1 2011-05-22 00:01:00 <diki> luke you there?
 2 2011-05-22 00:01:07 <stuhood> phantomcircuit: actually, the client only connects to nodes that don't share the first 2 octets of the ip address
 3 2011-05-22 00:01:31 <phantomcircuit> stuhood, so what? getting a ton of ip's is (relatively) trivial
 4 2011-05-22 00:02:32 <stuhood> phantomcircuit: yes. i'm just wondering how much of an improvement the xor metric would even be over the existing approach
 5 2011-05-22 00:03:39 <phantomcircuit> not much
 6 2011-05-22 00:04:06 <stuhood> if an attacker can get ip addresses that are distributed throughout the address space, there isn't much you can do
 7 2011-05-22 00:04:07 <phantomcircuit> a combination of "hot" peers and known good ndoes is necessary
 8 2011-05-22 00:04:23 <stuhood> yea
 9 2011-05-22 00:04:29 <phantomcircuit> stuhood, actually you can make a sybil attack pretty much impossible
10 2011-05-22 00:04:34 <phantomcircuit> think about it
11 2011-05-22 00:04:44 <phantomcircuit> if you get a block/transaction anytime within a 10 minute window you're fine
12 2011-05-22 00:04:47 <stuhood> well, that's the point of proof of work
13 2011-05-22 00:05:10 <phantomcircuit> controlling enough of the network to keep a peer from finding out something for 10 minutes when it's hopping every 30 seconds would be quite an achievement
14 2011-05-22 00:05:32 <stuhood> sorry, what's the hopping? still learning about the protocol
15 2011-05-22 00:05:42 <phantomcircuit> no client actually does this yet
16 2011-05-22 00:05:54 <phantomcircuit> the current network could have a dedicated attacker run a sybil attack
17 2011-05-22 00:06:02 <phantomcircuit> however that would cost more than it would be worth currently
18 2011-05-22 00:06:05 <phantomcircuit> so i doubt anybody will
19 2011-05-22 00:06:52 <stuhood> improving the distance metric and the number of connections might be a good idea then
20 2011-05-22 00:07:17 <phantomcircuit> not to mention removing the connection limit counts *cough*
21 2011-05-22 00:07:17 <stuhood> connect to at least 32 nodes at varying distances
22 2011-05-22 00:07:21 <stuhood> yea
23 2011-05-22 00:08:30 <XX01XX> How are you calculating distance?
24 2011-05-22 00:08:45 <stuhood> XX01XX: was thinking of kademlia's xor distance
25 2011-05-22 00:09:13 <stuhood> http://en.wikipedia.org/wiki/Kademlia#Accelerated_lookups
26 2011-05-22 00:13:34 <gmaxwell> to bad there is such as big cpu/gpu gap for the sha512 pow... you could have nodes generate expensive ID's used to peering decisions, but attackers with GPUs could generate lots of any level of pow that would be acceptable for CPUs.
27 2011-05-22 00:14:39 <stuhood> yea...
28 2011-05-22 00:15:09 <diki> zomg...i never knew pushpool was so complicated...
29 2011-05-22 00:15:15 <diki> *complex
30 2011-05-22 00:15:34 <diki> One function gets called in another function, which get's called in another function....
31 2011-05-22 00:15:39 <diki> and it keeps going and going
32 2011-05-22 00:15:46 <diki> until we reach main
33 2011-05-22 00:15:57 <XX01XX> Yo dawg...
34 2011-05-22 00:15:59 <diki> zomg, how does jgarzik follow all that
35 2011-05-22 00:16:39 <gmaxwell> "Man, you'd have to be a kernel developer or something to follow that shit"
36 2011-05-22 00:16:51 <lfm> um, I think that would be normal software
37 2011-05-22 00:17:01 <gmaxwell> diki: I didn't think it was that complicated. Not well commented however.
38 2011-05-22 00:17:16 <ArtForz> functions calling functions, the horror!
39 2011-05-22 00:17:24 <gmaxwell> And it does some dumb shit, like blocking getworks if the database takes a long time.
40 2011-05-22 00:18:03 <gmaxwell> sure as hell beats writing your own though.
41 2011-05-22 00:18:32 <diki> Inside main there is a function which calls another function, which in return calls another function and etc going over to submit_work...
42 2011-05-22 00:18:38 <diki> that was like 10 functions or so
43 2011-05-22 00:19:37 <ArtForz> way too simple
44 2011-05-22 00:20:11 <diki> way too complex
45 2011-05-22 00:20:14 <ArtForz> needs at least some recursion, a few calls to functions living in headers and a few complex macros
46 2011-05-22 00:20:33 <XX01XX> diki... can make things easier to maintain once you understand it
47 2011-05-22 00:20:37 <ArtForz> also, needs more cowbell.
48 2011-05-22 00:20:40 <gmaxwell> yea, a linked list of structs of function pointers.
49 2011-05-22 00:20:40 <lfm> how bout some libs with callbacks thrown in
50 2011-05-22 00:20:44 <diki> i am not insulting jgarzik here, i am in fact praising him
51 2011-05-22 00:20:56 <diki> following all that...you have to be a monster
52 2011-05-22 00:21:16 <lfm> you may call us monsters inc