1 2013-08-19 01:38:14 <jgarzik> hrm
2 2013-08-19 01:38:23 <jgarzik> this javascript block parsing is shockingly slow
3 2013-08-19 01:38:34 <Diablo-D3> >javascript
4 2013-08-19 01:38:35 <Diablo-D3> >slow
5 2013-08-19 01:38:37 <jgarzik> several minutes, just to get to block header 72000
6 2013-08-19 01:38:39 <Diablo-D3> GEE I WONDER
7 2013-08-19 01:38:59 <Diablo-D3> what browser
8 2013-08-19 01:39:08 <petertodd> jgarzik: Yeah, I'm still trying to get to block 100 on my babbage machine.
9 2013-08-19 01:39:10 <Diablo-D3> because chrome has shit js performance
10 2013-08-19 01:46:28 <jgarzik> Diablo-D3, This is not chrome, nor any browser
11 2013-08-19 01:46:48 <Diablo-D3> er, node.js?
12 2013-08-19 01:47:03 <Diablo-D3> might as well be chrome then, its v8
13 2013-08-19 01:47:11 <jgarzik> This is JIT'd node.js, which seems slower than non-JIT'd python.
14 2013-08-19 01:51:13 <Diablo-D3> jgarzik: I wouldnt doubt it
15 2013-08-19 01:52:16 <gmaxwell> jgarzik: there are a bunch of things you can do to improve JS performance, mostly that involve being sensitive to where your code is triggering copying and garbage collection. But I'm mostly clueless to that stuff, so I can't give much advice.
16 2013-08-19 01:53:45 <jgarzik> might be that I'm abusing it
17 2013-08-19 01:54:06 <jgarzik> my header database links each block object to the previous one
18 2013-08-19 01:54:25 <jgarzik> and there are secondary indices that add further references to block headers
19 2013-08-19 01:54:44 <jgarzik> simply storing in level db rather than memory might fix the problem
20 2013-08-19 01:55:00 <phantomcircuit> jgarzik, that's pretty much the worst possible case for the gc
21 2013-08-19 01:55:08 <phantomcircuit> which tried to do cycle detection
22 2013-08-19 01:55:10 <jgarzik> though it pains me to put the header db in leveldb, rather than RAM
23 2013-08-19 01:55:11 <phantomcircuit> tries*
24 2013-08-19 01:59:21 <Vinnie_win> so what's all this about the controversy over my pull request regarding the subtree?
25 2013-08-19 02:01:21 <jgarzik> I also wonder how well node's hash tables can handle a table of size 250,000
26 2013-08-19 02:05:06 <phantomcircuit> jgarzik, i would guess pretty well actually
27 2013-08-19 02:05:18 <phantomcircuit> i bet a bunch of the internal language stuff is actually based on hash tables
28 2013-08-19 02:05:26 <jgarzik> one hopes
29 2013-08-19 02:05:55 <jgarzik> ACTION could probably rewrite to have an in-ram hash table of block headers, then store the hashes in various other data structures, rather than a direct reference
30 2013-08-19 02:06:22 <jgarzik> keeps everything happily refcount==1
31 2013-08-19 02:06:36 <jgarzik> no circular references
32 2013-08-19 02:06:47 <sipa> Vinnie_win: we rewrote history in the bitcoin/leveldb repo, to clearly build on the official leveldb repo, and then got-subtree merged it into bitcoin
33 2013-08-19 02:06:59 <sipa> Vinnie_win: worked surprisingly well
34 2013-08-19 02:24:54 <hardfalcon> Luke-Jr / sipa: seems that the openssl folks have fixed padlock support in the meantime, everything works as it should on my machine (this wasn't the case a few months ago, I'm now using 1.0.1e), and openssl does even use the hardware accelerated montgomery multiplicator for RSA and DSA
35 2013-08-19 02:25:55 <hardfalcon> given the fact that the bitcoin sources include <openssl/sha.h> and a few other openssl headers, I assume that bitcoin-qt uses those optimizations already
36 2013-08-19 02:26:53 <phantomcircuit> hardfalcon, maybe maybe not
37 2013-08-19 02:26:58 <phantomcircuit> iirc openssl needs you to pass an engine parameter to use padlock
38 2013-08-19 02:27:15 <hardfalcon> nope
39 2013-08-19 02:27:15 <hardfalcon> you can enable it in /etc/ssl/openssl.cnf
40 2013-08-19 02:27:15 <phantomcircuit> hardfalcon, you can set a default engine?
41 2013-08-19 02:27:15 <phantomcircuit> which bitcoin certainly doesn't provide
42 2013-08-19 02:27:23 <hardfalcon> and on my machine it is enabled by default
43 2013-08-19 02:27:37 <phantomcircuit> then maybe
44 2013-08-19 02:27:38 <hardfalcon> as I've set the padlock engine as default in that config file, of course
45 2013-08-19 02:27:54 <phantomcircuit> iirc openssl is used for ecdsa SHA256 RIPEMD160
46 2013-08-19 02:27:56 <phantomcircuit> and nothing else
47 2013-08-19 02:28:15 <hardfalcon> well, sha256 is hw accelerated then
48 2013-08-19 02:28:19 <phantomcircuit> im not sure any of those would see a significant improvement from padlock
49 2013-08-19 02:28:27 <hardfalcon> for ecdsa I don't know
50 2013-08-19 02:28:30 <phantomcircuit> hardfalcon, yeah except they're short hashes
51 2013-08-19 02:28:38 <phantomcircuit> so it might actually be slower
52 2013-08-19 02:29:09 <phantomcircuit> hw acceleration of hash functions tends to be focused on hashing large amounts of data
53 2013-08-19 02:29:12 <phantomcircuit> like ssl connections
54 2013-08-19 02:29:17 <phantomcircuit> not 80 byte block headers
55 2013-08-19 02:30:13 <hardfalcon> yes, Luke-Jr showed me a great paper about montgomery multiplication with the padlock engine and how efficiency scales there with different keysizes
56 2013-08-19 02:30:31 <hardfalcon> no the hash function is extremely performant
57 2013-08-19 02:30:58 <phantomcircuit> shrug
58 2013-08-19 02:31:11 <phantomcircuit> even so sha256 on a modern cpu is only like 2 microseconds
59 2013-08-19 02:31:28 <hardfalcon> well not on a via nano cpu ;)
60 2013-08-19 02:32:01 <hardfalcon> we're talking of a 1.2ghz dualcore CPU with a computational power in the range of an intel atom cpu
61 2013-08-19 02:34:01 <hardfalcon> openssl speed -evp sha256
62 2013-08-19 02:34:01 <hardfalcon> sha256 8821.02k 20872.01k 37887.05k 47520.11k 51252.23k
63 2013-08-19 02:34:01 <hardfalcon> type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
64 2013-08-19 02:34:08 <hardfalcon> (with the padlock engine)
65 2013-08-19 02:39:00 <phantomcircuit> hardfalcon,
66 2013-08-19 02:39:01 <phantomcircuit> sha256 31017.82k 72729.71k 131471.93k 163056.64k 175551.83k
67 2013-08-19 02:39:01 <phantomcircuit> type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
68 2013-08-19 02:39:04 <phantomcircuit> i3-2100
69 2013-08-19 02:39:12 <phantomcircuit> not even the fastest cpu i have on this desk
70 2013-08-19 02:39:23 <hardfalcon> argh sry
71 2013-08-19 02:39:30 <hardfalcon> got something confused
72 2013-08-19 02:40:18 <hardfalcon> the padlock engine doesn't work with the archlinux stock package of openss, it only works with a package that I compiled by mysef some time ago
73 2013-08-19 02:40:50 <hardfalcon> had forgotten that and reinstaled openssl from the normal repositories this afternoon because I had fucked up a config file
74 2013-08-19 02:45:07 <hardfalcon> but strange enough there is no difference on whether I use the padlock engine or not, the benchmark results stay the same
75 2013-08-19 02:45:38 <hardfalcon> even when I delete /usr/lib/engines/libpadlock.so
76 2013-08-19 02:48:32 <hardfalcon> anyhow, I've just checked with gnutls-cli --benchmark-ciphers and gnutls-cli --benchmark-soft-ciphers
77 2013-08-19 02:49:41 <hardfalcon> --benchmark-ciphers (this one uses the padlock engine) gives 0.31 GB/sec for SHA256, whilst --benchmark-soft-ciphers only gives 47.21 MB/sec (both with a payload size of 16384, though)
78 2013-08-19 04:04:17 <warren> Luke-Jr: sorry I keep losing this, what git command adds the origin-pull thing again?
79 2013-08-19 04:07:45 <warren> nevermind, https://gist.github.com/piscisaureus/3342247
80 2013-08-19 04:47:53 <oleganza> lianj: hi there
81 2013-08-19 04:48:12 <oleganza> well, looks like nickserv let me in finally.
82 2013-08-19 05:09:40 <swulf--> can someone confirm that there was two blocks at height 252481 generated?
83 2013-08-19 05:10:37 <swulf--> block 0000000000000049ffe667048ee8901f5609c48a9ff71319445d19bf7c1b09dc doesn't seem to be available on any block explorer, but I'm sitting here looking at it in my client
84 2013-08-19 05:25:45 <gmaxwell> 2013-08-16 16:30:22 got inventory: block 0000000000000049ffe667048ee8901f5609c48a9ff71319445d19bf7c1b09dc new
85 2013-08-19 05:25:48 <gmaxwell> 2013-08-16 16:30:22 askfor block 0000000000000049ffe667048ee8901f5609c48a9ff71319445d19bf7c1b09dc 0 (00:00:00)
86 2013-08-19 05:25:51 <gmaxwell> 2013-08-16 16:30:22 sending getdata: block 0000000000000049ffe667048ee8901f5609c48a9ff71319445d19bf7c1b09dc
87 2013-08-19 05:25:54 <gmaxwell> 2013-08-16 16:30:23 received block 0000000000000049ffe667048ee8901f5609c48a9ff71319445d19bf7c1b09dc
88 2013-08-19 05:26:43 <gmaxwell> I never setbest it, so I assume it was already orphan when I fetched it.
89 2013-08-19 05:27:32 <swulf--> ahh
90 2013-08-19 05:27:39 <swulf--> so it does exist
91 2013-08-19 05:27:55 <swulf--> ACTION shakes a finger at blockchain.info for not letting me view orphaned blocks
92 2013-08-19 05:32:43 <theymos> Did you guys see that Liraz Siri released the blockexplorer.com source? He's looking for help in improving it. https://github.com/lirazsiri/blockexplorer
93 2013-08-19 07:11:29 <Luke-Jr> maaku: your signature on the 0.9 ML thread is invalid O.o
94 2013-08-19 07:12:03 <sipa> and you should feel invalid.
95 2013-08-19 07:12:55 <Luke-Jr> o.O
96 2013-08-19 07:13:17 <Luke-Jr> he should?
97 2013-08-19 07:13:45 <sipa> (it's a meme, don't take it seriously)
98 2013-08-19 07:18:00 <Luke-Jr> o
99 2013-08-19 09:02:05 <drizztbsd> hi, do you know BDB is migrated to AGPL? will you always use it?
100 2013-08-19 09:03:03 <sipa> we've been wanting to get rid of it for a long time
101 2013-08-19 09:03:12 <sipa> but it means incompatible wallets, etc
102 2013-08-19 09:03:41 <drizztbsd> so? no way? :)
103 2013-08-19 09:04:00 <sipa> sure way
104 2013-08-19 09:04:11 <sipa> just needs some time
105 2013-08-19 09:04:17 <drizztbsd> oh ok
106 2013-08-19 09:11:14 <CodeShark> for a time the bdb support could be retained in parallel with some other storage engine to allow people to easily convert their wallets
107 2013-08-19 09:11:33 <sipa> sure
108 2013-08-19 09:11:44 <sipa> and after that a separate conversion tool could be shipped
109 2013-08-19 09:11:55 <sipa> but first we need a different wallet format in the first place :)
110 2013-08-19 09:14:51 <drizztbsd> xml?
111 2013-08-19 09:14:52 <drizztbsd> ACTION hides
112 2013-08-19 09:14:57 <CodeShark> lol
113 2013-08-19 09:15:24 <sipa> how about ASN.1?
114 2013-08-19 09:15:36 <TD> a brilliant format for sure
115 2013-08-19 09:15:50 <sipa> with a EBCDIC for string characters
116 2013-08-19 10:49:53 <theorbtwo> Don't forget to encrypt it, and then ascii-armor it... and then convert the ascii to utf16.
117 2013-08-19 10:50:29 <marcusw> b32 it, then b64 the b32 just to be sure
118 2013-08-19 10:51:07 <sipa> please add some error correction codes. kthxbye.
119 2013-08-19 10:52:40 <drizztbsd> and rot13
120 2013-08-19 10:52:49 <sipa> triple-rot13
121 2013-08-19 10:53:01 <drizztbsd> lol sounds like a triple-xor :P
122 2013-08-19 10:53:26 <marcusw> crc32 it
123 2013-08-19 10:53:31 <tgs3> also add BOM in front
124 2013-08-19 10:53:37 <marcusw> I read that's a good thing to do on forbes
125 2013-08-19 10:53:43 <tgs3> so it's more notepad.exe compatible
126 2013-08-19 10:54:41 <marcusw> MS newlines obviously
127 2013-08-19 10:57:37 <sipa> drizztbsd: yes but triple xor is just xor; triple-rot13 is obviously rot39
128 2013-08-19 10:58:21 <CodeShark> what group is generated by xor and rot13? :)
129 2013-08-19 10:59:20 <CodeShark> and the vector of alternating bits
130 2013-08-19 11:02:29 <CodeShark> the Klein four-group?
131 2013-08-19 13:18:12 <TD> gmaxwell: how comes the OTR whitespace trick uses whitespaces that are so verbose and annoying? i just had to switch off OTR because my brother was complaining that every line i sent had tons of random whitespace after it
132 2013-08-19 13:19:25 <handle> lol yeah, I've noticed that as well
133 2013-08-19 13:19:49 <handle> maybe it's so that more information can be stored in the whitespace? though I think it's the same every time and is just used to identify OTR
134 2013-08-19 13:20:02 <handle> that said, isn't it just a combination of tabs and spaces?
135 2013-08-19 13:21:28 <TD> i assume it's so large it causes a lot of wrapping in small text boxes
136 2013-08-19 13:21:41 <TD> there must be a better way to do that. unicode provides a variety of non-printing control characters
137 2013-08-19 13:21:48 <TD> you could reverse text direction three times and i bet that'd be a unique signature
138 2013-08-19 13:21:57 <TD> or at least not append the whitespace to every single message!
139 2013-08-19 13:24:00 <sipa_> TD: i bet you'll get complains about people seeing tons of weird characters if you use unicode :D
140 2013-08-19 13:25:15 <marcusw> your broken unicode, your problem
141 2013-08-19 13:25:50 <sipa_> hahaha
142 2013-08-19 13:25:53 <sipa_> right :D
143 2013-08-19 13:26:43 <sipa> good luck with that
144 2013-08-19 13:28:39 <Luke-Jr> yeah
145 2013-08-19 13:28:51 <Luke-Jr> I added unicode to BFGMiner. Now I had to make it opt-in :/
146 2013-08-19 13:30:05 <sipa> works fine for me :)
147 2013-08-19 13:31:13 <marcusw> meanwhile on nameless irc, unicode allowed everywhere
148 2013-08-19 13:31:20 <marcusw> even in nicks and chan names
149 2013-08-19 13:32:11 <Luke-Jr> sipa: you don't use Chinese Windows :/
150 2013-08-19 13:32:19 <sipa> Luke-Jr: thank god.
151 2013-08-19 13:32:38 <Luke-Jr> ACTION would have thought Chinese OS had *better* Unicode support, but .. wtf
152 2013-08-19 13:33:57 <_ingsoc> Who the hell uses a Chinese OS?
153 2013-08-19 13:34:04 <_ingsoc> Except for the Chinese, etc.
154 2013-08-19 13:34:11 <Graet> mainly chinese people
155 2013-08-19 13:34:33 <sipa> maybe there are more of those than people using an English OS :)
156 2013-08-19 13:35:01 <_ingsoc> You're right, sipa. I'll be honest and say my world isn't very big. :(
157 2013-08-19 13:37:03 <marcusw> sometimes, I feel like it's a good thing they have their own LAN
158 2013-08-19 13:40:20 <Graet> anyone with ops in #bitcoin ?
159 2013-08-19 14:05:51 <arioBarzan> is it possible to mine on local network?
160 2013-08-19 14:06:10 <arioBarzan> without connecting to any other node
161 2013-08-19 14:06:23 <Scrat> I have a local blockchain :)
162 2013-08-19 14:06:32 <Scrat> problem is that noone else is accepting it
163 2013-08-19 14:07:05 <arioBarzan> that is not a problem for me. I need it just for learning some stuff
164 2013-08-19 14:07:24 <arioBarzan> how could I have a local blockchain
165 2013-08-19 14:07:27 <arioBarzan> ?
166 2013-08-19 14:07:45 <drizztbsd> arioBarzan: testnet
167 2013-08-19 14:08:21 <arioBarzan> isn't it possible on main net?
168 2013-08-19 14:08:46 <drizztbsd> without connection to any node?
169 2013-08-19 14:08:52 <arioBarzan> yes
170 2013-08-19 14:09:00 <arioBarzan> When satoshi started mining, I suspect there were no other node. So it should work. But I don't know how.
171 2013-08-19 14:10:29 <arioBarzan> Scrat: how did you run your bitcoind for a local blockchain?
172 2013-08-19 14:10:43 <TheLordOfTime> arioBarzan, probably as testnetinabox or something
173 2013-08-19 14:10:57 <TheLordOfTime> at least for testing
174 2013-08-19 14:11:06 <Scrat> what TLOT said
175 2013-08-19 14:12:09 <arioBarzan> TheLordOfTime: how Satoshi could mine on main net so?
176 2013-08-19 14:12:28 <TheLordOfTime> arioBarzan, you might reread what Scrat said
177 2013-08-19 14:12:32 <TheLordOfTime> <Scrat> I have a local blockchain :)
178 2013-08-19 14:12:32 <TheLordOfTime> <Scrat> problem is that noone else is accepting it
179 2013-08-19 14:13:31 <TheLordOfTime> if you create your own local chain, it'll always be your own network and only your own network, so how Satoshi started mining is irrelevant if you want to actually mine and get usable worth-something coins
180 2013-08-19 14:13:52 <TheLordOfTime> 'course i mine testnet as part of generating testnet coin for me to test my RPC call programs, but I'm not a full-blown miner ;P
181 2013-08-19 14:14:02 <TheLordOfTime> (testnet in a box helps with that xD)
182 2013-08-19 14:14:22 <drizztbsd> arioBarzan: testnet or main net is the same
183 2013-08-19 14:14:30 <drizztbsd> (protocol & co)
184 2013-08-19 14:15:46 <arioBarzan> I don't care about others accepting those worthless coins that I would make on a local network.
185 2013-08-19 14:16:51 <arioBarzan> I need to mine a block ( which of course would be worthless for others ) just for testing some transactions on a local node.
186 2013-08-19 14:17:13 <drizztbsd> I hope not on real current difficult :P
187 2013-08-19 14:17:19 <jgarzik> mornin'
188 2013-08-19 14:17:22 <handle> that sounds like something that might be best done via unit tests
189 2013-08-19 14:18:04 <arioBarzan> current difficulty is 1 on my local node
190 2013-08-19 14:19:23 <helo> something about needing two nodes...
191 2013-08-19 14:20:04 <arioBarzan> do I need to have at least running two nodes?
192 2013-08-19 14:21:01 <helo> i think i recall someone saying that a while back regarding running an isolated test network...
193 2013-08-19 14:21:24 <arioBarzan> helo: thansk, I will try that.
194 2013-08-19 14:23:10 <Luke-Jr> arioBarzan: this is what testnet is for.
195 2013-08-19 14:24:04 <arioBarzan> Luke-Jr: I know. But my question is how satoshi could mine on his node at the first place?
196 2013-08-19 14:24:25 <Luke-Jr> arioBarzan: Satoshi did not start Bitcoin alone.
197 2013-08-19 14:24:58 <arioBarzan> Luke-Jr: So what is the minimum number of nodes for a isolated network?
198 2013-08-19 14:25:15 <Luke-Jr> arioBarzan: 2
199 2013-08-19 14:25:23 <arioBarzan> Luke-Jr: thanks a lot.
200 2013-08-19 14:25:23 <Luke-Jr> arioBarzan: this is by definition testnet-in-a-box
201 2013-08-19 14:26:19 <Luke-Jr> np
202 2013-08-19 14:27:11 <arioBarzan> Luke-Jr: is there any special trick for that , or it should work if I run the daemons on two nodes on the same local network listening on 8333?
203 2013-08-19 14:27:44 <helo> -addnode=other_node_ip
204 2013-08-19 14:28:25 <arioBarzan> helo: thanks
205 2013-08-19 15:37:20 <Luke-Jr> petertodd: it occurs to me that as long as I'm storing my email in a git repo, I should probably timestamp the HEAD commit hash.. you have something setup for that already so I don't bloat the blockchain?
206 2013-08-19 15:37:52 <kinlo> wouldn't it make more sense to put something like that into namecoin?
207 2013-08-19 15:38:02 <kinlo> you can assign a clear name and assign a clear value
208 2013-08-19 15:38:21 <kinlo> instead of hiding the data in a chain not intended for that purpose
209 2013-08-19 15:38:50 <Luke-Jr> kinlo: well, it really should just be its own timestamp data merged-mined separately from any blockchain
210 2013-08-19 15:39:03 <Luke-Jr> but petertodd I believe has something already timestamping data en masse
211 2013-08-19 15:39:35 <kinlo> still :)
212 2013-08-19 16:26:07 <gmaxwell> sipa: jgarzik: luke-jr: RE pull 2910 do try to be somewhat gentle with the new contributor, (though I think I did already warn them that depending on checkpoints was not a good way to go about that)
213 2013-08-19 16:26:31 <Luke-Jr> gmaxwell: yeah, I just misread it the first time, sorry
214 2013-08-19 16:26:58 <Luke-Jr> I thought they were trying to eliminate the blockchain before the last checkpoint entirely
215 2013-08-19 16:28:08 <gmaxwell> yea, I actually read the title of the pull the same way.
216 2013-08-19 16:29:06 <sipa> oh, i wasn't aware i was being unfriendly - i certainly didn't intend to be :(
217 2013-08-19 16:30:32 <sipa> i agree with the idea in general, just said that the not storing part won't be needed
218 2013-08-19 16:30:52 <an3k> my bitcoin-qt 0.8.3-beta client cannot connect to other nodes anymore, thus I get no data
219 2013-08-19 16:30:57 <gmaxwell> ::nods:: no biggie, it was just a bit of a pile on. :)
220 2013-08-19 16:31:08 <gmaxwell> an3k: how do you know it can't connect?
221 2013-08-19 16:31:12 <gmaxwell> an3k: are you using a proxy?
222 2013-08-19 16:31:17 <an3k> In the debug window I see connections jumping from 0 to 2 back to 0, etc.
223 2013-08-19 16:31:20 <an3k> no proxy
224 2013-08-19 16:32:09 <an3k> the debug log says it tries to connect to a node, gets connected, sends version message and then disconnects.
225 2013-08-19 16:32:29 <gmaxwell> an3k: how long have you left it running? is this on a network which is known for censoring its users?
226 2013-08-19 16:33:04 <an3k> some hours but not a single block was processed. no censoring.
227 2013-08-19 16:33:40 <an3k> what I don't understand in the debug.log: "connected to self at 10.0.0.1:58663, disconnecting"
228 2013-08-19 16:33:52 <an3k> 10.0.0.1 is NOT the computer the bitcoin-qt client runs on
229 2013-08-19 16:34:16 <gmaxwell> do you have any connect or addnode in your bitcoin.conf?
230 2013-08-19 16:34:36 <an3k> where is that file located?
231 2013-08-19 16:34:48 <gmaxwell> same place the debug.log exists, if you have one at all.
232 2013-08-19 16:35:00 <an3k> don't have one
233 2013-08-19 16:35:22 <an3k> .locl, db.log, debug.log, peers.dat and wallet.dat
234 2013-08-19 16:35:30 <an3k> i mean .lock
235 2013-08-19 16:36:39 <gmaxwell> can you email me your debug.log? It may contain your IP address and IDs of your transactions, but otherwise no secret data.
236 2013-08-19 16:36:52 <an3k> sure
237 2013-08-19 16:38:32 <an3k> to what address should I send it?
238 2013-08-19 16:39:26 <gmaxwell> greg@xiph.org works.
239 2013-08-19 16:41:33 <an3k> wow, never knew you can have an image inside a public key :)
240 2013-08-19 16:42:10 <gmaxwell> Yea, almost no one uses that feature.
241 2013-08-19 16:43:46 <Luke-Jr> lol
242 2013-08-19 16:44:12 <Luke-Jr> tidbit: plaintext email supports a single image too
243 2013-08-19 16:44:32 <an3k> mail sent
244 2013-08-19 16:44:34 <Luke-Jr> http://en.wikipedia.org/wiki/X-Face
245 2013-08-19 16:45:13 <TD> ACTION put a photo into his public key when he uploaded it
246 2013-08-19 16:45:18 <TD> that reminds me. i need to get you guys to sign it :)
247 2013-08-19 16:45:27 <Luke-Jr> TD: you should have reminded us in San Jose! :p
248 2013-08-19 16:45:30 <gmaxwell> ACTION signs only TD's face.
249 2013-08-19 16:45:32 <an3k> ;p
250 2013-08-19 16:45:36 <TD> i didn't use PGP back then
251 2013-08-19 16:45:56 <Luke-Jr> ACTION facepalms
252 2013-08-19 16:45:58 <TD> i guess what gmaxwell said about how we should build upon our cryptographic heritage resonated, especially w.r.t. recent disclosures
253 2013-08-19 16:46:16 <Luke-Jr> gmaxwell: can you do that?
254 2013-08-19 16:46:33 <sipa> you sign individual identities, so i guess yes
255 2013-08-19 16:46:47 <TD> well, PGP has a lot of problems. i mostly care about signing and all my mail is signed automatically for a long time already, by DKIM. having a stronger system in place can't really hurt though, at least for occasional use. of course then the issue is, if you don't sign everything, a MITM can just drop the signature and nobody will know
256 2013-08-19 16:46:52 <sipa> you don't sign the key - you sign the connection between the key and the claimed identity
257 2013-08-19 16:47:00 <TD> it's not really ideal. i want to get time to compile and install Pond at some point.
258 2013-08-19 16:47:13 <an3k> I would prefer a new cryptography in case the current used I already "hacked" by NSA or whoever
259 2013-08-19 16:47:16 <TD> unfortunately it's written in Go and uses GTK3
260 2013-08-19 16:47:21 <gmaxwell> Luke-Jr: absolutely. Best practice (which it seems no one does in pgp) is to sign each identity seperately, and then email each signature to each address, thereby validating that they can recieve mail at those addresses. I'm not sure how you transmit a signature to a face though. :P
261 2013-08-19 16:47:27 <TD> so is pretty much a nightmare to compile on a Mac, and when I'm at work I have more important things to do :)
262 2013-08-19 16:47:51 <Luke-Jr> TD: KMail is most excellent if you get the 4.3 version before KDE rewrote it <.<
263 2013-08-19 16:48:16 <Luke-Jr> gmaxwell: QR-Code hand delivery
264 2013-08-19 16:48:25 <gmaxwell> Luke-Jr: The only person I've ever seen do the split stuff is Bradley Kuhn, though supposidly there is some perl script out there that makes it easy if you have sendmail working on your local host.
265 2013-08-19 16:48:54 <sipa> gmaxwell: yup
266 2013-08-19 16:49:03 <sipa> i've used that script once after a key-signing party
267 2013-08-19 16:49:14 <sipa> (it involved 200 people; i wasn't manually going to send 200 mails...)
268 2013-08-19 16:49:18 <an3k> But the whole system is only secure as long as nobody has access to the private key and password and exactly there I see a problem.
269 2013-08-19 16:49:34 <Luke-Jr> an3k: ???
270 2013-08-19 16:49:41 <an3k> Luke-Jr?
271 2013-08-19 16:50:28 <an3k> Don't you think the NSA has some backdoors in newer operating systems or at least a "illegal" solution to hack into?
272 2013-08-19 16:50:47 <Luke-Jr> an3k: maybe for the few people who still run Mac/Windows
273 2013-08-19 16:51:38 <an3k> well, on linux you're far more secure, you're absolutely right but sadly most people use windows on intel :(
274 2013-08-19 16:52:11 <gmaxwell> an3k: Security is a continuum, I'm sure the NSA (^wlockheed martin) has some awesome back doors, but they're not likely to go using on them on things which aren't the greatest threat they expect to have in the near future. So your goal is to not be the NSA's greatest threat and to be secure against everyone _except_ the NSA. :P
275 2013-08-19 16:52:38 <Luke-Jr> http://bitcoinmagazine.com/germany-sets-standard-for-bitcoin-regulation/
276 2013-08-19 16:52:53 <an3k> hehe
277 2013-08-19 16:52:55 <sipa> gmaxwell: or "I'm not trying to outrun the lion; I'm trying to outrun you." ?
278 2013-08-19 16:52:56 <Cusipzzz> ACTION stays behind Luke-Jr o the NSA threat board..
279 2013-08-19 16:53:19 <gmaxwell> sipa: Yep. :)
280 2013-08-19 16:55:44 <an3k> you heard what the time reported said? or what happened to the husband of Glenn Greenwald? Or what the NZ gov wants to do? Or that the USA still have nearly full control over eg. germany?
281 2013-08-19 16:57:13 <sipa> #bitcoin please
282 2013-08-19 16:57:27 <handle> s/oin/h/
283 2013-08-19 16:57:27 <handle> sorry
284 2013-08-19 17:01:35 <jrmithdobbs> lol
285 2013-08-19 17:01:52 <jrmithdobbs> Luke-Jr: really? 11:50 < Luke-Jr> an3k: maybe for the few people who still run Mac/Windows
286 2013-08-19 17:02:09 <sipa> in this channel, perhaps :)
287 2013-08-19 17:21:59 <petertodd> Luke-Jr: I did, although I turned the server off as the person who was using it stopped.
288 2013-08-19 17:22:27 <petertodd> Luke-Jr: I mean, I can turn it back on for you, which wouldn't be bloating the blockchain at all :/
289 2013-08-19 17:23:59 <petertodd> Luke-Jr: git is interesting re: timestamps - you can use a git commit hash to timestamp any individual file, but you have to reveal parts of the directory structure of your repo to do it!
290 2013-08-19 17:26:28 <an3k> gmaxwell: any idea what's going on with my client?
291 2013-08-19 17:26:47 <Luke-Jr> petertodd: true
292 2013-08-19 17:26:56 <Luke-Jr> petertodd: how would turning it back on not bloat the blockchain?
293 2013-08-19 17:28:59 <petertodd> Luke-Jr: I mean, it's O(1) bloat, but if there is one user, it's indistinguishable from that user just doing timestamping tx's
294 2013-08-19 17:29:24 <Luke-Jr> rather not then
295 2013-08-19 17:30:40 <petertodd> heh, you could do worse: change eligius to add timestamp digests to people's scriptSigs...
296 2013-08-19 17:30:55 <Luke-Jr> >_<
297 2013-08-19 17:31:04 <petertodd> although you can already do that with multisig tx's even if you aren't a miner
298 2013-08-19 17:31:22 <handle> what's the worst possible thing you could add to scriptsigs?
299 2013-08-19 17:31:38 <petertodd> handle: child porn?
300 2013-08-19 17:31:48 <handle> eh, I didn't mean like that
301 2013-08-19 17:31:50 <Luke-Jr> handle: you had to ask
302 2013-08-19 17:31:51 <handle> I meant computationally
303 2013-08-19 17:31:58 <handle> lol
304 2013-08-19 17:32:02 <Luke-Jr> handle: oh, that CVE???
305 2013-08-19 17:32:19 <Luke-Jr> https://en.bitcoin.it/wiki/CVEs#CVE-2013-2292
306 2013-08-19 17:32:31 <Luke-Jr> you'd lock up the whole network for a few minutes
307 2013-08-19 17:34:56 <handle> Luke-Jr: interesting idea
308 2013-08-19 17:35:00 <handle> I assume it's fixed now?
309 2013-08-19 17:35:03 <Luke-Jr> nope
310 2013-08-19 17:35:07 <handle> o.O
311 2013-08-19 17:35:13 <handle> brb making mining pool
312 2013-08-19 17:35:16 <Luke-Jr> ???
313 2013-08-19 17:35:19 <handle> :P
314 2013-08-19 17:35:31 <handle> I'm only kidding of course - still though, interesting that they could do that
315 2013-08-19 17:36:10 <gmaxwell> an3k: I think when trying to forward the port you managed to remap your outbound connections back to yourself!
316 2013-08-19 17:36:17 <Luke-Jr> LOL
317 2013-08-19 17:36:23 <gmaxwell> an3k: did you setup a port forward manually, or is this some kind of really awesome UPNP bug.
318 2013-08-19 17:36:38 <gmaxwell> an3k: sorry I took so long to get back to you, my email isn't always super fast.
319 2013-08-19 17:36:45 <an3k> np
320 2013-08-19 17:36:48 <an3k> its uPNP
321 2013-08-19 17:37:07 <an3k> but only the bitcoin-qt has a problem
322 2013-08-19 17:37:16 <gmaxwell> hm. I think your UPNP router may have some kind of crazy bug.
323 2013-08-19 17:37:38 <gmaxwell> an3k: can you turn of UPNP in bitcoin-qt and .. uh. I'm not sure how to clear the UPNP state out of your router without waiting an hour.
324 2013-08-19 17:38:02 <an3k> np, the router is fully managable :)
325 2013-08-19 17:38:14 <gmaxwell> In any case, what is happening is that DNS seed works normally, and then every single connection you make outbound ends up connecting back to you.
326 2013-08-19 17:38:18 <gmaxwell> Also, what router is this?!
327 2013-08-19 17:38:27 <an3k> Cisco RV082
328 2013-08-19 17:38:51 <Luke-Jr> figures
329 2013-08-19 17:38:53 <Luke-Jr> Cisco crap
330 2013-08-19 17:39:07 <handle> Cisco probably mines LTC
331 2013-08-19 17:39:12 <Luke-Jr> lol
332 2013-08-19 17:39:40 <an3k> turning upnp off in bitcoin removed the upnp entry in the router
333 2013-08-19 17:40:01 <an3k> and now i'm getting blocks very fast
334 2013-08-19 17:40:30 <an3k> 100% cpu usage :D
335 2013-08-19 17:40:51 <gmaxwell> Victory. Crazy. Now I wonder if other people have expirenced this and if there is anything we can do about it.
336 2013-08-19 17:41:09 <gmaxwell> Perhaps we're actually using UPNP wrong and every other device is smart enough to not believe our request?
337 2013-08-19 17:41:27 <an3k> I heard from someone about the same problem but don't know if he was using upnp or if his problem is caused by something else
338 2013-08-19 17:43:06 <an3k> well, the port getting mapped is 8333 and that is used to connect to nodes but when the connection to that node is established another connection to "itself" is created onto port 58663. But because of the fact that 10.0.0.1 is the router and not the computer the client runs on i think the client is doing something wrong
339 2013-08-19 17:44:20 <gmaxwell> It connects out from localhost:random to internet:8333 and then the router captures that and rewrites it to router:random -> localhost:8333
340 2013-08-19 17:44:34 <gmaxwell> r = UPNP_AddPortMapping(urls.controlURL, data.first.servicetype,
341 2013-08-19 17:44:34 <gmaxwell> the operative call to our UPNP library is:
342 2013-08-19 17:44:38 <gmaxwell> port.c_str(), port.c_str(), lanaddr, strDesc.c_str(), "TCP", 0, "0");
343 2013-08-19 17:44:39 <Luke-Jr> gmaxwell: could the router be redirecting DNS to its own IP?
344 2013-08-19 17:45:03 <gmaxwell> Luke-Jr: no, we're actually trying to connect to a bunch of internet addresses, not the router.
345 2013-08-19 17:45:05 <an3k> yes but it also connects from localhost:random to router:58663 thinking it would be localhost:58663
346 2013-08-19 17:45:33 <an3k> accepted connection 10.0.0.1:58663
347 2013-08-19 17:45:33 <an3k> trying connection 66.41.226.2:8333 lastseen=2.2hrs
348 2013-08-19 17:45:40 <gmaxwell> e.g.
349 2013-08-19 17:45:41 <gmaxwell> connected 72.179.167.123:8333
350 2013-08-19 17:45:41 <gmaxwell> send version message: version 70001, blocks=249565, us=37.4.17.39:8333, them=72.179.167.123:8333, peer=72.179.167.123:8333
351 2013-08-19 17:45:41 <gmaxwell> trying connection 72.179.167.123:8333 lastseen=505.1hrs
352 2013-08-19 17:45:44 <gmaxwell> accepted connection 10.0.0.1:58683
353 2013-08-19 17:45:47 <gmaxwell> connected to self at 10.0.0.1:58683, disconnecting
354 2013-08-19 17:45:56 <an3k> exactly
355 2013-08-19 17:46:02 <gmaxwell> just some random example data I cooked up.
356 2013-08-19 17:46:04 <an3k> connected to self at 10.0.0.1:58683, disconnecting
357 2013-08-19 17:46:14 <an3k> 10.0.0.1 is NOT "self"
358 2013-08-19 17:46:40 <Luke-Jr> it is, that's the point
359 2013-08-19 17:46:43 <an3k> why does bitcoin-qt thinks 10.0.0.1 is itself? i think fixing this will solve the whole problem
360 2013-08-19 17:46:53 <gmaxwell> an3k: because it is.
361 2013-08-19 17:46:55 <an3k> Luke-Jr: it's not
362 2013-08-19 17:46:55 <Luke-Jr> it's talking to itself
363 2013-08-19 17:47:06 <an3k> nah, how can it talk to itself using ip 10.0.0.1?
364 2013-08-19 17:47:13 <Luke-Jr> your router is taking the outgoing connection and forwarding it back to you
365 2013-08-19 17:47:16 <gmaxwell> an3k: it knows its talking to itself. It sends an identification cookie and sees its own traffic.
366 2013-08-19 17:47:46 <gmaxwell> the router is acting as a loopback, likely because its UPNP is directing _all_ connections to 8333 back to your host. E.g. it's working "too well".
367 2013-08-19 17:48:43 <an3k> ok but without upnp activated the log looks different. there is no second connection to 10.0.0.1
368 2013-08-19 17:49:25 <an3k> and the second connection to itself is NOT working because there is no mapping for that port, perhaps because the client didn't created it
369 2013-08-19 17:49:25 <Luke-Jr> never was
370 2013-08-19 17:49:30 <Luke-Jr> it was a connection FROM 10.0.0.1
371 2013-08-19 17:49:49 <sipa> indeed
372 2013-08-19 17:50:00 <sipa> it's your own connection coming back
373 2013-08-19 17:50:41 <phantomcircuit> gmaxwell, from here upnp appears to work correctly
374 2013-08-19 17:50:58 <an3k> I checked the router logs. absolutely nothing about port 58663
375 2013-08-19 17:51:44 <an3k> but I can see successful mappings of port 8333
376 2013-08-19 17:51:47 <sipa> i doubt it logs its own natted source port fir forwarded connections
377 2013-08-19 17:51:51 <an3k> it does
378 2013-08-19 17:52:07 <sipa> k
379 2013-08-19 17:52:25 <phantomcircuit> sipa, most modern routers log all connections forwarded from the outside
380 2013-08-19 17:52:56 <sipa> well, there is no doubt that you connected to yourself
381 2013-08-19 17:53:03 <sipa> however that happened
382 2013-08-19 17:53:25 <gmaxwell> It's unambigious whats happening here, the only question is if your router's UPNP is bugged or ours is.
383 2013-08-19 17:53:32 <phantomcircuit> there's a lot of buggy upnp implementations
384 2013-08-19 17:53:58 <phantomcircuit> gmaxwell, i can verify upnp works correctly with bitcoin-qt on windows and a netgear WNDR4500
385 2013-08-19 17:54:05 <gmaxwell> I'd say that the fact that it works for lots of people suggests that its on your end, but since a lot of firewalls can't actually hairpin a connection like that at all, I'm not sure.
386 2013-08-19 17:54:11 <an3k> gmaxwell: in reply to the code you posted above: Aug 19 19:33:11 2013\tKernel\t upnpd[415]: AddPortMap: DevUDN: uuid:75802409-bccb-40e7-8e6c-fa095ecce13e ServiceID: urn:upnp-org:serviceId:WANIPConn1 RemoteHost: (null) Prot: TCP ExtPort: 8333 Int: 10.0.0.31.8333
387 2013-08-19 17:54:33 <sipa> and 10..31 is your internal ip?
388 2013-08-19 17:54:39 <an3k> yes
389 2013-08-19 17:54:45 <sipa> looka correct so far
390 2013-08-19 17:54:58 <gmaxwell> the fact that RemoteHost: (null) is interesting perhaps.
391 2013-08-19 17:55:12 <an3k> well, i never had any upnp problem. even sabnzbd uses it correctly
392 2013-08-19 17:55:20 <sipa> i assume it means "any"
393 2013-08-19 17:55:23 <phantomcircuit> gmaxwell, upnp allows for specifying a specific remote host
394 2013-08-19 17:55:30 <phantomcircuit> i assume it's 0.0.0.0
395 2013-08-19 17:55:52 <sipa> an3k: could you post a log entry for sabnzbdb requesting a mapping?
396 2013-08-19 17:56:05 <gmaxwell> phantomcircuit: right, but is there something special you need to do to distingush "from the internet" from "all connections, including my own"
397 2013-08-19 17:56:12 <gmaxwell> sipa: good idea.
398 2013-08-19 17:56:31 <phantomcircuit> gmaxwell, i suspect you'd have to connect to external_ip:8333 to end up with a loop back
399 2013-08-19 17:56:36 <an3k> sipa: sure, one sec
400 2013-08-19 17:57:02 <phantomcircuit> it's possible that's being incorrectly implemented as an iptables rule which doesn't specify the destination ip
401 2013-08-19 17:57:23 <phantomcircuit> but then outgoing traffic to port 8333 should all be trapped
402 2013-08-19 17:57:30 <phantomcircuit> an3k, what kind of router is this?
403 2013-08-19 17:57:37 <gmaxwell> phantomcircuit: yea, thats why I asked if he did it manually.
404 2013-08-19 17:58:05 <sipa> which would mean we beed an outgoing connection filtering rule that says "remote port != local listen port"
405 2013-08-19 17:58:06 <gmaxwell> Although, I actually don't think IPtables dnat can actually do this. (a direct hairpin, as that requires rewriting the source IP address too)
406 2013-08-19 17:59:10 <phantomcircuit> gmaxwell, iirc it can rewrite the source ip also
407 2013-08-19 17:59:37 <phantomcircuit> interesting
408 2013-08-19 17:59:45 <gmaxwell> (also, as an aside if it's always rewriting the source IP even for non-hairpin connections, that will totally bugger our DoS stuff)
409 2013-08-19 17:59:46 <phantomcircuit> this client w/ upnp is now reporting 12 connections
410 2013-08-19 18:00:18 <gmaxwell> phantomcircuit: perhaps, but you don't get a rule that does that accidentally.
411 2013-08-19 18:09:21 <an3k> have to restart NAS to get it to use upnp ... brb
412 2013-08-19 18:14:46 <jrmithdobbs> phantomcircuit: upnp works on anything running miniupnpd which is like 95% of home routers at this point
413 2013-08-19 18:15:01 <jrmithdobbs> but miniupnpd has problems, lots of them =/
414 2013-08-19 18:31:12 <an3k> Aug 19 22:30:22 2013\tKernel\t upnpd[7504]: AddPortMap: DevUDN: uuid:75802409-bccb-40e7-8e6c-fa095ecce13e ServiceID: urn:upnp-org:serviceId:WANIPConn1 RemoteHost: (null) Prot: UDP ExtPort: 6881 Int: 10.0.0.21.6881
415 2013-08-19 18:31:55 <gmaxwell> well, that one is UDP not TCP but otherwise looks identical. Alas.
416 2013-08-19 18:32:24 <phantomcircuit> an3k, what kind of router is this?
417 2013-08-19 18:32:32 <an3k> Cisco RV082
418 2013-08-19 18:33:10 <sipa> it'd be interesting to see if a UDP packet to internet:6881 also comes back
419 2013-08-19 18:33:41 <phantomcircuit> an3k, is it running vpn stuff?
420 2013-08-19 18:34:01 <an3k> the router? no
421 2013-08-19 18:34:07 <an3k> sipa: how? :)
422 2013-08-19 18:34:22 <phantomcircuit> that router is a "Dual WAN VPN Router"
423 2013-08-19 18:34:31 <phantomcircuit> so im guessing it does weird stuff to make that work
424 2013-08-19 18:34:40 <phantomcircuit> which probably doesn't interact well with upnp
425 2013-08-19 18:34:48 <an3k> phantomcircuit exactly and it's configured to use dual wan, not wan1 as primary and wan2 as backup
426 2013-08-19 18:35:05 <an3k> nah, it does work well with upnp. just bitcoin-qt has some problems
427 2013-08-19 19:06:32 <dansmithbtc2> Hi, are there any plans to mitigate the crippling of bitcoin network with a difficulty pumping and dumping attack? What will it take to make the difficulty change not every 2016 blocks but acc.to some other algorithm?
428 2013-08-19 19:07:19 <sipa> realistically: an alt coin
429 2013-08-19 19:08:50 <dansmithbtc2> How about a simple change in the protocol, like - if no blocks found within 2 hours, then readjust the difficulty?
430 2013-08-19 19:09:12 <gmaxwell> dansmithbtc2: creates trivial vulnerabilties.
431 2013-08-19 19:09:49 <gmaxwell> And uneeded. If the network has failed due to that, it's probably no longer economically interesting as the same party could reverse and respend at will or DOS with empty/junk blocks.
432 2013-08-19 19:10:14 <sipa> ;;tblb 7200
433 2013-08-19 19:10:15 <gribble> The expected time between blocks taking 2 hours and 0 seconds to generate is 120 years, 42 weeks, 6 days, 23 hours, 12 minutes, and 18 seconds
434 2013-08-19 19:10:41 <gmaxwell> (not only creates trivial vulnerabilities, but in fact altcoins have adopted exactly that and been exploited as a result!)
435 2013-08-19 19:11:29 <gmaxwell> dansmithbtc2: an example vulnerability, ??? bitcoin is intended to be fairly robust against a network attacker. I partition your node for more than two hours (connecting it only to my sybil nodes), and now I can feed you low diffiuclty blocks and you think you're getting real confirmations.
436 2013-08-19 19:19:57 <Luke-Jr> maybe we should display confirmation in total work with some exponential unit?
437 2013-08-19 19:20:36 <petertodd> Luke-Jr: heh, I nearly read that as "in tonal work"...
438 2013-08-19 19:20:38 <gmaxwell> Anyone have a semi-recent signature R value dump handy? it takes forever to run, TD suggests it might be interesting here https://bitcointalk.org/index.php?topic=277595.0
439 2013-08-19 19:21:08 <gmaxwell> Luke-Jr: sipa suggested displaying time-at-current-hashrate (or, perhaps, peak hashrate)
440 2013-08-19 19:21:09 <petertodd> Luke-Jr: Sounds reasonable, though probably sub-exponential is fine from a UI point of view - like first a tiny sliver of the circle, then a bigger chunk, and an even bigger chunk etc.
441 2013-08-19 19:21:41 <gmaxwell> sadly the entirely history of bitcoin is only a couple months at the current difficulty!
442 2013-08-19 19:21:54 <k9quaint> not sad
443 2013-08-19 19:22:06 <gmaxwell> well good news for the future!
444 2013-08-19 19:22:07 <petertodd> Luke-Jr: vessenes was talking about doing a report on confirmations with regard to cost to an attacker and stuff like that FWIW
445 2013-08-19 19:22:20 <Luke-Jr> petertodd: well, I mean so users don't see 293284914234242 after a block because difficulty is high :p
446 2013-08-19 19:22:24 <gmaxwell> petertodd: I ran numbers on that, and didn't like the results.
447 2013-08-19 19:23:01 <petertodd> gmaxwell: I'm not surprised - has he talked to you yet?
448 2013-08-19 19:23:30 <gmaxwell> petertodd: e.g. the conservative thing to do would be to assume 100% of the transaction in the block with your txn are double spends, and ask how many confirms you need before the attacker with a given hashrate has a negative expectency.
449 2013-08-19 19:23:57 <gmaxwell> (and take the hasrate of the largest pool you know about)
450 2013-08-19 19:24:23 <petertodd> gmaxwell: I know - a really sophisticated hacker can get a heck of a lot of value out of a malicious block - in part you are relying on the good security of others in your own security assumptions.
451 2013-08-19 19:24:29 <sipa> the genesis block has 93 days worth of confirmations...
452 2013-08-19 19:24:51 <gmaxwell> and the result (when I checked it) came out to be like 20 blocks or something, of course more when there is more transaction value.
453 2013-08-19 19:24:54 <Luke-Jr> gmaxwell: and assume any other largeish pool is DDoS'd
454 2013-08-19 19:25:07 <gmaxwell> Luke-Jr: I wasn't assuming that, but thats a point too... :(
455 2013-08-19 19:25:35 <petertodd> Luke-Jr: yeah, might as well assume they all are then :(
456 2013-08-19 19:25:41 <gmaxwell> if you assume that, then uh.. I think nothing is safe except waiting for long enough that you'd hear of the DOS.
457 2013-08-19 19:26:01 <Luke-Jr> >_<
458 2013-08-19 19:26:10 <phantomcircuit> gmaxwell, for most things where high certainty of transactions is necessary the difference between a single high value tx and lots of low value tx's is irrelevant
459 2013-08-19 19:26:17 <petertodd> Well, you can look at the rate of new blocks, and make assumptions about how much hashing power the attacker might have, but that's ugly fast...
460 2013-08-19 19:26:53 <Luke-Jr> once upon a time I had a pullrequest working toward disabling confirmations if the network hashrate drops significantly..
461 2013-08-19 19:27:07 <gmaxwell> phantomcircuit: the point there is that if you are conservative the value of your txn doesn't matter for the attacker's profit. Because an attacker can fill a block with attack transactions, and attack many parties at once.
462 2013-08-19 19:27:35 <phantomcircuit> gmaxwell, right
463 2013-08-19 19:27:49 <petertodd> Luke-Jr: IMO a good thing - merchants should be doing stuff like that for sure even at the cost of occasional outages due to bad mining luck
464 2013-08-19 19:28:02 <gmaxwell> realistically they likely kind find enough victims, but I was trying to a conservative figure... hoping it would produce a number like ... 6. :)
465 2013-08-19 19:28:02 <petertodd> Luke-Jr: but that's all business logic really...
466 2013-08-19 19:28:26 <gmaxwell> MTGOX deposits are pretty clever in a what that causes occasional delays.
467 2013-08-19 19:28:34 <gmaxwell> They look for six confirms _longer_ than any chain they've seen.
468 2013-08-19 19:28:43 <gmaxwell> so if there is a short fork, they need more confirms.
469 2013-08-19 19:29:25 <Luke-Jr> gmaxwell: meh, if someone were doing that kind of attack, they wouldn't publish it until it was done
470 2013-08-19 19:29:28 <gmaxwell> though really the number six is pure cargo cult. With single pools kissing 40% hashrate, there is no rational basis for the number six anymore.
471 2013-08-19 19:29:35 <CodeShark> gmaxwell, that's how I measured confirmation counts in one implementation I did
472 2013-08-19 19:30:02 <gmaxwell> Luke-Jr: it increases their security slightly against doublespends that happen in the case of things like bug induced major forks.
473 2013-08-19 19:30:15 <petertodd> gmaxwell: That's exactly the kind of stuff that I think vessenes wants written up, ideally made into libraries for people to attack to their business logic. But that's lots of gruntwork...
474 2013-08-19 19:37:06 <CodeShark> the thing is that these types of differences in confirmation metrics don't really matter the vast majority of the time - but when they do make a difference, it can be a big difference
475 2013-08-19 19:38:56 <gmaxwell> the mtgox metric is snazzy but causes randomly increased delays and confirmation non-determinism on a regular basis though. :(
476 2013-08-19 19:39:09 <gmaxwell> e.g. many times a day your required confirm count is increased by 1.
477 2013-08-19 19:40:00 <petertodd> sounds like the rest of mt. gox re: payment non-determinism...
478 2013-08-19 19:40:21 <CodeShark> if it's any consolation, their dollar deposit/withdrawal is even less deterministic and presents far, far,
479 2013-08-19 19:40:28 <CodeShark> FAAAR longer delays
480 2013-08-19 19:40:50 <CodeShark> I've been waiting a month for a withdrawal
481 2013-08-19 19:41:12 <CodeShark> reminded me never to use Gox for that again
482 2013-08-19 19:50:10 <helo> yeah, gox is on its way out if it doesn't fix that soon
483 2013-08-19 19:50:49 <phantomcircuit> helo, it's plausible that there isn't a plausible fix
484 2013-08-19 19:51:01 <phantomcircuit> iirc they're literally filling out paper forms for wires
485 2013-08-19 19:51:18 <phantomcircuit> short of becoming a bank im not sure how to get around that
486 2013-08-19 19:51:27 <helo> 15% depreciation of mtgoxusd is pretty serious
487 2013-08-19 19:51:42 <phantomcircuit> sure but for now it's allowing be to make a killing
488 2013-08-19 19:51:44 <phantomcircuit> woot woot
489 2013-08-19 19:52:15 <helo> hopefully all exchanges don't go this way after they get popular
490 2013-08-19 19:52:51 <phantomcircuit> helo, afaict the problem is without dwolla the wire volume on gox exploded with people trying to transfer like $500 around
491 2013-08-19 19:53:00 <phantomcircuit> which caused their bank problems
492 2013-08-19 19:53:07 <phantomcircuit> which means their bank makes it their problem
493 2013-08-19 19:55:23 <Luke-Jr> phantomcircuit: they should have required the full id nonsense for wires
494 2013-08-19 19:55:46 <Luke-Jr> now they're not only pissing off/losing Dwolla customers, but also annoying those of us who had been using wires without problems for a long time
495 2013-08-19 19:56:50 <Luke-Jr> or at least prioritised larger amounts
496 2013-08-19 19:57:19 <Luke-Jr> otoh, I don't know if my $10k is on the high or low end for MtGox..
497 2013-08-19 20:03:07 <phantomcircuit> Luke-Jr, that's not even enough for the us though
498 2013-08-19 20:03:20 <Luke-Jr> phantomcircuit: ?
499 2013-08-19 20:03:25 <phantomcircuit> the states require you be licensed regardless of whether you follow the other rules
500 2013-08-19 20:03:40 <phantomcircuit> afaict they are currently in the process of registering
501 2013-08-19 20:03:42 <phantomcircuit> but it takes years
502 2013-08-19 20:08:37 <phantomcircuit> lol
503 2013-08-19 20:08:45 <phantomcircuit> the progress bar is depressingly accurate
504 2013-08-19 20:08:54 <phantomcircuit> 19 weeks behind, maybe 30% full
505 2013-08-19 20:29:33 <phantomcircuit> 2013-08-19 22:28:52 ERROR: CTransaction::CheckTransaction() : vin empty
506 2013-08-19 20:29:33 <phantomcircuit> 2013-08-19 22:28:52 ERROR: CTxMemPool::accept() : CheckTransaction failed
507 2013-08-19 20:29:33 <phantomcircuit> 2013-08-19 22:28:52 Misbehaving: 65.207.52.66:64744 (30 -> 40)
508 2013-08-19 20:29:34 <phantomcircuit> wat
509 2013-08-19 20:29:42 <phantomcircuit> what are you doing 65.207.52.66
510 2013-08-19 20:29:46 <phantomcircuit> go home you're drunk
511 2013-08-19 20:30:18 <gmaxwell> it sent a transaction 00000000 nodes that do that appear to do it because they somehow end up with it in their wallet.
512 2013-08-19 20:30:20 <gmaxwell> I don't think we know why.
513 2013-08-19 20:30:37 <gmaxwell> but its apparently happened with unmodified reference client nodes.
514 2013-08-19 20:31:56 <phantomcircuit> NORTH SHORE HOSPITALITY ASSOC LP
515 2013-08-19 20:32:04 <phantomcircuit> someone on hotel wifi doing something weird?
516 2013-08-19 20:32:49 <phantomcircuit> i literally think i could walk into that hotel if i was in pittsburgh
517 2013-08-19 20:33:05 <phantomcircuit> lol ip is lol
518 2013-08-19 20:38:50 <rubino123> *?
519 2013-08-19 20:38:50 <rubino123> 8?
520 2013-08-19 20:38:50 <rubino123> how many confirmations before the coins are useable?
521 2013-08-19 20:39:56 <Luke-Jr> rubino123: what coins?
522 2013-08-19 20:40:07 <Luke-Jr> and usable from whose perspective?
523 2013-08-19 20:40:40 <sipa> for the protocol, 0
524 2013-08-19 20:40:41 <rubino123> useable so I can transfer them from one account on the wallet to another
525 2013-08-19 20:40:48 <rubino123> account
526 2013-08-19 20:40:51 <sipa> for most clients, 1
527 2013-08-19 20:41:01 <sipa> for some merchant accounts, 6
528 2013-08-19 20:41:06 <Luke-Jr> rubino123: 0
529 2013-08-19 20:41:18 <Luke-Jr> rubino123: accounts are just beancounters, they don't interact with coins
530 2013-08-19 20:41:29 <rubino123> Luke-Jr: is it possible to turn off the computer networkinterface and use the bitcoind daemon
531 2013-08-19 20:41:39 <rubino123> to make internal shuffles
532 2013-08-19 20:41:42 <sipa> yes
533 2013-08-19 20:41:43 <Luke-Jr> sure
534 2013-08-19 20:42:15 <sipa> not anything that needs a tramsaction though
535 2013-08-19 20:42:24 <sipa> so you can't move coins around
536 2013-08-19 20:42:40 <Luke-Jr> unless you do a bunch of complex stuff with the raw transaction API
537 2013-08-19 20:42:41 <sipa> but you can create new addresses, or do moves between accounts in a wallet
538 2013-08-19 20:43:40 <rubino123> if I move x bitcoins from account A to account B and then the network gets brought back online will new coin that get sent to account A get reconciled into acocunt A?
539 2013-08-19 20:43:55 <sipa> accounts _only_ exist locally
540 2013-08-19 20:43:57 <rubino123> My guess is no but better to be safe ask and not get it wrong
541 2013-08-19 20:44:05 <sipa> the have _nothing_ to do with coins or addresses
542 2013-08-19 20:44:24 <sipa> they are really just numbers
543 2013-08-19 20:44:34 <sipa> in a shared wallet
544 2013-08-19 20:44:45 <sipa> all accounts share all coins in a wallet
545 2013-08-19 20:45:03 <Luke-Jr> rubino123: coins received on an address, go into the account that address credits.
546 2013-08-19 20:45:26 <Luke-Jr> rubino123: moving balances between accounts has no effect on address->account associations
547 2013-08-19 20:45:32 <sipa> s/go into/are assigned to/
548 2013-08-19 20:45:50 <sipa> it's not that the received coins belong to that account
549 2013-08-19 20:45:52 <Luke-Jr> right, the coins don't go into the account
550 2013-08-19 20:45:58 <sipa> coins belong to the wallet as a whole
551 2013-08-19 20:46:02 <Luke-Jr> it JUST credits the account
552 2013-08-19 20:46:08 <sipa> indeed
553 2013-08-19 20:46:09 <rubino123> so here is the use use case: the provider is running a service that requires payment by a certain time for the service to begin. At the predetermined time the provider turns off the network connection, moves all coins from the reception account into another account.
554 2013-08-19 20:46:34 <Luke-Jr> you don't move coins, you move balances.
555 2013-08-19 20:46:40 <sipa> coins do not belong to accounts
556 2013-08-19 20:46:44 <rubino123> The provider then turns the connection back on and any coins being sent to the recpetion account at this time get sent back
557 2013-08-19 20:47:00 <sipa> are you talking about accounts or addresses?
558 2013-08-19 20:47:11 <Luke-Jr> rubino123: that would work, IFF you collect a "send back" address beforehand
559 2013-08-19 20:47:31 <Luke-Jr> rubino123: you don't even need to turn the connection off
560 2013-08-19 20:47:41 <rubino123> it would be aggreed that the sending addressis the send back address
561 2013-08-19 20:47:48 <sipa> bah
562 2013-08-19 20:47:49 <Luke-Jr> rubino123: there is no sending address
563 2013-08-19 20:47:52 <sipa> don't do that
564 2013-08-19 20:47:54 <Luke-Jr> transactions only have a destination
565 2013-08-19 20:47:56 <rubino123> I am trying to get away with not using a db for user account balances
566 2013-08-19 20:48:08 <rubino123> for the provider
567 2013-08-19 20:48:14 <sipa> if you need to send coins back, ask for a refund address
568 2013-08-19 20:48:20 <sipa> please