it really doesn't work if you have httpseverywhere installed
565 2012-11-08 17:51:25 <drizztbsd> he's the same user that does not check the signature :P
566 2012-11-08 17:51:32 <Luke-Jr> gmaxwell: well, then we probably need to go the cage route :/
567 2012-11-08 17:51:39 <phantomcircuit> drizztbsd, those are very different groups of users
568 2012-11-08 17:51:44 <Luke-Jr> a normal dedi isn't going to be as secure as SF I think
569 2012-11-08 17:51:53 <gmaxwell> Luke-Jr: I don't know about that. It's not like sourceforge's infrastructure is well secured!
570 2012-11-08 17:51:59 <Luke-Jr> gmaxwell: no? :|
571 2012-11-08 17:51:59 <phantomcircuit> checking for https and checking that the signature is correct are not even close to being the same class of attack
572 2012-11-08 17:52:13 <Luke-Jr> I'd have expected SF was big enough that they only use cages
573 2012-11-08 17:52:25 <sipa> they have mirrors everywhere
574 2012-11-08 17:52:30 <gmaxwell> yup.
575 2012-11-08 17:52:32 <drizztbsd> I check the signature when I build the archlinux official packages ;)
576 2012-11-08 17:52:35 <sipa> i doubt those are all cagrd
577 2012-11-08 17:52:37 <gavinandresen> I think the first step is detection/alert of corrupted downloads. We need that in any case.
578 2012-11-08 17:52:52 <BlueMatt> 
ACTION votes for gitian-based auto-updates

579 2012-11-08 17:53:18 <gavinandresen> BlueMatt: ... but after we code-sign binaries on Mac/Windows....
580 2012-11-08 17:53:20 <sipa> meh python executables
581 2012-11-08 17:53:36 <BlueMatt> gavinandresen: there was a post on the forum that it is easy to work around in (at least) windows
582 2012-11-08 17:53:47 <gmaxwell> In any case, we're getting defocused now.
583 2012-11-08 17:53:48 <drizztbsd> BlueMatt: using a better system than https://gist.github.com/806265
584 2012-11-08 17:53:59 <gavinandresen> So, all of this makes for a pretty heft infrastructure TODO.  And yes, lets focus
585 2012-11-08 17:54:20 <BlueMatt> we finished the agenda a while ago, no?
586 2012-11-08 17:54:27 <gavinandresen> Step 1:  decide on a hosting company or companies and server config.
587 2012-11-08 17:54:37 <gmaxwell> drizztbsd: HSTS makes downgrading attacks pointless.
588 2012-11-08 17:54:52 <sipa> hsts?
589 2012-11-08 17:55:13 <sipa> oh, got it
590 2012-11-08 17:55:17 <gavinandresen> HSTS is the "only use https to contact me?"
591 2012-11-08 17:55:23 <gmaxwell> Yes.
592 2012-11-08 17:55:35 <gavinandresen> right, yes, that's a Good Idea.
593 2012-11-08 17:55:45 <drizztbsd> and obliviusly it does not work on IE
594 2012-11-08 17:55:46 <drizztbsd> :P
595 2012-11-08 17:55:57 <gmaxwell> just make sure you have a backup of the cert so when hackers wipe the site you can put it back up... (lessons learned from bitcoinica)
596 2012-11-08 17:56:40 <gmaxwell> drizztbsd: it doesn't harm browsers that don't do it.. but roughly half the browser out there (and probably way more than half of the bitcoin users' browsers) use it.
597 2012-11-08 17:57:47 <BlueMatt> I suppose this meeting is over, Ima go...I hope to the bloom stuff updated soon (next few days) and it should be ready for more review
598 2012-11-08 17:58:15 <gavinandresen> ACK.  Meeting Officially Adjourned.
599 2012-11-08 17:58:18 <Diablo-D3> http://imgur.com/gallery/YBlrZ
600 2012-11-08 17:58:35 <sipa> while we're at it: acks on -reindex ?
601 2012-11-08 17:58:37 <phantomcircuit> gavinandresen, ps i serously doubt a bootstrap.dat download server would need to be unmetered
602 2012-11-08 17:58:39 <gmaxwell> It's great I think we actually didn't spend a moment discussing actual software. :P
603 2012-11-08 17:58:47 <phantomcircuit> gmaxwell, lol
604 2012-11-08 17:59:07 <rg> hello
605 2012-11-08 17:59:19 <rg> i heard you guys were in need of a dedicated server.
606 2012-11-08 17:59:21 <gavinandresen> phantomcircuit: 3GB per download (assuming http, not torrent) times a few thousand a month....  is a lot of bandwidth
607 2012-11-08 17:59:24 <Luke-Jr> lol
608 2012-11-08 17:59:31 <gavinandresen> phantomcircuit: in any case, we don't have the budget for that
609 2012-11-08 17:59:40 <Diablo-D3> ha ha rg
610 2012-11-08 17:59:47 <rg> you can talk to the sales guy at my company
611 2012-11-08 17:59:51 <rg> if youd like
612 2012-11-08 17:59:55 <Diablo-D3> quit potching the customers I wish I had :<
613 2012-11-08 18:00:01 <rg> Diablo-D3: im working for enzu now
614 2012-11-08 18:00:12 <Diablo-D3> rg: go fuck yourself.
615 2012-11-08 18:00:16 <rg> just think, had you gone through with it
616 2012-11-08 18:00:21 <rg> you wouldve got to complain directly at me
617 2012-11-08 18:00:29 <Luke-Jr> rg: do you guys have cages available btw?
618 2012-11-08 18:00:35 <Diablo-D3> gmaxwell, sipa: do NO do business with enzu
619 2012-11-08 18:00:35 <rg> luke-jr: yeah in los angeles
620 2012-11-08 18:00:42 <Diablo-D3> they're the fucks that deleted cia.vc out of spite
621 2012-11-08 18:00:50 <rg> Diablo-D3: fud
622 2012-11-08 18:01:12 <gavinandresen> rg: can you email me contact information?  gavin@bitcoinfoundation.org
623 2012-11-08 18:01:20 <rg> sure gavin
624 2012-11-08 18:01:24 <rg> he's actually online right now
625 2012-11-08 18:01:28 <rg> you want me to have him contact you directly?
626 2012-11-08 18:01:31 <gavinandresen> rg: sure
627 2012-11-08 18:01:32 <Diablo-D3> gavinandresen: seriously, do NOT do business with them
628 2012-11-08 18:01:40 <Diablo-D3> gavinandresen: they're incompetent at best, scam artists at worst
629 2012-11-08 18:01:42 <phantomcircuit> 
ACTION grabs the popcorn

630 2012-11-08 18:01:46 <rg> diablo-d3's unrecommendation should show you that you SHOULD use them
631 2012-11-08 18:01:49 <gavinandresen> Diablo-D3: email me your concerns
632 2012-11-08 18:01:50 <rg> as he's a major troll
633 2012-11-08 18:01:56 <drizztbsd> I used to use ovh servers
634 2012-11-08 18:01:59 <rg> gavinandresen: diablo has no real concerns
635 2012-11-08 18:02:01 <drizztbsd> unlimited bandwidth
636 2012-11-08 18:02:07 <rg> and the concerns he has , he has no proof of anything
637 2012-11-08 18:02:17 <gavinandresen> rg: ok.  I don't want to get into it here.
638 2012-11-08 18:02:29 <gavinandresen> (whatever "it" is)
639 2012-11-08 18:02:29 <rg> yeah me nither
640 2012-11-08 18:02:31 <Diablo-D3> gavinandresen: I just said them above, they deleted cia.vc, one of the FOSS community's biggest resources, and something Bitcoin itself used, out of spite over an employee quitting
641 2012-11-08 18:02:36 <phantomcircuit> drizztbsd, lol "unlimited" except you can never get past like 5% utilization ;)
642 2012-11-08 18:02:38 <gavinandresen> Diablo-D3: email
643 2012-11-08 18:02:45 <rg> lol
644 2012-11-08 18:02:55 <rg> diablo: message me the rest
645 2012-11-08 18:02:57 <rg> i wanna hear the story
646 2012-11-08 18:03:37 <rg> 
ACTION loves to compare what really happened vs what people are saying happened

647 2012-11-08 18:03:46 <rg> and if you have doubts, just search webhostingtalk
648 2012-11-08 18:03:56 <rg> or ask the other 10,000 customers who're happy
649 2012-11-08 18:04:05 <Diablo-D3> rg: webhostingtalk my ass, we both know thats a troll forum
650 2012-11-08 18:04:12 <rg> LOL
651 2012-11-08 18:04:23 <phantomcircuit> this isn't the appropriate place for this :)
652 2012-11-08 18:04:43 <drizztbsd> phantomcircuit: I don't know, I don't have many traffic :P
653 2012-11-08 18:04:58 <drizztbsd> (except for fuckPsn, but I use github)
654 2012-11-08 18:05:03 <phantomcircuit> drizztbsd, you cant even get 1gbps between ovh servers in the same dc ffs
655 2012-11-08 18:06:44 <Luke-Jr> phantomcircuit: hey, you know where genjix been? :P
656 2012-11-08 18:07:08 <phantomcircuit> Luke-Jr, somewhere in london last i checked
657 2012-11-08 18:07:37 <phantomcircuit> he's avoiding irc and friends because he's realized he wasted a huge % of his time just chatting with people about nothing
658 2012-11-08 18:08:02 <phantomcircuit> he responds to emails usually
659 2012-11-08 18:08:31 <rg> lol
660 2012-11-08 18:08:35 <rg> i was wondering whath appened to him
661 2012-11-08 18:08:40 <rg> he prety much just dropped off the face of the earyth
662 2012-11-08 18:09:08 <phantomcircuit> he got tired of arguing with people about things he doesn't actually care about
663 2012-11-08 18:09:16 <phantomcircuit> so he just stopped being available for arguments
664 2012-11-08 18:09:23 <Luke-Jr> lol
665 2012-11-08 18:09:41 <phantomcircuit> not a bad plan
666 2012-11-08 18:12:06 <rg> lol
667 2012-11-08 18:12:16 <rg> sometimes i wish i never /join #bitcoin-*
668 2012-11-08 18:13:14 <sipa> why?
669 2012-11-08 18:15:41 <phantomcircuit> sipa, huge massive timesink unless you are very good at keeping yourself focused
670 2012-11-08 18:27:09 <midnightmagic> perhaps there're more reasons for genjix to have stopped keeping in touch with the people on irc?
671 2012-11-08 18:32:04 <phantomcircuit> midnightmagic, he told me because it's a waste of time
672 2012-11-08 18:32:12 <phantomcircuit> and experience agrees with his assessment
673 2012-11-08 18:35:55 <jgarzik> word
674 2012-11-08 18:36:39 <jgarzik> phantomcircuit: One merely needs to sharpen one's ignoring skills ;p
675 2012-11-08 18:36:46 <jgarzik> then IRC is great
676 2012-11-08 18:37:13 <sipa> /ignore *
677 2012-11-08 18:38:13 <phantomcircuit> just smashed my head on the ceiling >.>
678 2012-11-08 18:44:03 <jgarzik> gavinandresen, sipa, gmaxwell: Sorry I missed the meeting.  Sometimes kids are demanding ;p  Re TODO: peer selection, peer selection, peer selection.  Post-ultraprune, that will be an obvious, user-facing sore point remaining.
679 2012-11-08 18:44:11 <jgarzik> RE p2p rewrite
680 2012-11-08 18:44:20 <jgarzik> I've been tempted to put everything under a boost.asio io_service
681 2012-11-08 18:44:34 <sipa> i believe libcoin did that too
682 2012-11-08 18:44:35 <jgarzik> that permits threaded and/or async P2P and RPC connections
683 2012-11-08 18:44:43 <gavinandresen> jgarzik: that'd be spiffy.
684 2012-11-08 18:44:56 <jgarzik> I've already done so with https://github.com/jgarzik/rpcsrv
685 2012-11-08 18:45:08 <jgarzik> which is a super-duper high-performance HTTP server for JSON-RPC
686 2012-11-08 18:45:25 <jgarzik> (which is a skeleton, and has no "guts" beyond answering 'echo'-style RPCs)
687 2012-11-08 18:45:32 <jgarzik> should be easy to adapt to bitcoin
688 2012-11-08 18:45:57 <jgarzik> picocoin also gets by just fine without send/receive buffers
689 2012-11-08 18:46:01 <jgarzik> bitcoind could too
690 2012-11-08 18:46:21 <gavinandresen> is anybody running into JSON-RPC bottlenecks?  I think the unoptimized/unindexed database access is the issue with high-RPC-call services
691 2012-11-08 18:46:38 <gavinandresen> (and I've been disappointed not to receive patches to fix that from high-RPC-call services....)
692 2012-11-08 18:46:47 <sipa> the issue is cs_main
693 2012-11-08 18:47:25 <jgarzik> kernel/OS buffers + a size-limited list of queued messages-to-send works great. No need for tx/rx buffers.
694 2012-11-08 18:47:32 <jgarzik> indeed -- RPC still under One Big Lock
695 2012-11-08 18:47:41 <jgarzik> and P2P messages hold that lock, too
696 2012-11-08 18:47:54 <sipa> processing them, yes
697 2012-11-08 18:47:59 <sipa> sending/receiving them, no
698 2012-11-08 18:48:02 <jgarzik> nod
699 2012-11-08 18:48:17 <gmaxwell> jgarzik: I mentioned peer selection indirectly. :) 10:18 < gmaxwell> E.g. block download (not just initial but for hosts that have been offline a few days) matter much more now.
700 2012-11-08 18:48:36 <sipa> yes, real solution is headers-first mode
701 2012-11-08 18:48:47 <sipa> but probably we'll need to stop-gap solution for 0.8
702 2012-11-08 18:49:24 <gmaxwell> jgarzik: the memory usage reductions from getting rid of buffers would be very welcome to some people.
703 2012-11-08 18:49:56 <jgarzik> I like the "get impatient" metric.  If (a) no blocks received in past 60 seconds, and (b) you think more blocks exist "out there" (based on collective examination of nStartingHeight or whatever), then switch to new peer, request blocks.
704 2012-11-08 18:50:34 <sipa> i may try to get peer rotation done for 0.8
705 2012-11-08 18:50:38 <gmaxwell> starting height is a very coarse metric.
706 2012-11-08 18:51:19 <gmaxwell> sipa: I'm ... struggling with the privacy impacts of peer rotation and wondering how to weigh it against isolation risk.
707 2012-11-08 18:52:08 <sipa> hmm?
708 2012-11-08 18:52:45 <gmaxwell> Right now if you're making regular transactions only your few peers which you don't really change without nodes getting bounced... will identify your IP for sure (if they're looking)
709 2012-11-08 18:53:21 <gmaxwell> maybe that concern is just out of scope now... "use tor"
710 2012-11-08 18:53:40 <gmaxwell> though fundimentally I think it's interesting that initial txn announcement is a different problem than general p2p behavior.
711 2012-11-08 18:53:50 <jgarzik> Of course, there is always the Google approach of asking N peers in parallel the same question
712 2012-11-08 18:53:56 <jgarzik> and judging based on answers
713 2012-11-08 18:54:19 <jgarzik> wastes bandwidth, but gives you fast service
714 2012-11-08 18:55:10 <gmaxwell> well we could be timing the version/addr messages when the connection goes up.
715 2012-11-08 18:55:22 <gmaxwell> and prefer to pull from the peer that was fastest.
716 2012-11-08 18:55:26 <jgarzik> RE P2P networking: we do not disable Nagle (TCP_NODELAY), so the kernel delays our small packets sometimes, in anticipation of more local sending.
717 2012-11-08 18:55:59 <sipa> since we send in bursts anyway, that is probably not necessary
718 2012-11-08 18:56:16 <sipa> (i.e., TCP_NODELAY wouldn't hurt)
719 2012-11-08 18:56:17 <gmaxwell> jgarzik: we have so many problems before nagle is a concern... :P
720 2012-11-08 18:56:42 <sipa> huh? memset being optimized out?
721 2012-11-08 18:56:48 <sipa> maybe it gets inlined
722 2012-11-08 18:56:49 <gmaxwell> next you'll suggest https://www.usenix.org/conference/nsdi12/minion-unordered-delivery-wire-compatible-tcp-and-tls
723 2012-11-08 18:57:26 <gmaxwell> sipa: I'm pretty sure GCC can, at least of stack variables when it uses the compiler internal one and knows you wont read it.
724 2012-11-08 19:00:00 <jgarzik> I'd like proof of gcc behavior
725 2012-11-08 19:00:10 <jgarzik> rather than supposition
726 2012-11-08 19:09:24 <jgarzik> Apparently Ruby doesn't scale.  Solution?  Java!  Almost 10,000 tweets per second.  http://www.theregister.co.uk/2012/11/08/twitter_epic_traffic_saved_by_java/
727 2012-11-08 19:12:24 <phantomcircuit> gavinandresen, anybody calling the rpc api a lot is probably doing something wrong ;)
728 2012-11-08 19:16:55 <phantomcircuit> jgarzik, it probably has more to do with the fact that they now have engieers with the common sense to drop ruby in the first place
729 2012-11-08 19:17:09 <rg> Java over Ruby? jeez
730 2012-11-08 19:17:38 <rg> Drowning over fire death
731 2012-11-08 19:18:39 <phantomcircuit> rg, it's a trade off of developer time over efficient hardware utilization
732 2012-11-08 19:18:52 <phantomcircuit> which actually a lot of the language wars stuff basically seems to come down to
733 2012-11-08 19:48:56 <rg> gavin
734 2012-11-08 19:48:58 <rg> we need to speak
735 2012-11-08 19:49:04 <rg> however im currently in th emiddle of two installs
736 2012-11-08 19:49:13 <rg> wll you be around in a little while?
737 2012-11-08 19:49:48 <gavinandresen> rg: no, I'm about to disappear.  Back tomorrow....
738 2012-11-08 19:56:12 <rg> ok
739 2012-11-08 19:56:14 <rg> that's no prob
740 2012-11-08 19:56:21 <rg> i just wanted to let you know that paying in Bitcoin will be fnie
741 2012-11-08 19:56:27 <gavinandresen> cool
742 2012-11-08 19:56:31 <rg> tfine*
743 2012-11-08 20:12:39 <D34TH> #1984
744 2012-11-08 20:23:35 <edcba> D34TH: #old
745 2012-11-08 20:26:44 <BlueMatt> sipa: ping
746 2012-11-08 20:44:39 <sipa> BlueMatt: pong
747 2012-11-08 20:45:30 <BlueMatt> sipa: so Im trying to get bloom filters to pass with your CPartialMerkleTree...I clearly did something wrong but Im not sure where...
748 2012-11-08 20:46:15 <sipa> what doesn't work?
749 2012-11-08 20:46:50 <BlueMatt> well, first of all...git push...give me a minute
750 2012-11-08 20:47:22 <sipa> so you'd construct a CPartialMerkleTree by passing it a list of txids and a bool vector to select a subset from it, serialize & send it, deserialize, deconstruct resulting in a matched txid list + merkle root, verify that merkle root against the one in the block header
751 2012-11-08 20:47:54 <BlueMatt> sipa: yep, pretty sure thats exactly what I did
752 2012-11-08 20:48:05 <sipa> maybe there's a bug :)
753 2012-11-08 20:48:08 <BlueMatt> (Im sure its something obvious, just hoping you can spot it)
754 2012-11-08 20:48:15 <sipa> ok, will have a look
755 2012-11-08 20:48:54 <BlueMatt> sipa: https://github.com/TheBlueMatt/bitcoin/tree/newbloom
756 2012-11-08 20:49:00 <BlueMatt> specifically the last commit
757 2012-11-08 20:49:25 <BlueMatt> (error is in bloom tests:192 (among many others)
758 2012-11-08 20:52:10 <sipa> where is CMerkleBlock implemented?
759 2012-11-08 20:53:13 <sipa> oh, never mind
760 2012-11-08 20:54:23 <sipa> BlueMatt: do you call CBlock::CalcMerkleRoot?
761 2012-11-08 20:55:10 <BlueMatt> dont think so...what did I miss?
762 2012-11-08 20:55:26 <sipa> eh, that shouldn't be necessary of course
763 2012-11-08 20:55:39 <sipa> assuming the serialized form contains the correct merkle root
764 2012-11-08 20:56:24 <BlueMatt> I didnt change the serialized form since it was working with the old code (and the serialized form was copied/pasted out of a HexStr() call)
765 2012-11-08 21:03:45 <sipa> BlueMatt: which other lines fail?
766 2012-11-08 21:04:21 <BlueMatt> all of the " vMatched[i] == merkleBlock.vMatchedTxn[i].first" and "vMatched.size() == merkleBlock.vMatchedTxn.size()" and "merkleBlock.txn.ExtractMatches(vMatched) == block.hashMerkleRoot" checks afaict
767 2012-11-08 21:06:16 <sipa> shouldn't it be .second ?
768 2012-11-08 21:06:37 <BlueMatt> did I really...
769 2012-11-08 21:07:44 <sipa> for the rest... it may be useful to add some error printf before the return 0's in ExtractMatches
770 2012-11-08 21:07:59 <BlueMatt> tried that...its not returning 0 either
771 2012-11-08 21:08:24 <sipa> hmm, what is vMatched.size() ?
772 2012-11-08 21:09:37 <BlueMatt> 0
773 2012-11-08 21:09:39 <BlueMatt> on 192
774 2012-11-08 21:10:37 <sipa> and merkleBlock.txn.vHash.size() ?
775 2012-11-08 21:13:07 <sipa> and vBits?
776 2012-11-08 21:13:39 <sipa> (this weekend i may try experimenting myself a bit, but not now, if you can wait)
777 2012-11-08 21:13:45 <BlueMatt> sure, I can wait
778 2012-11-08 21:13:49 <BlueMatt> I have to go now anyway
779 2012-11-08 21:13:56 <BlueMatt> see you later
780 2012-11-08 21:44:53 <nanotube> BlueMatt: the issues feed should be fixed now. :)
781 2012-11-08 22:11:31 <sudog> ;;bc,stats
782 2012-11-08 22:11:33 <gribble> Current Blocks: 207102 | Current Difficulty: 3304356.3929903 | Next Difficulty At Block: 207647 | Next Difficulty In: 545 blocks | Next Difficulty In About: 3 days, 14 hours, 26 minutes, and 35 seconds | Next Difficulty Estimate: 3398863.17791331 | Estimated Percent Change: 2.8600663392