1 2010-10-27 00:00:01 <brocktice> Works pretty well for our scientific applications...
  2 2010-10-27 00:01:21 <ArtForz> theoretical throughput for a GTX480: 101Mhash/s, HD5870: 313Mhash/s
  3 2010-10-27 00:02:20 <ArtForz> and for extra fun the 480 uses about 30% more power
  4 2010-10-27 00:02:33 <brocktice> I've only got a GTS250 and some other lower end card
  5 2010-10-27 00:02:44 <brocktice> 9500 GT
  6 2010-10-27 00:02:51 <ArtForz> thats gonna be quite slow
  7 2010-10-27 00:02:59 <Diablo-D3> 480 costs more, uses more power, and is a third of the opencl performance
  8 2010-10-27 00:03:01 <Diablo-D3> EPIC FAIL
  9 2010-10-27 00:03:08 <ArtForz> 9800GT is ~ 25Mh/s peak
 10 2010-10-27 00:03:13 <ArtForz> 9400GT ... a lot less
 11 2010-10-27 00:03:23 <ArtForz> errr.. 9500
 12 2010-10-27 00:03:26 <Diablo-D3> 480 and 5870 perform the same on 3D
 13 2010-10-27 00:03:43 <brocktice> Yeah I bought the 9500 GT before I started using CUDA.
 14 2010-10-27 00:03:51 <brocktice> Then we did that, plus I wanted a 4-display setup
 15 2010-10-27 00:03:55 <Diablo-D3> brocktice: CUDA is dead
 16 2010-10-27 00:03:56 <brocktice> so, added the GTS250
 17 2010-10-27 00:03:58 <Diablo-D3> its nvidia hardware only
 18 2010-10-27 00:04:05 <ArtForz> yep, for integer stuff ATI blows nvidia completely out of the water
 19 2010-10-27 00:04:05 <Diablo-D3> and its not as well designed or supported as opencl
 20 2010-10-27 00:04:16 <Diablo-D3> and nvidia is going to go bankrupt anyhow
 21 2010-10-27 00:04:21 <brocktice> It's pretty well supported in HPC/scientific computing right now.
 22 2010-10-27 00:04:25 <brocktice> So that's what we're using.
 23 2010-10-27 00:04:31 <Diablo-D3> brocktice: only because nvidia pays them off
 24 2010-10-27 00:04:37 <Diablo-D3> ATI supports _standards_
 25 2010-10-27 00:04:40 <Diablo-D3> nvidia supports retardation
 26 2010-10-27 00:04:49 <Diablo-D3> its obvious which company I'd go with
 27 2010-10-27 00:04:52 <ArtForz> "millions of flies can't be wrong" ;)
 28 2010-10-27 00:05:07 <brocktice> Well, when ATI can reliably run my displays with acceleration, in 64-bit Linux, I'll consider switching.
 29 2010-10-27 00:05:22 <brocktice> Every time I buy an ATI card I end up blowing two days troubleshooting and then return the damn thing.
 30 2010-10-27 00:06:07 <brocktice> I'm not saying you're wrong, only that until recently the alternative wasn't really tenable.
 31 2010-10-27 00:06:14 <ArtForz> *shrug* works more-or-less fine here
 32 2010-10-27 00:06:15 <brocktice> I'll be quite happy if that's changed (or changing)
 33 2010-10-27 00:06:52 <brocktice> Aaanyway, let me re-ask my question. Is there some way to do gpu accelerated hashing on Linux now?
 34 2010-10-27 00:07:09 <brocktice> That doesn't involve a BitCoin transaction followed by silence?
 35 2010-10-27 00:07:17 <ArtForz> yes
 36 2010-10-27 00:07:34 <ArtForz> there's open source CUDA and OpenCL miners
 37 2010-10-27 00:07:50 <Diablo-D3> brocktice: erm
 38 2010-10-27 00:07:55 <Diablo-D3> [10:05:07] <brocktice> Well, when ATI can reliably run my displays with acceleration, in 64-bit Linux, I'll consider switching.
 39 2010-10-27 00:07:58 <Diablo-D3> so 2+ years ago?
 40 2010-10-27 00:08:10 <brocktice> I tried not long ago (< 2 years) and that was not the case
 41 2010-10-27 00:08:14 <brocktice> not for multiple monitors
 42 2010-10-27 00:08:19 <brocktice> Oh yes, I remember, this spring.
 43 2010-10-27 00:08:23 <ArtForz> Diablo-D3: multihead + accel = eadache
 44 2010-10-27 00:08:36 <ArtForz> of course it doesn't work properly on nvidia either
 45 2010-10-27 00:08:37 <brocktice> Yeah, sorry, forgot that specification
 46 2010-10-27 00:08:43 <brocktice> It works a hell of a lot better
 47 2010-10-27 00:08:52 <brocktice> I was going to do this nifty CrossFireX thing, but no.
 48 2010-10-27 00:09:15 <brocktice> With NVidia I had to use FakeXinerama and then all was well.
 49 2010-10-27 00:09:26 <ArtForz> well, it somewhat works on nvidia if you use single-large-surface mode
 50 2010-10-27 00:09:35 <ArtForz> good luck with xinerama though
 51 2010-10-27 00:09:43 <brocktice> FakeXinerama did the trick
 52 2010-10-27 00:10:00 <ArtForz> now try it with one screen rotated by 90??? ;)
 53 2010-10-27 00:10:18 <brocktice> Yeah I don't hate myself that much.
 54 2010-10-27 00:10:24 <ArtForz> I finally gave up on that
 55 2010-10-27 00:10:26 <brocktice> took me long enough to get this setup working.
 56 2010-10-27 00:10:27 <Diablo-D3> brocktice: well, for all the people I know, fglrx works fine on multihead
 57 2010-10-27 00:10:37 <brocktice> For 2 heads, probably.
 58 2010-10-27 00:10:43 <Diablo-D3> and they've renewed even better support due to the 6 head stuff
 59 2010-10-27 00:10:51 <ArtForz> now running 1920x1200 center flanked by + 2 1600x1200
 60 2010-10-27 00:10:53 <brocktice> Yeah I hear the very newest stuff is better.
 61 2010-10-27 00:10:56 <Diablo-D3> brocktice: dont mistake hardware support for accel support
 62 2010-10-27 00:11:21 <brocktice> I've got 2 1600x1200 on the bottom and 2 1280x1024 on top
 63 2010-10-27 00:11:25 <Diablo-D3> older hardware has issue with accel on heads that arent on the first card in a crossfire solution
 64 2010-10-27 00:11:45 <brocktice> Yeah plus I was trying to make the most of the card on the motherboard, probably a mistake.
 65 2010-10-27 00:11:55 <Diablo-D3> you can
 66 2010-10-27 00:12:03 <Diablo-D3> it'll crossfire fine
 67 2010-10-27 00:12:07 <ArtForz> only problem is my side screens are CRTs, and trying to get fglrx to run em at 95Hz vrefresh is a very frustrating experience
 68 2010-10-27 00:12:19 <Diablo-D3> but you may or may not have issues trying to get good accel on the head plugged into the mobo
 69 2010-10-27 00:12:32 <ArtForz> especially as they provide flat out wrong EDID info :/
 70 2010-10-27 00:12:33 <brocktice> oh, yeah, one of my FPs was hooked up with d-sub originally because I was short a cable.
 71 2010-10-27 00:12:36 <brocktice> Boy was that a pain.
 72 2010-10-27 00:12:51 <Diablo-D3> brocktice: it happens in windows too
 73 2010-10-27 00:12:58 <Diablo-D3> and its pretty much tied to how the hardware works
 74 2010-10-27 00:13:00 <Diablo-D3> nvidia gets it too
 75 2010-10-27 00:13:13 <Diablo-D3> heads on SLI slaves dont entirely totally 100% function right
 76 2010-10-27 00:13:32 <brocktice> Right, I'm not doing SLI
 77 2010-10-27 00:13:42 <Diablo-D3> shouldnt be an issue for 2 heads.
 78 2010-10-27 00:13:43 <brocktice> Apparently Matrox makes some boxes that do what I want.
 79 2010-10-27 00:13:45 <Diablo-D3> since thats all you have
 80 2010-10-27 00:13:48 <Diablo-D3> UGH
 81 2010-10-27 00:13:51 <Diablo-D3> those matrox boxes suck
 82 2010-10-27 00:13:56 <brocktice> Yeah they were pretty expensive too
 83 2010-10-27 00:13:57 <Diablo-D3> basically the box lies about what it is
 84 2010-10-27 00:14:00 <brocktice> so I avoided that.
 85 2010-10-27 00:14:05 <Diablo-D3> it says its like a 3xxx x 1024 screen
 86 2010-10-27 00:14:08 <Diablo-D3> and just manually splits it out
 87 2010-10-27 00:14:22 <Diablo-D3> brocktice: btw, how did you get > 2 heads?
 88 2010-10-27 00:14:38 <brocktice> I have 2 DVI ports on each card
 89 2010-10-27 00:14:49 <brocktice> running TwinView on each, then Xinerama across them.
 90 2010-10-27 00:14:49 <Diablo-D3> so you're using crossfire?
 91 2010-10-27 00:14:53 <Diablo-D3> oh EWW
 92 2010-10-27 00:14:56 <brocktice> No, I'm on NVIDIA
 93 2010-10-27 00:14:56 <Diablo-D3> eww eww eww
 94 2010-10-27 00:14:58 <Diablo-D3> eww
 95 2010-10-27 00:15:04 <Diablo-D3> why arent you using sli?
 96 2010-10-27 00:15:06 <brocktice> Hey, it works with full acceleration.
 97 2010-10-27 00:15:08 <ArtForz> basic problem is my CRTs advertise a sgnificantly lower max hsync via EDID than what they can do
 98 2010-10-27 00:15:11 <brocktice> Not identical cards.
 99 2010-10-27 00:16:00 <brocktice> It wasn't clear to me whether I could do SLI anyway, last time I was involved with SLI was in the VooDoo2 days.
100 2010-10-27 00:16:29 <Diablo-D3> yes, but nvidia bought 3dfx
101 2010-10-27 00:16:35 <brocktice> bad pic, but you get the idea: http://www.flickr.com/photos/brocktice/4606055067/
102 2010-10-27 00:16:38 <Diablo-D3> sli and crossfire and 3dfx's sli are all the same shit
103 2010-10-27 00:16:44 <Diablo-D3> take more than one card, and render the same scene with it
104 2010-10-27 00:16:50 <ArtForz> well, not quite
105 2010-10-27 00:16:53 <Diablo-D3> its just gotten more advanced
106 2010-10-27 00:17:05 <ArtForz> original 3dfx sli was Scan Line Interleave
107 2010-10-27 00:17:08 <Diablo-D3> they either split tiles or split frames, depending on what the app is doing
108 2010-10-27 00:17:14 <Diablo-D3> ArtForz: well
109 2010-10-27 00:17:17 <Diablo-D3> they CAN do that too
110 2010-10-27 00:17:21 <Diablo-D3> its just so fucking not optimal
111 2010-10-27 00:17:22 <brocktice> The big problem here is that without buying a really really expensive card, it wasn't possible to get 4x DVI outs on a single board.
112 2010-10-27 00:17:25 <ArtForz> yep
113 2010-10-27 00:17:32 <brocktice> With DisplayPort that is changing.
114 2010-10-27 00:17:36 <Diablo-D3> tile rendering was the best idea yet
115 2010-10-27 00:17:47 <ArtForz> completely useless with modern shader based paths
116 2010-10-27 00:17:57 <Diablo-D3> brocktice: yeah, you can just use eyefinity cards
117 2010-10-27 00:18:02 <ArtForz> so nowadays they do tiling or alternate frame rendering
118 2010-10-27 00:18:08 <Diablo-D3> brocktice: 6 outs on a card
119 2010-10-27 00:18:26 <brocktice> Right, next build I'll do that.
120 2010-10-27 00:18:28 <Diablo-D3> 2 native dvi/hdmi, 6 displayport, and you can use 6 at a time
121 2010-10-27 00:18:30 <Diablo-D3> (buy displayport->hdmi bridges for >2 dvi/hdmi)
122 2010-10-27 00:19:13 <ArtForz> there should be cheap active DP -> single link DVI/HDMI bridges now
123 2010-10-27 00:19:38 <ArtForz> the older dual-link capable active bridges kinda suck
124 2010-10-27 00:19:57 <Diablo-D3> cheap? sort of
125 2010-10-27 00:20:01 <Diablo-D3> they're like $30 or some shit
126 2010-10-27 00:20:09 <ArtForz> = they just blank out FNAR and work fine after you re-plug em... for a while
127 2010-10-27 00:20:14 <Diablo-D3> and dp->dual link dvi is sorta pointless
128 2010-10-27 00:20:34 <Diablo-D3> if you have >1920x1200, JUST BUY A FUCKING NEW MONITOR YOU RICH BASTARD
129 2010-10-27 00:20:56 <ArtForz> well, unless you already have a bunch of 2560xsometing DVI screens around :P
130 2010-10-27 00:20:57 <brocktice> Yeah, I had two old monitors and two newer hand-me-downs
131 2010-10-27 00:21:00 <brocktice> so no HDMI
132 2010-10-27 00:21:23 <brocktice> The old viewsonics are better monitors, but lower resolution than the samsungs.
133 2010-10-27 00:21:30 <ArtForz> HDMI and single-link DVI are pretty much the same thing
134 2010-10-27 00:24:00 <ArtForz> hrrrm... since quad 5970 works... I wonder what the limit is
135 2010-10-27 00:24:22 <Diablo-D3> yes, it is the same thing
136 2010-10-27 00:24:39 <Diablo-D3> they just threw in extra pins to do something suspiciously like digital audio
137 2010-10-27 00:24:41 <Malouin_> I try to find image or video of 1 rig with 4 radeon 5970, but i don't find any. How come?
138 2010-10-27 00:24:57 <Diablo-D3> as in, I can be a dumb fuck, take a really cheap fucking controller, and split it out to normal old school digital
139 2010-10-27 00:25:14 <Diablo-D3> Malouin_: because until a little bit ago, you couldnt
140 2010-10-27 00:25:15 <ArtForz> http://www.youtube.com/watch?v=YYSSnqpchaw
141 2010-10-27 00:25:29 <Diablo-D3> Malouin_: driver had a hardwired max of 4 gpus, 2x 5970 = 4
142 2010-10-27 00:25:34 <Diablo-D3> but now
143 2010-10-27 00:25:36 <Diablo-D3> NOW
144 2010-10-27 00:25:44 <Diablo-D3> WE SHALL SEE OUR ENEMIES DRIVEN BEFORE US
145 2010-10-27 00:25:56 <Malouin_> that explain why, thanks. And thanks for the video, that was the one I was looking for
146 2010-10-27 00:25:59 <Diablo-D3> AND HEAR THE LAMANATIONS OF THEIR WOMEN
147 2010-10-27 00:26:05 <Malouin_> lol
148 2010-10-27 00:26:07 <ArtForz> btw, that asshat is cooking his middle cards, they're throttling at >90???C
149 2010-10-27 00:26:29 <brocktice> What else can he do?
150 2010-10-27 00:26:38 <brocktice> Take off the fans/sinks and do liquid cooling?
151 2010-10-27 00:26:53 <ArtForz> flexible PCIe extenders
152 2010-10-27 00:27:08 <Malouin_> Ouch, 90c
153 2010-10-27 00:27:33 <brocktice> did not know such a thing existed, good to know.
154 2010-10-27 00:27:59 <ArtForz> I built a quad 5970 rig, used 4" flexible PCIe extenders and staggered the cards
155 2010-10-27 00:28:19 <Malouin_> He can't possibly run that on 1 psu.. Can he ?
156 2010-10-27 00:28:23 <brocktice> For what? I hope not just mining.
157 2010-10-27 00:28:36 <ArtForz> just mining
158 2010-10-27 00:29:01 <brocktice> And, you know, shits and giggles. I can't imagine that will pay itself off for a very long time.
159 2010-10-27 00:29:13 <ArtForz> it already did
160 2010-10-27 00:29:19 <brocktice> When did you build it?
161 2010-10-27 00:29:27 <ArtForz> about 2 months ago
162 2010-10-27 00:29:36 <brocktice> Wow, well, I stand corrected.
163 2010-10-27 00:29:40 <ArtForz> well, it used to be 2 dual-5970 rigs
164 2010-10-27 00:29:55 <brocktice> yeah but that's actually more expensive. :)
165 2010-10-27 00:30:07 <ArtForz> airflow is still a massive problem, even with card fans @ 100% and 2 22W 120x38mm fans the hottest core gets over 80???C
166 2010-10-27 00:30:26 <brocktice> At least heating the room it's in shouldn't be an issue in the winter.
167 2010-10-27 00:30:53 <ArtForz> yup
168 2010-10-27 00:31:23 <Malouin_> thats the argument I told to my GF to explain the computer 24/7
169 2010-10-27 00:31:29 <ArtForz> I still wonder if the driver would work with >4 5970s
170 2010-10-27 00:32:31 <Malouin_> We will know it before 2011
171 2010-10-27 00:32:35 <brocktice> It does help in my basement office with my current workstation, and I'm not running such a large thermal envelope.
172 2010-10-27 00:33:13 <ArtForz> 7 5970s = 32Tflop MULADD in a box (or for mining about 3.75Ghash/s)
173 2010-10-27 00:34:48 <Malouin_> I plan to attain the 1.2Ghash/s by the end of november, if everything go well
174 2010-10-27 00:35:11 <ArtForz> and yes, 4 5970s on a 1200W PSU is just asking for trouble
175 2010-10-27 00:35:53 <Malouin_> I'm gonna use my 1000watt for 2 5970, and my 1200W for the main computer + 2 other 5970, I mean the final rig
176 2010-10-27 00:36:16 <Malouin_> I tried to find where he were hiding his second psu, but didn't find it.. that why i asked
177 2010-10-27 00:36:29 <ArtForz> as each 5970 is 300W, either the rest of the system is running on pixie dust or he's overloading his PSU
178 2010-10-27 00:37:05 <ArtForz> but I guess if you're running cards packed like that you don't really care if the rig blows up
179 2010-10-27 00:37:28 <Malouin_> hahaha
180 2010-10-27 00:37:52 <ArtForz> btw, for the "how the fuck do you run 7 cards in a box"... http://fastra2.ua.ac.be/?page_id=38
181 2010-10-27 00:38:31 <ArtForz> thats 6 GTX295s and a GTX275
182 2010-10-27 00:39:32 <ArtForz> that page is what pretty much gave me the idea of using PCIe extenders to stagger cards
183 2010-10-27 00:39:39 <Malouin_> what did he put between each card ?
184 2010-10-27 00:39:56 <ArtForz> on that FASTRA box? nothing
185 2010-10-27 00:40:14 <Malouin_> thats worst than the 4 5970..
186 2010-10-27 00:41:08 <ArtForz> http://fastra2.ua.ac.be/wp-content/gallery/fastra2/IMG_0413.JPG <- I guess that fan isn't standing there randomly...
187 2010-10-27 00:41:37 <Diablo-D3> [10:37:52] <ArtForz> btw, for the "how the fuck do you run 7 cards in a box"... http://fastra2.ua.ac.be/?page_id=38
188 2010-10-27 00:41:39 <Diablo-D3> "very carefully"
189 2010-10-27 00:42:19 <ArtForz> btw, it looks like they would've had space to space out the cards a bit
190 2010-10-27 00:42:57 <ArtForz> then just add a bunch of industrial fans and it should be somewhat workable
191 2010-10-27 00:43:01 <Malouin_> yeah, some space on the top one
192 2010-10-27 00:43:12 <Diablo-D3> see
193 2010-10-27 00:43:15 <Diablo-D3> Im thinking
194 2010-10-27 00:43:19 <Diablo-D3> why slot the cards in slots at all
195 2010-10-27 00:43:56 <Diablo-D3> there should be some sort of way to, like, glue them to the sides of a giant block of aluminym
196 2010-10-27 00:44:07 <Diablo-D3> and use the extenders to plug them in
197 2010-10-27 00:45:16 <Diablo-D3> you dont have to worry about heatsink weight
198 2010-10-27 00:45:23 <Diablo-D3> since they'd just be, like, laying on the heatsink
199 2010-10-27 00:46:04 <ArtForz> if money were no concern I'd just watercool em
200 2010-10-27 00:46:33 <ArtForz> with a decent full cover block most 5970s have no problem hitting 1GHz
201 2010-10-27 00:47:00 <Malouin_> 5970s are monsters
202 2010-10-27 00:47:41 <Diablo-D3> ArtForz: actually
203 2010-10-27 00:47:45 <Diablo-D3> why dont you just go water?
204 2010-10-27 00:47:55 <Diablo-D3> for the shit you do, its not THAT much of an overhead
205 2010-10-27 00:47:59 <Diablo-D3> like, what, $200 more?
206 2010-10-27 00:48:02 <ArtForz> decent waterblocks don't grow on trees
207 2010-10-27 00:48:02 <Diablo-D3> and its a one time cost?
208 2010-10-27 00:48:32 <Kiba> just go water?
209 2010-10-27 00:48:37 <ArtForz> and 910 to 1000 is only 10%
210 2010-10-27 00:48:41 <Kiba> is there a difference between water cooling and waterblock?
211 2010-10-27 00:48:44 <ArtForz> nope
212 2010-10-27 00:48:52 <Diablo-D3> Kiba: the block is the "heatsink" that attaches to the device
213 2010-10-27 00:49:47 <Diablo-D3> heh
214 2010-10-27 00:49:53 <ArtForz> hrrrm... actually immersion cooling might work...
215 2010-10-27 00:51:33 <ArtForz> aka, just put the whole block of cards under oil
216 2010-10-27 00:51:49 <Malouin_> wasn't that a fake rumor?
217 2010-10-27 00:51:53 <ArtForz> nope
218 2010-10-27 00:52:02 <ArtForz> works quite well
219 2010-10-27 00:52:23 <Diablo-D3> Malouin_: no, it works
220 2010-10-27 00:52:39 <Diablo-D3> everything but shit with moving parts (ie, hdd, etc) go into a vat
221 2010-10-27 00:52:51 <Malouin_> Hmmmm
222 2010-10-27 00:53:16 <Diablo-D3> have to use, obviously, non-conductive and non-corrosive/reactive fluid
223 2010-10-27 00:53:21 <ArtForz> yep
224 2010-10-27 00:53:50 <ArtForz> we tried it with pure parrafin oil
225 2010-10-27 00:54:10 <ArtForz> yeah, a computer running in... kerosene
226 2010-10-27 00:54:12 <Kiba> what about lead oiling?
227 2010-10-27 00:54:19 <ArtForz> nothing could go wring with that :P
228 2010-10-27 00:54:28 <Kiba> s/oiling/posioning?
229 2010-10-27 00:54:50 <ArtForz> huh?
230 2010-10-27 00:55:02 <Diablo-D3> lol
231 2010-10-27 00:55:26 <ArtForz> who the fuck drinks kerosene?
232 2010-10-27 00:55:31 <Malouin_> :P
233 2010-10-27 00:55:34 <Kiba> cooking oil
234 2010-10-27 00:55:44 <Diablo-D3> ArtForz: "lead oiling"
235 2010-10-27 00:55:50 <Diablo-D3> not something I'd want around
236 2010-10-27 00:57:04 <ArtForz> well, we didnt want to use vegetable oil so we don't get a biology experiment
237 2010-10-27 00:57:57 <ArtForz> I suspect you'll still need a pump and radiator at these power levels though
238 2010-10-27 00:58:17 <Diablo-D3> well, high grade pure biodiesel should work in theory
239 2010-10-27 00:58:27 <Diablo-D3> since it has non-corrosive addatives
240 2010-10-27 00:58:28 <Diablo-D3> but meh
241 2010-10-27 00:59:14 <Malouin_> i'm out, see you tomorrow
242 2010-10-27 00:59:23 <ArtForz> unless you make the tank act as a huge passive rad
243 2010-10-27 00:59:31 <Diablo-D3> ArtForz: well thats the point
244 2010-10-27 00:59:36 <Diablo-D3> you use a giant fucking tank
245 2010-10-27 00:59:39 <Diablo-D3> and let the heat diffuse
246 2010-10-27 00:59:58 <Diablo-D3> well not BIG big
247 2010-10-27 01:00:00 <Diablo-D3> but tall
248 2010-10-27 01:00:19 <Diablo-D3> the convection currents in the medium should continually drag heat up
249 2010-10-27 01:00:46 <ArtForz> yep
250 2010-10-27 01:01:38 <ArtForz> but I suspect for this power level you'd need a huge tank to get enough surface area
251 2010-10-27 01:02:20 <ArtForz> or you have to rib/fin/whatever the sides of the tank = passive radiator
252 2010-10-27 01:02:28 <Diablo-D3> ArtForz: well
253 2010-10-27 01:02:31 <Diablo-D3> withb a convection current
254 2010-10-27 01:02:37 <Diablo-D3> the hot medium will move up
255 2010-10-27 01:02:49 <ArtForz> the problem is the tank->air heat transfer
256 2010-10-27 01:03:03 <Diablo-D3> and wont come down until either the temp of the tank equalizes (and thus the convection current stops until the temp gradent increases again...)
257 2010-10-27 01:03:09 <ArtForz> unless you want a vat of very hot oil
258 2010-10-27 01:03:13 <Diablo-D3> or hotter stuff comes up
259 2010-10-27 01:03:15 <Diablo-D3> ArtForz: well basically yes
260 2010-10-27 01:03:23 <Diablo-D3> but the top will be hotter than the bottom
261 2010-10-27 01:03:27 <ArtForz> yep
262 2010-10-27 01:03:30 <ArtForz> but not by much
263 2010-10-27 01:03:37 <Diablo-D3> depends how tall the tank is
264 2010-10-27 01:04:11 <ArtForz> my guess is you get a kind of toroid flow
265 2010-10-27 01:04:26 <Diablo-D3> yeah, in a way
266 2010-10-27 01:04:34 <Diablo-D3> column comes up in the middle, sinks down the sides
267 2010-10-27 01:04:38 <ArtForz> yep
268 2010-10-27 01:28:28 <Kiba> hmm some of the forum members sure post a lot
269 2010-10-27 01:28:50 <noagendamarket> like me? lol
270 2010-10-27 01:29:06 <Diablo-D3> heh
271 2010-10-27 01:29:10 <Diablo-D3> I hardly ever post
272 2010-10-27 01:29:45 <Kiba> if you think about it
273 2010-10-27 01:29:58 <Kiba> most of the recent thread in the bitcoin section are by me
274 2010-10-27 01:30:10 <Diablo-D3> I DO NOT HAVE TIME TO THINK
275 2010-10-27 01:30:13 <Diablo-D3> THINKING IS FOR BABIES
276 2010-10-27 01:30:23 <Diablo-D3> WHAT SICK MAN SENDS BABIES TO FIGHT
277 2010-10-27 01:31:29 <AAA_awright> NOMMIES!
278 2010-10-27 01:31:44 <AAA_awright> Eww baby
279 2010-10-27 02:29:13 <Diablo-D3> ArtForz: hey, the header has unix time, right?
280 2010-10-27 02:29:28 <ArtForz> yup
281 2010-10-27 02:29:37 <Diablo-D3> how far drift to I get to have?
282 2010-10-27 02:29:45 <ArtForz> err, what?
283 2010-10-27 02:30:21 <Diablo-D3> well, its not included in the state I get from getwork, right?
284 2010-10-27 02:30:29 <ArtForz> yes it is
285 2010-10-27 02:30:35 <Diablo-D3> oh, it is?
286 2010-10-27 02:31:12 <Diablo-D3> ArtForz: so wait
287 2010-10-27 02:31:27 <Diablo-D3> then what part do I need to hash into the existing state it hands me?
288 2010-10-27 02:31:29 <Diablo-D3> just the nonce?
289 2010-10-27 02:32:02 <ArtForz> last 16 bytes of the 80 byte block header and the constant sha256 padding
290 2010-10-27 02:32:17 <ArtForz> you increment the nonce which is the last 4 bytes of the block header
291 2010-10-27 02:32:36 <Diablo-D3> so it has done EVERYTHING but the last 4 bytes and the padding?
292 2010-10-27 02:32:43 <Diablo-D3> so thats ALL I have to sha?
293 2010-10-27 02:32:46 <ArtForz> last 16 bytes
294 2010-10-27 02:33:05 <ArtForz> = last 4 bytes of hashMerkleRoot, nTime, nBits, nNonce
295 2010-10-27 02:33:22 <Diablo-D3> goddamnit.
296 2010-10-27 02:33:30 <Diablo-D3> wtf is a merklroot?
297 2010-10-27 02:33:48 <ArtForz> it's the second 32-byte hash in the block header
298 2010-10-27 02:34:44 <ArtForz> = 256 bit, and it straddles the first and second sha256 block, thats why the second block starts with last 4 bytes of it
299 2010-10-27 02:34:59 <Diablo-D3> why the fuck did they make this so difficult
300 2010-10-27 02:35:07 <ArtForz> it's fucking easy as shit
301 2010-10-27 02:35:36 <ArtForz> you get 80 bytes block header, 32 bytes state
302 2010-10-27 02:35:50 <ArtForz> you init your sha256 state to state
303 2010-10-27 02:36:04 <ArtForz> one sha256 block is 64 bytes
304 2010-10-27 02:36:20 <Diablo-D3> A through H is state?
305 2010-10-27 02:36:22 <ArtForz> your 64 byte block is the last 16 bytes of the 80 byte header, folloewd by sha padding (= 80000000...)
306 2010-10-27 02:36:24 <ArtForz> yes
307 2010-10-27 02:37:04 <ArtForz> and the last 4 bytes in your block is 00000280
308 2010-10-27 02:37:13 <ArtForz> which is 640, or the length of the dat hashed in bits
309 2010-10-27 02:37:24 <ArtForz> = standard fucking sha256 padding
310 2010-10-27 02:37:39 <ArtForz> then you get your sha256 output
311 2010-10-27 02:38:05 <ArtForz> you put that in a scratch area followed by sha256 padding for a 32-byte message (as you're hasing your hash)
312 2010-10-27 02:38:32 <ArtForz> thats the input data to another block of sha256 with normal initial state
313 2010-10-27 02:38:43 <ArtForz> and the output of that is the hash you compare to target
314 2010-10-27 02:38:53 <ArtForz> like I said, completely trivial
315 2010-10-27 02:41:16 <Keefe> lol
316 2010-10-27 02:42:04 <ArtForz> it gets a bit more complicated if you start moving around or eliminating endianness conversions in between and save the state after the first 3 rounds of the first block
317 2010-10-27 02:43:10 <Keefe> i figured it all out by reading the wikipedia article on sha2 and studying solar's code and bitcoin source
318 2010-10-27 02:43:18 <ArtForz> yup
319 2010-10-27 02:43:30 <ArtForz> the wikipedia article on sha2 is quite decent
320 2010-10-27 02:43:48 <appamatto> how goes the btc mining?
321 2010-10-27 02:44:35 <ArtForz> well, thanks to the dead PSU I'm running at reduced capacity, only 1900btc today
322 2010-10-27 02:45:15 <appamatto> wtf
323 2010-10-27 02:45:17 <Kiba> what you're going to do with it?
324 2010-10-27 02:45:19 <doublec> hehe
325 2010-10-27 02:45:25 <appamatto> across how many cards?
326 2010-10-27 02:45:28 <Kiba> 7200 btc a day, right?
327 2010-10-27 02:45:45 <ArtForz> yep, total *should* be 7200 a day
328 2010-10-27 02:45:54 <appamatto> what the
329 2010-10-27 02:45:59 <appamatto> what % of btc is that?
330 2010-10-27 02:46:02 <ArtForz> for the whole network
331 2010-10-27 02:46:03 <Kiba> aren't you competing with others miners for btc?
332 2010-10-27 02:46:06 <ArtForz> yes
333 2010-10-27 02:46:18 <ArtForz> normally I'd be doing about 2200 currently
334 2010-10-27 02:46:18 <Kiba> what's the typical generation?
335 2010-10-27 02:46:30 <ArtForz> well, nominal is 7200/day for the whole network
336 2010-10-27 02:46:45 <ArtForz> currently we're at about 11000/day
337 2010-10-27 02:47:18 <Kiba> what you do with your bitcoins you generated though?
338 2010-10-27 02:47:31 <ArtForz> sell most of em, stockpile the rest
339 2010-10-27 02:47:46 <Keefe> i'm about to shift into stockpiling half
340 2010-10-27 02:48:06 <Keefe> i'm close to paying off my hw investment
341 2010-10-27 02:48:24 <ArtForz> I'm stockpiling about 20%
342 2010-10-27 02:49:27 <Kiba> I see that the bitcoin economy is now about half a million
343 2010-10-27 02:49:34 <ArtForz> rest get sold off and reinvested in GPUs
344 2010-10-27 02:49:39 <Keefe> Kiba: should have put a few K $ into btc a month ago
345 2010-10-27 02:49:49 <ArtForz> yup
346 2010-10-27 02:50:08 <ArtForz> well, I started with about $50 in hardware
347 2010-10-27 02:50:46 <ArtForz> now I have ... about $8k in hardware and 30kbtc
348 2010-10-27 02:51:01 <Kiba> so it's a quite profitable investment
349 2010-10-27 02:51:12 <Keefe> if it wasn't for ArtForz sorta bragging about his successes, i wouldn't have considered it ;)
350 2010-10-27 02:51:18 <ArtForz> it used to be more profitable
351 2010-10-27 02:51:44 <Keefe> started with $300 in hw, now $950
352 2010-10-27 02:51:50 <ArtForz> but it's still decent profit
353 2010-10-27 02:52:50 <ArtForz> $1k in hardware produces about $25/day profit
354 2010-10-27 02:52:59 <Keefe> which reminds me, i am going to send you some btc someday soon ArtForz, as a thank you for the inspiration and help
355 2010-10-27 02:53:30 <Kiba> hmm, maybe ArtForz would be interested in buying a share in my microcompany ;)
356 2010-10-27 02:53:57 <Kiba> but I supposed he could start his own microcompany if he wanted
357 2010-10-27 02:54:10 <Kiba> it'll be interested to do a corporation without lawyer and all that nonsense
358 2010-10-27 02:54:17 <Kiba> s/interested/interesting
359 2010-10-27 02:54:27 <ArtForz> and even if something more efficient comes along, you still have a pile of realtively recent high-end GPUs to sell
360 2010-10-27 02:55:45 <ArtForz> *shrug*, I don't really see the point in 3d printing
361 2010-10-27 02:56:41 <Kiba> it is already possible to have medium-scale production for 3D objects
362 2010-10-27 02:56:57 <ArtForz> I do have a lathe and a vertical mill
363 2010-10-27 02:56:59 <Kiba> ArtForz: Well, maybe with my cloneblocks, I'll be able to conquer the building toys market
364 2010-10-27 02:57:15 <ArtForz> rmeinds me, I want to convert that mill to CNC one day
365 2010-10-27 02:58:48 <ArtForz> already have all the parts, just too lazy to actually mount it
366 2010-10-27 03:00:16 <nanotube> ArtForz: 3d printing requires less skill on the part of the user than a lathe or a cnc rig. :) that's the point in 3d printing.
367 2010-10-27 03:00:24 <nanotube> (i think... there may be other points.... :) )
368 2010-10-27 03:00:32 <nanotube> ;;bc,blocks
369 2010-10-27 03:00:33 <gribble> 87852
370 2010-10-27 03:01:16 <ArtForz> well, I do mainly one-offs for personal use, so just sketching and manufacturing it is faster than doing a CAD drawing + CNC
371 2010-10-27 03:09:40 <Kiba> nanotube: I think 3D printers ARE CNC
372 2010-10-27 03:10:12 <Kiba> now thing-o-matics and others printers are just basically minifactories
373 2010-10-27 03:10:21 <Kiba> but I don't think large factories are going to disappear
374 2010-10-27 03:10:44 <Kiba> one advantage that they have over desktop factories is alway going to be scale
375 2010-10-27 03:12:44 <Kiba> someday, I am going top everyone with a...self replicating logistic network!
376 2010-10-27 03:14:30 <nanotube> heh
377 2010-10-27 03:26:35 <nanotube> ;;bc,blocks
378 2010-10-27 03:26:36 <gribble> 87858
379 2010-10-27 03:33:12 <Kiba> ;mtgox
380 2010-10-27 03:36:14 <nanotube> Kiba: bitbot's dead
381 2010-10-27 03:38:05 <Kiba> I know.
382 2010-10-27 03:41:43 <nanotube> <weep> :)
383 2010-10-27 03:47:52 <AAA_awright> BRILLIANT IDEA
384 2010-10-27 03:48:07 <Kiba> BRILLIANT WHAT?
385 2010-10-27 03:48:19 <AAA_awright> BitCoin is trying to create a network where you can send messages to people, but without a central address authority to assign addresses, yes?
386 2010-10-27 03:48:40 <Kiba> and what about it?
387 2010-10-27 03:48:47 <AAA_awright> Could you optimize this messaging network more, and create an IP protocol that doesn't have a central number authority?
388 2010-10-27 03:49:07 <Kiba> but what if an address conflict?
389 2010-10-27 03:49:21 <AAA_awright> BitCoin doesn't have any address conflicts I hope
390 2010-10-27 03:49:38 <AAA_awright> Here is the full idea... Each IP is simply a SHA1 or other hash of a public key
391 2010-10-27 03:49:38 <Kiba> well, the probability is very low, but it is very possible I guess
392 2010-10-27 03:50:18 <AAA_awright> All the packets are encrypted and signed with this, so simply knowing the address, you also know the packet hasn't been tampered wtih
393 2010-10-27 03:50:33 <AAA_awright> But how do you route that?
394 2010-10-27 03:50:49 <Kiba> get everyone to download the same block?
395 2010-10-27 03:51:07 <AAA_awright> You could use bloom filters, and keep a first-order hash of the hashes of all the routers you are linked to... then you assemble a second-order list of all of those filters, and so on...
396 2010-10-27 03:51:18 <AAA_awright> Assuming no malicous nodes
397 2010-10-27 03:51:29 <Kiba> bitcoin is based on not trusting anyone...
398 2010-10-27 03:51:36 <Kiba> you gottach do better than that
399 2010-10-27 03:51:41 <AAA_awright> You just have to pass the packet along to the router that says they have the shortest connection to the target
400 2010-10-27 03:52:07 <Kiba> well, I am going to go to sleep
401 2010-10-27 03:52:13 <Kiba> let me know if you're going to code this idea up
402 2010-10-27 03:52:18 <Kiba> I think it would be fanastic'
403 2010-10-27 03:52:23 <edcba> lol
404 2010-10-27 03:52:50 <edcba> maybe that's what i2p already do
405 2010-10-27 03:52:55 <edcba> or something else
406 2010-10-27 03:54:16 <AAA_awright> You could store the routing information for a billion computers with a 0.1% error in 10 gigabytes
407 2010-10-27 03:55:03 <AAA_awright> A 64-bit router could handle that no problem
408 2010-10-27 03:55:24 <AAA_awright> hmm
409 2010-10-27 03:55:49 <AAA_awright> Obvious security hole aside it seems too good to be true
410 2010-10-27 03:56:13 <AAA_awright> 704-bit headers, if you use SHA1 hashes
411 2010-10-27 03:58:28 <AAA_awright> You could possibly mitigate the security hole with some analytic magic to whitelist trusted routers
412 2010-10-27 03:58:39 <AAA_awright> idk, maybe it needs some more mathematics
413 2010-10-27 04:04:28 <noagendamarket> A messaging system using xmpp would be better
414 2010-10-27 04:05:18 <appamatto> AAA, sounds interesting to me
415 2010-10-27 04:05:55 <Diablo-D3> shit ALWAYS needs more math
416 2010-10-27 04:06:28 <appamatto> if your packets get dropped you can start "blaming" the node
417 2010-10-27 04:06:56 <appamatto> doesn't sound like a huge problem
418 2010-10-27 04:09:28 <Keefe> B4C is just teasing us btc sellers :)
419 2010-10-27 04:10:13 <Keefe> almost never any cash on hand, but highest rate around
420 2010-10-27 04:12:22 <joe_1> last night i tried sending everyone in the channel 10 BTC as a promotion for cashcow casino, but out of the whole channel only 4 people had IP addresses accepting bitcoins
421 2010-10-27 04:12:40 <Keefe> guess that's why it's not called cash4bitcoins ;)
422 2010-10-27 04:13:08 <appamatto> I didn't know IP could receive btc
423 2010-10-27 04:13:30 <joe_1> yeah if you click send coins, you can put an IP address where you would normally put the bitcoin address
424 2010-10-27 04:13:40 <appamatto> how does that work?
425 2010-10-27 04:14:02 <Keefe> guess my cloak cost me 10 btc, lol
426 2010-10-27 04:14:25 <joe_1> it's nothing special the software asks the recipient to provide the sender with a bitcoin address, then it goes how you would normally send money to an address.
427 2010-10-27 04:19:59 <edcba> btc service gives a btc addr to each connnected node
428 2010-10-27 04:20:23 <edcba> at connnect
429 2010-10-27 04:41:39 <noagendamarket> http://www.bitcoin.org/smf/index.php?topic=1579.msg18786;topicseen#msg18786
430 2010-10-27 04:42:15 <noagendamarket> bitbot is on my vps lol
431 2010-10-27 04:44:20 <srb123> is your vps in US, or AUS though?
432 2010-10-27 04:52:08 <phantomcircuit> Question, would a sustained DDoS attack against nodes cause the target to drop?
433 2010-10-27 04:52:18 <phantomcircuit> or rather to rise and the difficulty to drop
434 2010-10-27 04:53:06 <lfm> hard to say what would happen. some might crash, some might just quit
435 2010-10-27 04:53:41 <lfm> might depend on the exact nature of the attack
436 2010-10-27 04:53:53 <Keefe> i'd expect such an attack to one way or another slow down block generation or apparent generation rate, which would eventually lower the difficulty
437 2010-10-27 04:54:57 <phantomcircuit> so theoretically someone could DDoS a large % of the network and take advantage of the reduced target to generate relatively cheap coins?
438 2010-10-27 04:54:58 <lfm> if the client survives the attack it would still most likley lose all connectioons and I think generation stops if you have zero connections
439 2010-10-27 04:55:25 <lfm> phantomcircuit, I dont think it would work that way
440 2010-10-27 04:55:45 <phantomcircuit> lfm, why not?
441 2010-10-27 04:55:57 <lfm> you gonna sustain it for months?
442 2010-10-27 04:56:17 <phantomcircuit> a DDoS against mostly home upstream?
443 2010-10-27 04:56:37 <phantomcircuit> yes actually i could marshall the resources to do that against several hundred thousand home connections
444 2010-10-27 04:56:45 <phantomcircuit> marshal
445 2010-10-27 04:57:28 <lfm> well if you did, you would probably find few people willing to buy the result
446 2010-10-27 04:57:49 <phantomcircuit> hmm good point
447 2010-10-27 04:58:37 <phantomcircuit> although it would be quite difficult to prove who had done it, as a certain % of the network would have to be left up in order to generate blocks that anybody would accept
448 2010-10-27 04:58:51 <lfm> basically it owuld destroy yhe whole ecology of bitcoin Id guess.
449 2010-10-27 04:59:25 <phantomcircuit> so more or less a ddos would cause a complete shutdown of the bitcoin economy
450 2010-10-27 04:59:29 <phantomcircuit> uh huh
451 2010-10-27 04:59:49 <lfm> Ya , i think so, youd be stuck with lots of worthless bitcoins
452 2010-10-27 05:00:41 <phantomcircuit> actually it would require significantly more than a month most likely
453 2010-10-27 05:01:02 <phantomcircuit> as the initial target would assume a large # of generators
454 2010-10-27 05:02:00 <lfm> there might develop a whack-a-mole situation too where more nodes pop up
455 2010-10-27 05:03:09 <lfm> especially if you actually kept the basic system going long enf for the difficulty to significantly drop
456 2010-10-27 05:03:31 <srb123> what is the legality of organising a DDoS, i remember them talking about last week re: gene simmons vs 4chan
457 2010-10-27 05:04:04 <srb123> im guessing it illeagal but hard to trace?
458 2010-10-27 05:04:06 <lfm> depends where you are
459 2010-10-27 05:04:16 <srb123> illegal*
460 2010-10-27 05:04:45 <noagendamarket> well someone is bound to try it sooner or later
461 2010-10-27 05:05:14 <phantomcircuit> srb123, totally illegal
462 2010-10-27 05:05:20 <phantomcircuit> almost impossible to trace
463 2010-10-27 05:05:23 <Diablo-D3> illegal damnit
464 2010-10-27 05:05:30 <phantomcircuit> unless the attacker is retarded
465 2010-10-27 05:05:38 <Diablo-D3> srb123: btw, you mean 4chan vs gene simmons
466 2010-10-27 05:05:41 <Diablo-D3> and 4chan won that round
467 2010-10-27 05:05:42 <lfm> maybe, if they actually find it profitable in some way especially
468 2010-10-27 05:06:24 <srb123> so why did you come here to a IRC channel that has logged your IP, to discuss it?
469 2010-10-27 05:06:31 <phantomcircuit> how many nodes are there right now?
470 2010-10-27 05:06:45 <phantomcircuit> srb123, 7 proxies bro
471 2010-10-27 05:06:48 <lfm> on and off a few thousand I think
472 2010-10-27 05:06:58 <phantomcircuit> a few thousand?
473 2010-10-27 05:07:28 <lfm> hard to say exactly cuz lots are not full time
474 2010-10-27 05:07:35 <phantomcircuit> assuming mostly residential ISPs a relatively modest botnet could keep most of the network down for months
475 2010-10-27 05:08:07 <phantomcircuit> hell there is probably some kind of DoS in the client itself if i looked hard enough
476 2010-10-27 05:08:27 <lfm> but every node you ddos is one less possilble customer for the bitcoins you are planning on creating
477 2010-10-27 05:09:03 <phantomcircuit> lfm, i thought the clients all just assume that the most computationally difficult chain is the valid one
478 2010-10-27 05:09:20 <lfm> whats that got to do with it?
479 2010-10-27 05:09:45 <phantomcircuit> they'd have to know what was happening and specifically black list the generated chain
480 2010-10-27 05:09:57 <lfm> Im talking about the ddosed nodes dropping out of the bitcoin economy due to discouragment
481 2010-10-27 05:10:25 <phantomcircuit> nah you could do something like iono
482 2010-10-27 05:10:31 <lfm> not to mention they would be cut off the internet completely
483 2010-10-27 05:10:33 <phantomcircuit> cycle the ddos around
484 2010-10-27 05:10:48 <phantomcircuit> i mena you'd only need to hold a % of the network down for a while
485 2010-10-27 05:10:55 <phantomcircuit> wouldn't have to be the same nodes the entire time
486 2010-10-27 05:11:37 <lfm> phantomcircuit, ummm, not hwo that would work then
487 2010-10-27 05:12:14 <phantomcircuit> just have to keep individual nodes off the network long enough that they waste time computing w/o the right transaction tree
488 2010-10-27 05:12:43 <lfm> phantomcircuit, umm, if they're cut off, they wont even try to make any tree
489 2010-10-27 05:12:54 <phantomcircuit> so same effect then
490 2010-10-27 05:13:08 <phantomcircuit> but no need to keep individual nodes down for an entire month
491 2010-10-27 05:13:12 <phantomcircuit> just a few minutes
492 2010-10-27 05:13:20 <phantomcircuit> then let them up
493 2010-10-27 05:13:28 <phantomcircuit> rinse and repeat with a new set of nodes
494 2010-10-27 05:13:30 <lfm> they would start up again when you let off them
495 2010-10-27 05:13:37 <phantomcircuit> yes they would
496 2010-10-27 05:13:47 <phantomcircuit> but they would have spent that time not doing anything
497 2010-10-27 05:14:33 <lfm> so youd only drop the rate by that percentage then, not hardly significant it seems
498 2010-10-27 05:15:06 <phantomcircuit> i bet it could be made significant
499 2010-10-27 05:15:30 <lfm> 30% is what I would consider not significantt for this
500 2010-10-27 05:16:27 <phantomcircuit> still could raise the target by 30%
501 2010-10-27 05:16:31 <lfm> the difficult almost doubled last week
502 2010-10-27 05:17:19 <lfm> so if you knock it down 30% and its going up 50% you still lose
503 2010-10-27 05:17:30 <phantomcircuit> not exactly
504 2010-10-27 05:17:36 <phantomcircuit> i knock it down 30%
505 2010-10-27 05:17:42 <phantomcircuit> and then take everybody out
506 2010-10-27 05:17:55 <phantomcircuit> and generate the now slightly easier block
507 2010-10-27 05:18:15 <Diablo-D3> ArtForz: does bitcoind do json-rpc over http?
508 2010-10-27 05:18:34 <lfm> what you meant "take everybody out"? thats 100% then?
509 2010-10-27 05:18:42 <phantomcircuit> not to mention what happens when someone designs an ASIC for bitcoin generation?
510 2010-10-27 05:19:09 <phantomcircuit> lfm, yup, obviously people who have already generated a substantial number of bc aren't just going to get discouraged and quit
511 2010-10-27 05:19:46 <lfm> phantomcircuit, maybe, maybe not, if they just descide to sell all their bitcoin then the market will crash
512 2010-10-27 05:22:40 <phantomcircuit> lfm, well assuming people would cry faul
513 2010-10-27 05:22:41 <lfm> phantomcircuit, maybe you dont realize how small the bitcoin market is. It is infantile. You could easily make it crash with just a few thousand investment if yoou wanted to throw your money away on it.
514 2010-10-27 05:22:45 <phantomcircuit> what about an ASIC?
515 2010-10-27 05:23:43 <lfm> what about it. it might be competitive with the gpgpus and it might not, Im not sure.
516 2010-10-27 05:24:13 <phantomcircuit> an ASIC could easily out perform a GPU
517 2010-10-27 05:24:34 <lfm> kinda depends on the nature of the application tho doesnt it?
518 2010-10-27 05:25:05 <phantomcircuit> repetitive sha hashing is a perfect application for an ASIC
519 2010-10-27 05:25:26 <lfm> ok, if you're right then you definatly should do it.
520 2010-10-27 05:26:15 <phantomcircuit> nah
521 2010-10-27 05:26:18 <phantomcircuit> idgaf
522 2010-10-27 05:26:24 <lfm> you're not right?
523 2010-10-27 05:26:34 <phantomcircuit> im just pointing out that this can never go anywhere due to inherent limits
524 2010-10-27 05:26:38 <lfm> you don't like money?
525 2010-10-27 05:27:01 <phantomcircuit> i like money that the IRS will take
526 2010-10-27 05:27:14 <lfm> they take maybe half
527 2010-10-27 05:27:32 <phantomcircuit> did you know that those guys have the balls to require people who barter to pay them in USD?
528 2010-10-27 05:27:38 <phantomcircuit> what USD? they barter
529 2010-10-27 05:27:53 <lfm> they can barter for some usd then
530 2010-10-27 05:28:00 <joe_1> irs does not accept gold?
531 2010-10-27 05:28:08 <phantomcircuit> joe_1, lold
532 2010-10-27 05:28:13 <lfm> hehe, I dont blame em, gold is crap
533 2010-10-27 05:28:42 <lfm> unless you need some balast to keep your sailboat upright
534 2010-10-27 05:29:33 <doublec> Diablo-D3, yes it does do json-rpc over http
535 2010-10-27 05:29:34 <phantomcircuit> gold is valuable as a hedge against inflation
536 2010-10-27 05:30:02 <lfm> sometimes, sometimes gold drops in price too
537 2010-10-27 05:30:09 <Diablo-D3> I wonder how hard it is to code a json-rpc client
538 2010-10-27 05:31:06 <lfm> Diablo-D3, with the right libs it should be pretty easy
539 2010-10-27 05:31:19 <doublec> is it not possible to enter a sell order in mtgox at the moment?
540 2010-10-27 05:31:38 <lfm> doublec, you sure?
541 2010-10-27 05:31:42 <doublec> I get the message that my order is stored in my open orders but it isn't actually there.
542 2010-10-27 05:31:50 <doublec> and the bitcoin amount isn't subtracted from my amount.
543 2010-10-27 05:32:12 <lfm> double you may be running into round off errors, try one cent less
544 2010-10-27 05:32:33 <doublec> for example, I just tried 10 coins at 0.2 as a test
545 2010-10-27 05:32:41 <doublec> it says that it's stored as an open order
546 2010-10-27 05:32:44 <doublec> but it isn't
547 2010-10-27 05:33:02 <doublec> Diablo-D3, json-rpc clients are pretty easy
548 2010-10-27 05:33:22 <doublec> even better, most languages have a library for it
549 2010-10-27 05:33:54 <phantomcircuit> lol does that seriously value a single bitcoin at $0.20 ?
550 2010-10-27 05:34:24 <lfm> I sold 990 BTC at 0.185 yesterday
551 2010-10-27 05:34:33 <doublec> phantomcircuit, I picked a number greater than the current value to ensure it should be stored as an open order
552 2010-10-27 05:34:43 <doublec> to show that open orders aren't working
553 2010-10-27 05:35:15 <phantomcircuit> lfm, so you got 183.15USD cash money?
554 2010-10-27 05:35:21 <phantomcircuit> (or cash equivolent)
555 2010-10-27 05:35:25 <lfm> I think the high for the last few days on mtgox has been 0.19
556 2010-10-27 05:35:48 <joe_1> i may cash out when it reaches 10.00 USD per coin
557 2010-10-27 05:36:00 <lfm> phantomcircuit, just credit at the exchange so far, I havn't tried to transfer it out yet.
558 2010-10-27 05:36:18 <phantomcircuit> joe_1, lold
559 2010-10-27 05:37:03 <phantomcircuit> doublec, for $0.30 ?
560 2010-10-27 05:38:26 <lfm> doublec, I was having trouble getting mtgox to show me my trading history today. I didnt try any new orders to sell
561 2010-10-27 05:39:04 <doublec> odd
562 2010-10-27 05:40:59 <doublec> I've emailed them about it
563 2010-10-27 05:46:41 <lfm> I been spammed, they payed me 10btc tho so I dont mind! hehe ----From: http://cashcow.no-ip.orgMessage:
564 2010-10-27 05:47:35 <edcba> so you are one of the 4 getting it :)
565 2010-10-27 05:47:50 <lfm> edcba, you sent that?
566 2010-10-27 05:48:07 <edcba> nope
567 2010-10-27 05:48:12 <nathan7> la la la
568 2010-10-27 05:48:20 <lfm> hehe
569 2010-10-27 05:49:41 <edcba> i should try spamming bitcoiners too
570 2010-10-27 05:49:53 <edcba> not a bad idea lol
571 2010-10-27 05:50:12 <lfm> go thru the thread where people leave their ip so you cqan attach message
572 2010-10-27 05:57:16 <edcba> i have a prog that lists network ips
573 2010-10-27 05:57:24 <edcba> sent through btc network
574 2010-10-27 05:59:03 <lfm> nathan7, The requested URL /myaccount/cashier.php was not found on this server.
575 2010-10-27 06:00:41 <nathan7> :o
576 2010-10-27 06:00:50 <nathan7> How am I to blame?
577 2010-10-27 06:01:12 <lfm> just guessing
578 2010-10-27 06:02:59 <lfm> based on your "lalala". I may have misinterpretted it tho I spoze
579 2010-10-27 06:04:04 <joe_1> thanks for the report. I need to add a page there. This is the cashier button in the game; it leads to a URL that does not exist.
580 2010-10-27 06:04:57 <lfm> ya, you have the page accessible from login, just not linked to that page
581 2010-10-27 06:04:59 <edcba> the account page seems to complex for me
582 2010-10-27 06:05:23 <lfm> agree, not clear why you need the chequing account
583 2010-10-27 06:05:56 <edcba> i'd have put actions in table redirecting to another page with form to fill
584 2010-10-27 06:06:11 <joe_1> I was thinknig of expanding it to be a bank where you can swap contracts for real world currencies, and the casino was going to be an extra feature. But for now the casino is the main feature, so it doesn't make sense.
585 2010-10-27 06:06:32 <lfm> ya separate pages for buy in and cash out Id prefer
586 2010-10-27 06:06:39 <joe_1> ok
587 2010-10-27 06:06:40 <edcba> and maybe at login just give an addr to send the btc to
588 2010-10-27 06:07:48 <joe_1> before login? then i would tie the bitcoin address to the ip address?
589 2010-10-27 06:08:06 <lfm> after login of course
590 2010-10-27 06:08:25 <lfm> like that banner that is already there
591 2010-10-27 06:09:29 <lfm> joe not a good idea to tie accounts to ip addresses cuz ip addresses change sometimes
592 2010-10-27 06:10:24 <joe_1> right, that was the idea of placing an address at the login screen. The way it is implemented now, it is tied to the password protected account.
593 2010-10-27 06:11:34 <joe_1> So move it down to just one account instead of checking + casino account?
594 2010-10-27 06:11:36 <lfm> the bitcoin address could be the account name I spoze
595 2010-10-27 06:13:08 <lfm> joe_1, ya, I think just casino and your bitcoin account is the the checking
596 2010-10-27 06:15:18 <lfm> good luck joe_1 , I gotta log, see ya
597 2010-10-27 06:15:27 <joe_1> ok thanks for the suggestions
598 2010-10-27 06:15:41 <lfm> thanks for the 10btc
599 2010-10-27 06:15:45 <joe_1> np
600 2010-10-27 11:38:36 <Malouin_> Rainy day again =(
601 2010-10-27 11:54:40 <Malouin_> So the 6970 will have a big upgrade of power compared to the 5970?
602 2010-10-27 12:34:30 <gavinandresen> jgarzik: you awake?
603 2010-10-27 12:53:47 <glassresistor> so i was blissfully suprised when i took a look a my farm boxes and the market
604 2010-10-27 12:54:03 <glassresistor> looks like my btc has some actual value now
605 2010-10-27 13:02:31 <gavinandresen> mtgox or dwdollar or m0mchil:  any of you have some time to brainstorm?  I'm working on a redesign of the "label" part of the JSON-RPC API...
606 2010-10-27 13:03:54 <dwdollar> hey gavin, I'm here
607 2010-10-27 13:04:10 <mtgox> me too
608 2010-10-27 13:04:20 <dwdollar> It's been awhile since I've looked at the API though, heh
609 2010-10-27 13:05:05 <m0mchil> yup
610 2010-10-27 13:05:07 <gavinandresen> Right, so my goal is to teach bitcoin to do proper accounting
611 2010-10-27 13:05:10 <mtgox> I just want events
612 2010-10-27 13:05:22 <dwdollar> yes, events would be nice
613 2010-10-27 13:05:52 <mtgox> but most importantly of all I want the wallet to back itself up
614 2010-10-27 13:05:54 <gavinandresen> ... so Bad Things don't happen if your connection to bitcoind happens to crap out between the time you call sendtoaddress and the server sends you back the JSON-RPC result
615 2010-10-27 13:05:58 <dwdollar> proper accounting would also be good
616 2010-10-27 13:06:33 <jgarzik> gavinandresen: something transactional, like a two-phase commit?
617 2010-10-27 13:06:51 <jgarzik> gavinandresen: prepare-to-send step + send-ack step?
618 2010-10-27 13:07:55 <gavinandresen> jgarzik:  well... no, I don't think so.  I think all-or-nothing should work just fine, but that means batching commands
619 2010-10-27 13:08:57 <Teppy> gavinandresen: I've had some accounting challenges with Dragon's Tale - there's one specific thing that would make my solution cleaner that I can describe.
620 2010-10-27 13:09:05 <gavinandresen> ... and batching commands is an issue because when I checked a month ago none of the JSON-RPC libraries supported JSON-2.0 (which defines command batching)
621 2010-10-27 13:09:18 <gavinandresen> Teppy:  what's that?
622 2010-10-27 13:09:54 <glassresistor> gavinandresen: i noticed this problem when i first started using bitcoin, shouldn't a simple handshake work
623 2010-10-27 13:10:03 <Teppy> My game server doesn't use an external database at all. Every 3 minutes it dumps the entire world state to a file.
624 2010-10-27 13:10:41 <jgarzik> Teppy: what happens if crash occurs during state save?
625 2010-10-27 13:10:59 <ArtForz> I'd guess rollback
626 2010-10-27 13:11:30 <Teppy> Exactly - so what I do now is I track, for each user Total BTC Deposited (into Dragon's Tale) and Total BTC Withdrawn (from Dragon's Tale)
627 2010-10-27 13:12:31 <Teppy> Both numbers are nondecreasing.
628 2010-10-27 13:13:34 <Teppy> I can then resync a balance by querying bitcoind using getreceivedbylabel and the listtransactions patch that you developed, jgarzik.
629 2010-10-27 13:14:08 <ArtForz> the problem is outgoing TX -> user account mapping
630 2010-10-27 13:14:08 <glassresistor> Teppy: are you all keeping this in an open repo?
631 2010-10-27 13:14:21 <Teppy> Exactly.
632 2010-10-27 13:14:27 <Teppy> Open repo?
633 2010-10-27 13:15:03 <Teppy> Oh, you mean the source that does all this?
634 2010-10-27 13:15:12 <ArtForz> to resync incoming TX state between bitcoin and your accoutn DB you can just keep a DB of userid->bitcoin address(es) and use listtransactions
635 2010-10-27 13:15:21 <gavinandresen> All: my thoughts on redesign:   http://gist.github.com/649220
636 2010-10-27 13:15:28 <ArtForz> we can already attach a stored-in-wallet comment to outgoing transactions
637 2010-10-27 13:16:03 <ArtForz> so if there was some way to retreive that comment with listtransactions, you could just put the DB UserID there
638 2010-10-27 13:16:28 <Teppy> That's what jgarzik's patch does.
639 2010-10-27 13:16:36 <Teppy> There is already a comment field.
640 2010-10-27 13:16:51 <glassresistor> Teppy: yeah i'd like to see/workon yalls code, but running bitcoind for a while and find some of the tranactional aspects scary
641 2010-10-27 13:17:08 <ArtForz> well, then resyncing state between bitcoin and DB after crash shouldnt be too hard
642 2010-10-27 13:17:47 <Teppy> It's not, but using comments to store character names is sort of a hack.
643 2010-10-27 13:18:39 <ArtForz> well, you need to store some kind of unique per-user key
644 2010-10-27 13:19:22 <ArtForz> storing that in comment is a bit hackish, but as your app is the only thing accessing your wallet, it's imo fine
645 2010-10-27 13:19:29 <glassresistor> m0mchil: just found you miner opencl code, is it working?
646 2010-10-27 13:19:43 <ArtForz> glassresistor: yes
647 2010-10-27 13:19:53 <gavinandresen> Teppy:  you're exactly the use case I'm redesigning for-- characters == accounts, and bitcoin should keep track of, and be able to tell you, exactly what affected each account balance
648 2010-10-27 13:20:31 <ArtForz> oh, so you're talking about the planned multiple-account feature?
649 2010-10-27 13:20:36 <gavinandresen> Yup
650 2010-10-27 13:20:38 <glassresistor> ArtForz: m0mchil badass i hadn't seen anyone sharing opencl code for mining yes
651 2010-10-27 13:20:52 <glassresistor> so just run the .py file
652 2010-10-27 13:21:15 <glassresistor> where does it store the coins?
653 2010-10-27 13:21:16 <ArtForz> you also need a bitcoin node patched with the getwork patch
654 2010-10-27 13:22:15 <m0mchil> glassresistor: let's direct chat, the discussion about accounts is important
655 2010-10-27 13:22:24 <Teppy> gavinandresen: So there would still be an overall wallet balance, and that would come from summing all the accounts?
656 2010-10-27 13:22:42 <gavinandresen> Teppy: yep
657 2010-10-27 13:22:56 <Teppy> And no individual account could ever be < 0...
658 2010-10-27 13:23:04 <glassresistor> m0mchil: no problem
659 2010-10-27 13:23:27 <Teppy> Yeah, that fits my situation precisely.
660 2010-10-27 13:23:49 <gavinandresen> Teppy: that's up to your application.  But sendfrom <account> <amount> will fail if account doesn't have at least <amount> in it.
661 2010-10-27 13:23:57 <ArtForz> yeah, that sounds fine
662 2010-10-27 13:25:15 <ArtForz> so basically addr for incoming tx == account ID, and being able to specify which addr to use to source a outgoing transaction?
663 2010-10-27 13:25:50 <gavinandresen> ArtForz: yes.
664 2010-10-27 13:25:59 <Teppy> So getnewaddress would take an account name as a parameter?
665 2010-10-27 13:26:15 <ArtForz> either that or the address *is* the account ID ;)
666 2010-10-27 13:26:31 <gavinandresen> Teppy: yes (like it takes "label" now)
667 2010-10-27 13:26:43 <gavinandresen> ArtForz:  No, each account can have multiple addresses.
668 2010-10-27 13:26:50 <ArtForz> yeah
669 2010-10-27 13:27:02 <ArtForz> that work too of course
670 2010-10-27 13:27:15 <Teppy> That sounds clean.
671 2010-10-27 13:27:19 <gavinandresen> ... although you COULD enforce "account name == bcaddress" in your app
672 2010-10-27 13:27:43 <Teppy> Are you going down multiple levels or just 1?
673 2010-10-27 13:27:51 <jgarzik> <gribble> BCM|   NEW BID|PPUSD                  2000 @ $0.1701
674 2010-10-27 13:27:52 <jgarzik> <gribble> BCM|   NEW BID|LRUSD                  6000 @ $0.1600
675 2010-10-27 13:27:53 <jgarzik> wow
676 2010-10-27 13:27:55 <Teppy> IOW, is an account made of account?
677 2010-10-27 13:27:57 <ArtForz> and I guess have a "default" account for generation and incoming tx-to-ip?
678 2010-10-27 13:28:07 <gavinandresen> Teppy:  no multiple levels.
679 2010-10-27 13:28:33 <gavinandresen> ArtForz:  very good question....
680 2010-10-27 13:29:44 <gavinandresen> ArtForz:  I think for sanity's sake it probably makes sense that, internally, every wallet transaction is associated with SOME account.  Default would be the "empty-string-named" account, I guess
681 2010-10-27 13:29:54 <ArtForz> yeah, sounds fine
682 2010-10-27 13:30:15 <ArtForz> so if you don'tz use the account feature all incming ends up in "", and all outgoing comes from ""
683 2010-10-27 13:30:54 <gavinandresen> ArtForz: ... that implys that if you use sendtoaddress it will FAIL if the "" account has no coins.  I think.  Which is a breaking change....
684 2010-10-27 13:31:08 <ArtForz> yes
685 2010-10-27 13:31:14 <ArtForz> well, not breaking really
686 2010-10-27 13:31:34 <glassresistor> anyone worked on making bitcoin just a head for bitcoind, so that you can run bitcoind and then open bitcoin for the gui functionality?
687 2010-10-27 13:31:41 <ArtForz> it's one-way pretty much, yes
688 2010-10-27 13:32:06 <ArtForz> I'd imagine all existing transactions would end up in "" when initially enabling multi-account
689 2010-10-27 13:33:01 <gavinandresen> ArtForz:   how this all affects the GUI is a concern.  E.g. I've got a receiving address labeled "From MyBitcoin" ... I don't want to have to specifically send from the "From MyBitcoin" account when I send somebody coins.
690 2010-10-27 13:33:50 <ArtForz> yes, if you plan on handling multiple accounts in the GUI, it'd need a overhaul
691 2010-10-27 13:33:58 <gavinandresen> So I think I take it back-- sendfromaddress should probably just take coins from accounts in alphabetical order (where "" is first).
692 2010-10-27 13:34:12 <Teppy> Sorry, disconnect
693 2010-10-27 13:34:31 <gavinandresen> (err. sendTOaddress is the current JSON-API call...)
694 2010-10-27 13:35:08 <ArtForz> that would be possible, but you'd need a pretty big warning sign
695 2010-10-27 13:35:45 <Teppy> So sendtoaddress would be depricated, right?
696 2010-10-27 13:35:55 <ArtForz> as now using the GUI or using plain sendtoaddr will throw off your account balances
697 2010-10-27 13:36:10 <gavinandresen> Teppy: good question!
698 2010-10-27 13:36:27 <ArtForz> well, more like, indiscriminately take coins to send from accounts
699 2010-10-27 13:36:46 <gavinandresen> ArtForz:  I see two distinct use cases:   bitcoind users running services.  And plain-old users...
700 2010-10-27 13:37:24 <ArtForz> I don't see the use case for mutiple account for normal GUI users
701 2010-10-27 13:37:41 <gavinandresen> Perhaps sendtoaddress should be removed from the JSON api, so you always have to choose an account to send from.
702 2010-10-27 13:37:45 <gavinandresen> ArtForz: I agree
703 2010-10-27 13:37:53 <m0mchil> I tend to agree too
704 2010-10-27 13:38:16 <ArtForz> it's needed if you run some service where you have to keep track of per-user balances
705 2010-10-27 13:38:58 <Teppy> If you go the depricated route, then you'd have to have a default account. Having one might make the transition easier.
706 2010-10-27 13:39:26 <Teppy> You'll have to put existing wallet balances into *some* account.
707 2010-10-27 13:40:27 <ArtForz> yep
708 2010-10-27 13:40:49 <ArtForz> imo "" would make the most sense
709 2010-10-27 13:41:23 <gavinandresen> Teppy: well... no...   account balances are just numbers in the database, and we COULD have sum(accounts) != total_server_balance
710 2010-10-27 13:41:35 <gavinandresen> But I'm certain that is just wrong.
711 2010-10-27 13:41:51 <gavinandresen> And will cause massive confusion and heartache.
712 2010-10-27 13:43:04 <theymos> How is this scheme different from Satoshi's move and getaccountaddress?
713 2010-10-27 13:43:44 <gavinandresen> theymos: it is more fleshed out, and I'm trying to think like an accountant.
714 2010-10-27 13:44:18 <gavinandresen> In particular, it supports getting a list of all transactions that went into the final account balance.
715 2010-10-27 13:44:23 <ArtForz> it pretty much is thew same idea as satoshis
716 2010-10-27 13:44:33 <ArtForz> except with listtransactions
717 2010-10-27 13:45:13 <ArtForz> unless I'm missing something
718 2010-10-27 13:45:50 <gavinandresen> Yep.  Satoshi sent me code, and set me loose to improve on it.
719 2010-10-27 13:48:00 <gavinandresen> So.... questions for y'all:    would this be useful if it didn't support some way of sending multiple commands to create one transaction?  (I think yes, and maybe batch transactions can come later)
720 2010-10-27 13:48:37 <theymos> If each account kept strictly separate coins, it could be used for maintaining strong anonymity.
721 2010-10-27 13:49:21 <dwdollar> gavinandresen:  I'd say so.  This will be a great help to new developers.
722 2010-10-27 13:49:42 <gavinandresen> theymos: Yeah.... no.   Maybe later...
723 2010-10-27 13:52:12 <gavinandresen> Question 2:   any objections to renaming  label--> account  everywhere in the API ?     (could be deprecated,with both names in the API for a while...)
724 2010-10-27 13:52:50 <gavinandresen> (I guess that is Question 3: to deprecate or just break stuff?)
725 2010-10-27 13:53:34 <mtgox> please just deprecate
726 2010-10-27 13:55:10 <dwdollar> I don't mind breaking, but I'm bias since I'm rewriting everything.
727 2010-10-27 13:55:32 <gavinandresen> Deprecating is polite.  And I'm all about being polite.
728 2010-10-27 13:56:15 <m0mchil> this early, better break imo
729 2010-10-27 13:56:20 <Teppy> As I understand it, it solves my problem very cleanly.
730 2010-10-27 13:56:47 <Teppy> Depricating is clean in this case.
731 2010-10-27 13:57:54 <Teppy> In fact even long-term, forcing the use of accounts may not be wise because it complicates the API for a more typical user.
732 2010-10-27 13:58:41 <gavinandresen> Teppy: do you mean a typical bitcoind user or typical GUI user?
733 2010-10-27 13:58:56 <Teppy> Typical bitcoind user.
734 2010-10-27 13:59:17 <m0mchil> as an UI user I'd like to not see addresses... just accounts
735 2010-10-27 13:59:51 <mtgox> I vote to make the API as simple as possible and leave the DB stuff to a real DB
736 2010-10-27 14:01:13 <gavinandresen> mtgox:  the Berkeley DB inside bitcoin is a "real database" ....
737 2010-10-27 14:01:22 <gavinandresen> mtgox:  but I sympathize
738 2010-10-27 14:01:57 <m0mchil> but I guess this is too big change at the moment (accounts UI)
739 2010-10-27 14:01:59 <nanotube> speaking of databases... the bdb is a pain to back up... it should be made simpler to use alternative db backends.
740 2010-10-27 14:02:28 <gavinandresen> mtgox: KISS is good, but I've struggled to figure out how to properly handle edge cases for the escrow service I'm working on.
741 2010-10-27 14:03:22 <gavinandresen> mtgox:  If both bitcoin and my database are remembering who paid what, when, then if the connection between the two is even slightly unreliable Bad Things can happen....
742 2010-10-27 14:03:32 <ArtForz> nanotube: why?
743 2010-10-27 14:04:38 <gavinandresen> ArtForz:  it'd be cool to use BigTable or Amazon's SimpleDB as the database back-end for bitcoin, so you get automatic replication and backup...
744 2010-10-27 14:04:47 <mtgox> isn't there some issue with a backup of wallet.dat getting corrupted after getting a new address? this sounds like a way more important thing to fix
745 2010-10-27 14:04:53 <ArtForz> nope
746 2010-10-27 14:04:59 <gavinandresen> mtgox:  that's already been done with the keypool feature
747 2010-10-27 14:05:08 <ArtForz> current bitcoin now keeps a ~100 addr pool in the wallet
748 2010-10-27 14:05:09 <nanotube> ArtForz: replication, management tools, etc... ?
749 2010-10-27 14:05:43 <Teppy> What I would *love* is a simple stand-alone utility that I could run like this, when bitcoind isn't running:
750 2010-10-27 14:05:59 <ArtForz> shrug, the only "DB" worth saving is the wallet, everthing else is just internal stuff thats recreated when we start from a blank slate anyways
751 2010-10-27 14:06:12 <nanotube> ArtForz: also, for a large service provider, the db may grow to millions of addresses, if a new address is created for all transactions, how scalable is the bdb backend?
752 2010-10-27 14:06:12 <Teppy> bcutil move 100 wallet.dat walletbackup.dat
753 2010-10-27 14:07:04 <MacRohard> you could just transfer your coins to another bitcoind, erase the wallet and transfer them back to defrag it
754 2010-10-27 14:07:09 <nanotube> ArtForz: using a 'real db' things can be done to inspect the data and delete old/unused addresses that have no balance, etc. otherwise, currently, wallet.dat grows infinitley
755 2010-10-27 14:07:27 <glassresistor> i think ui users should still see both accound/address otherwise they could be mislead but a clever person
756 2010-10-27 14:07:40 <ArtForz> you should be able to do the same thing with bdb right now, we dropped DB_PRIVATE
757 2010-10-27 14:07:53 <ArtForz> but yes, scalability is a problem
758 2010-10-27 14:08:44 <ArtForz> with only ~10000 transactions in my TEST client wallet flushes already take >0.3 sec
759 2010-10-27 14:11:10 <gavinandresen> nanotube:  databases, even Berkeley DB, are good at ignoring old/unused data...
760 2010-10-27 14:11:40 <gavinandresen> ArtForz: any idea what is taking so long?  I haven't run into any problems with the Faucet wallet.db (which is almost 50MB right now)
761 2010-10-27 14:12:17 <ArtForz> probably because /home on that box is nfs mounted
762 2010-10-27 14:12:38 <ArtForz> wallet is ~5MB
763 2010-10-27 14:14:29 <ArtForz> guess the fsyncs take quite a few network roundtrips
764 2010-10-27 14:17:03 <gavinandresen> ArtForz:  "Doctor, it hurts when I put my arm behind my back like this......"
765 2010-10-27 14:17:40 <bonsaikitten> gavinandresen: your arm is broken!
766 2010-10-27 14:17:58 <gavinandresen> ArtForz:   That said, I WOULD be interested in knowing if sqlite or some other DB handles that case better....
767 2010-10-27 14:18:21 <ArtForz> well, I just like having having all /home s on a central server, makes backups a lot easier
768 2010-10-27 14:18:26 <gavinandresen> (but I'm not interested enough in knowing that I'm willing to do any work to figure it out)
769 2010-10-27 14:19:47 <bonsaikitten> no, sqlite has a similar pessimized behaviour
770 2010-10-27 14:19:55 <bonsaikitten> silly C and D parts of ACID
771 2010-10-27 14:20:04 <ArtForz> yup
772 2010-10-27 14:23:28 <ArtForz> I guess you could remove old fully-spent transactions from wallet, but I think the logic to keep track of this would be nontrivial
773 2010-10-27 14:24:02 <glassresistor> if you really want good transactions and caching go with postgres, maybe with reddis cacheing-backend
774 2010-10-27 14:24:52 <bonsaikitten> overkill much?
775 2010-10-27 14:25:50 <glassresistor> bonsaikitten: me? if your making a bank with 1000+ users its not overkill
776 2010-10-27 14:26:35 <bonsaikitten> heh
777 2010-10-27 14:26:48 <bonsaikitten> how many transactions per timeunit do you expect?
778 2010-10-27 14:27:06 <bonsaikitten> sqlite is a bit silly with concurrency, but if you have only one thread it's pretty fast
779 2010-10-27 14:27:19 <ArtForz> oi, 0.20 PPUSD on bcm
780 2010-10-27 14:27:38 <gavinandresen> Yeah, 1,000 users per day is only, at most, a couple/few thousand transactions per day.  That's not many.
781 2010-10-27 14:28:23 <bonsaikitten> at 1k concurrent users it gets naughty
782 2010-10-27 14:29:19 <ArtForz> well, probably would make most sense to use a generic SQL connector library
783 2010-10-27 14:30:20 <glassresistor> ArtForz: orm's are always the way to go, so for dev work the db doesn't matter but once it goes live i always "dress" for a shit starm
784 2010-10-27 14:31:36 <bonsaikitten> urgh
785 2010-10-27 14:31:46 <bonsaikitten> ORMs are only good for very specific access patterns
786 2010-10-27 14:31:48 <ArtForz> well, I think we already store more-or-less normalized data
787 2010-10-27 14:32:01 <bonsaikitten> as soon as you hit performance issues you end up circumventing them again
788 2010-10-27 14:32:54 <gavinandresen> Right!   Re-implementing off of Berkeley DB ain't gonna happen in the next month, so one more question, then I'm gonna go eat lunch:
789 2010-10-27 14:33:04 <ArtForz> of course not in the next month
790 2010-10-27 14:33:17 <gavinandresen> Question: do we need a  listaccounts   API call?  Would it ever get used?
791 2010-10-27 14:34:20 <glassresistor> bonsaikitten: idk i work with databases everyday, like currently, and although some ORMs(like say cakephps) are really clunky often they have little to no performance hit if they are LAZY, and typically your not doing much more than joins, and searches
792 2010-10-27 14:34:43 <bonsaikitten> glassresistor: if your access pattern is simple enough
793 2010-10-27 14:34:47 <ArtForz> but if transaction volume picks up by a few orders of magnitude big sites might start running into problems
794 2010-10-27 14:35:05 <bonsaikitten> glassresistor: as soon as you do the business logic in the DB (as it should be) the ORM fails hard
795 2010-10-27 14:35:11 <gavinandresen> ArtForz: that would be a very good problem to have.  I tend not to worry about "good problems to have"...
796 2010-10-27 14:35:53 <glassresistor> bonsaikitten: sure but i have access patterns like I need to do a full text search for one field and but only want answers that have True in the published column of a through model, and the django ORM handles it
797 2010-10-27 14:36:18 <ArtForz> yeah, best to not start crossing a bridge unless you're at least on the same planet :P
798 2010-10-27 14:36:44 <ArtForz> otherwise you end up with horribly over-engineered solutions
799 2010-10-27 14:36:47 <Teppy> gavinandresen: Yes, and to scale, it should accept a wildcard like "a*"
800 2010-10-27 14:36:57 <bonsaikitten> glassresistor: hmm, then you're doing "boring" things, but a lot of them ;)
801 2010-10-27 14:37:21 <ArtForz> gavin: I don't really see a use case for listaccounts, but adding it shouldnt hurt
802 2010-10-27 14:37:55 <Malouin_> true
803 2010-10-27 14:39:38 <glassresistor> bonsaikitten: well yes but almost all business logic is boring
804 2010-10-27 14:40:34 <gavinandresen> Teppy:  Yeah.... no.  Maybe wildcards in a later version
805 2010-10-27 14:40:49 <jgarzik> <gribble> BCM|  CONF TRD|PPUSD                   500 @ $0.2000
806 2010-10-27 14:40:51 <jgarzik> wow
807 2010-10-27 14:41:07 <nanotube> heh yea seriously...
808 2010-10-27 14:41:22 <glassresistor> bonsaikitten: i've rarely had problems with aggregates or using special db functions, even say like postgis stuff
809 2010-10-27 14:41:45 <glassresistor> bonsaikitten: it should be fine for 95% of use cases and its rapidly speeds up dev
810 2010-10-27 14:42:16 <bonsaikitten> glassresistor: I've seen cases where the ORM was slower by a factor of ~500 than the "smart" handwritten SQL
811 2010-10-27 14:42:34 <bonsaikitten> not that common, but when it happens it's sad
812 2010-10-27 14:42:40 <ArtForz> jgarzik: yeah, oi, like I said ~15 minutes ago ;)
813 2010-10-27 14:43:21 <jgarzik> if you have decent storage, running db4 w/ TX_NOSYNC is realistic
814 2010-10-27 14:43:31 <jgarzik> ie. battery-backed RAID or somesuch
815 2010-10-27 14:43:34 <ArtForz> and LR$ at 0.16 ...
816 2010-10-27 14:43:58 <ArtForz> well, the jump for LR$ wasnt unexpected, with mtgox switching to LR
817 2010-10-27 14:44:55 <jgarzik> ArtForz: well, mtgox is IMO still a closed ecosystem until LR withdrawals flow freely
818 2010-10-27 14:45:32 <ArtForz> well, LR API is a pain compared to SCI
819 2010-10-27 14:45:54 <glassresistor> bonsaikitten: sometimes with orms you have to either fallback to a RAW or a extra sql function to make things faster or actually add to the orm but often creative objects can be just as fast
820 2010-10-27 14:46:19 <jgarzik> ArtForz: ?   LR API is an SCI, is it not?
821 2010-10-27 14:46:24 <bonsaikitten> glassresistor: well, if all you do is a blog with comments then the ORM is pretty perfect :)
822 2010-10-27 14:46:30 <ArtForz> erm, well, not really
823 2010-10-27 14:46:40 <ArtForz> the normal merchant SCI only lets you receive funds
824 2010-10-27 14:46:44 <gavinandresen> Thanks for the brainstorming, everybody.  I'll post a thread on the forums about all this probably later today (gotta eat now)
825 2010-10-27 14:46:52 <ArtForz> to send you need to use the XML API
826 2010-10-27 14:46:57 <glassresistor> bonsaikitten: agreed but i do allot more that without much trouble but i can write sql when i need to and i know what sql the ORM writes
827 2010-10-27 14:47:06 <jgarzik> ArtForz: point
828 2010-10-27 14:47:14 <UukGoblin> but bcm price went up a lot too
829 2010-10-27 14:47:15 <bonsaikitten> glassresistor: django even likes to tell you what sql it uses :)
830 2010-10-27 14:47:22 <bonsaikitten> glassresistor: that's why it's so nice to use
831 2010-10-27 14:47:28 <glassresistor> bonsaikitten: yeah i know
832 2010-10-27 14:47:36 <glassresistor> bonsaikitten: i run a django shop
833 2010-10-27 14:47:38 <ArtForz> so I can understand automated payouts taking a while to implement
834 2010-10-27 14:47:59 <jgarzik> ArtForz: yeah.  at least it's better than the Pecunix payment API situation :/
835 2010-10-27 14:48:14 <ArtForz> ouch, yeah
836 2010-10-27 14:49:18 <jgarzik> <gribble> BCM|   NEW BID|PXGAU                   500 @ GAU0.003500
837 2010-10-27 14:49:19 <jgarzik> craziness
838 2010-10-27 14:49:50 <UukGoblin> 172615 <+gribble> BCM|  CONF TRD|PPUSD                   500 @ $0.2000
839 2010-10-27 14:49:56 <UukGoblin> THAT'S crazyness.
840 2010-10-27 14:50:05 <ArtForz> yep
841 2010-10-27 14:50:09 <jgarzik> indeed
842 2010-10-27 14:50:25 <UukGoblin> and I bloody have a 1k BTC frozen in that bitcoin exchange...
843 2010-10-27 14:50:35 <UukGoblin> Sirius, wake up! ;-P
844 2010-10-27 14:50:39 <ArtForz> I think thats pretty much the first time I've seen BCM PP$ going above mtgox
845 2010-10-27 14:51:15 <glassresistor> which repository are you keeping the accouting branch in?
846 2010-10-27 14:51:37 <jgarzik> the high prices at BCM seem to be due to a lack of bitcoin sellers, IMO
847 2010-10-27 14:51:43 <jgarzik> asks are very thin, compared to bids
848 2010-10-27 14:53:02 <ArtForz> isn't that pretty much the same situation on mtgox?
849 2010-10-27 14:54:15 <ArtForz> well, not that extreme
850 2010-10-27 14:54:47 <ArtForz> but yeah, asks are very thin, looks like a few hedge bets and thats it
851 2010-10-27 14:54:55 <glassresistor> or is all this conversation about spec and not being published
852 2010-10-27 15:02:36 <gavinandresen> glassresistor: conversation about spec right now.  But it's next on my "improve bitcoin" todo list, and satoshi agrees it's needed.
853 2010-10-27 15:02:54 <nanotube> jgarzik: well... rising prices of /anything/ generally indicate a surplus of buyers relative to sellers. :)
854 2010-10-27 15:03:14 <Malouin_> Thats good for us
855 2010-10-27 15:03:50 <ArtForz> yes, miners like demand outstripping supply ;)
856 2010-10-27 15:04:05 <Malouin_> hehe
857 2010-10-27 15:05:08 <brocktice> It's pretty smart to make it easy for the earliest adopters to get 'free' bitcoins.
858 2010-10-27 15:05:18 <brocktice> Then those people will have an incentive to bolster demand by offering services, etc.
859 2010-10-27 15:05:31 <Malouin_> yup
860 2010-10-27 15:05:49 <brocktice> Which ideally will snowball into a real market that accepts bitcoins.
861 2010-10-27 15:06:03 <brocktice> 'real' meaning enough to sustain ongoing activity
862 2010-10-27 15:06:09 <Malouin_> Its depend on the community
863 2010-10-27 15:07:13 <Malouin_> ArtForz: Any guess on what will be the difficulty on december 31st 2010 ? :p
864 2010-10-27 15:08:25 <ArtForz> about 7000
865 2010-10-27 15:08:51 <nanotube> ArtForz: your mining will no longer be profitable at diff 7000, would it?
866 2010-10-27 15:08:56 <ArtForz> sure would
867 2010-10-27 15:09:12 <nanotube> i recall you saying that you'll reach break-even point at a difficulty of about 4k?
868 2010-10-27 15:09:18 <nanotube> has that changed?
869 2010-10-27 15:09:35 <Malouin_> only 7k ? I would have guess higher
870 2010-10-27 15:09:42 <ArtForz> that was at $0.04/btc ...
871 2010-10-27 15:09:51 <nanotube> ArtForz: aaaaah of course :)
872 2010-10-27 15:10:25 <ArtForz> and with 5770s
873 2010-10-27 15:10:31 <brocktice> ArtForz: So, with that quad-radeon configuration, how many bitcoins do you generate per day?
874 2010-10-27 15:10:34 <brocktice> or week
875 2010-10-27 15:10:47 <Malouin_> I love those numbers
876 2010-10-27 15:10:59 <nanotube> so i guess the interesting question now is... will the price increases continue to make up for generation difficulty increases in the foreseeable future. :)
877 2010-10-27 15:11:26 <ArtForz> quad 5970 = about 1250 btc/day at current diff
878 2010-10-27 15:11:40 <brocktice> wow
879 2010-10-27 15:12:33 <brocktice> making me want to price out a machine. :)
880 2010-10-27 15:13:26 <ArtForz> I'd guess about $1600 for a dual 5970 box
881 2010-10-27 15:13:54 <brocktice> I don't think I can spare the time, but it's a fun though.
882 2010-10-27 15:13:57 <brocktice> *thought
883 2010-10-27 15:13:59 <brocktice> I'm such a gadget whore
884 2010-10-27 15:14:05 <mtgox> whoa ArtForz you must be raking it in now
885 2010-10-27 15:14:11 <Malouin_> I'm going to finish my final rig in about 2 week, it will become a quad-radeon 5970
886 2010-10-27 15:14:25 <ArtForz> well, about $300 in profit today
887 2010-10-27 15:14:34 <Malouin_> with those flexible pcie extender you recommended me
888 2010-10-27 15:15:02 <brocktice> Got space in my LackRack...
889 2010-10-27 15:15:04 <ArtForz> well, I'm going back to dual 5970
890 2010-10-27 15:15:09 <brocktice> oh?
891 2010-10-27 15:15:14 <brocktice> What's the speedup?
892 2010-10-27 15:15:30 <ArtForz> pretty much 0
893 2010-10-27 15:15:42 <ArtForz> dual setup is just way easier to cool
894 2010-10-27 15:15:54 <brocktice> I mean, like, what's the ratio for each additional card? is it 1:1?
895 2010-10-27 15:16:04 <ArtForz> pretty much
896 2010-10-27 15:16:08 <brocktice> nice
897 2010-10-27 15:16:18 <brocktice> just going to do 2 separate boxes?
898 2010-10-27 15:16:36 <Malouin_> but your card aren't THAT hot, wasn't around 82c ?
899 2010-10-27 15:16:48 <UukGoblin> 82c is hot