| --- Day changed Sat Apr 19 2008 | ||
| -!- Splicer2 [n=p@h174n1c1o261.bredband.skanova.com] has joined #hplusroadmap | 00:56 | |
| kanzure | fenn: I wonder if agx-make directly generates CAM or not. I suspect not. agx-make would be more like the API layer to /dev/ more than anything else. | 01:01 |
|---|---|---|
| kanzure | yes? | 01:01 |
| kanzure | it would aggregate all of the information relevant as inputs to the machinery in your local fab-lab-configuration | 01:01 |
| fenn | agx-make is a compiler that can call many different 'backend' CAM tools | 01:08 |
| kanzure | can we hack gcc for this purpose? | 01:08 |
| kanzure | port gcc to a new 'isa' | 01:08 |
| kanzure | where this 'isa' is the fablab instruments? | 01:08 |
| fenn | dunno, gcc is a C compiler mostly | 01:08 |
| fenn | using make might make sense | 01:09 |
| fenn | gnu make | 01:09 |
| kanzure | I'm just wondering to what extent we can consider it a compiler | 01:10 |
| kanzure | since technically gnu make is not a compiler in the truest sense, last I checked | 01:10 |
| fenn | ok, it's not a compiler, its a "make" whatever you call that | 01:10 |
| kanzure | plus we don't have to compile the firmware for each new package physical implementation | 01:10 |
| kanzure | ok | 01:10 |
| -!- Splicer [n=p@h238n1c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 01:10 | |
| -!- Splicer [n=p@h174n1c1o261.bredband.skanova.com] has joined #hplusroadmap | 01:11 | |
| fenn | its a frontend that causes software files for performing a particular physical operation to be compiled | 01:11 |
| fenn | where said physical operation is one that's been verified to be feasible with the current infrastructure/inventory | 01:12 |
| kanzure | yes | 01:12 |
| kanzure | my thoughts were that the skdb packages would generate yaml serialized data structs | 01:13 |
| kanzure | these structs would be the files that the tools are programmed to physically follow and/or implement | 01:13 |
| kanzure | simple enough, no? | 01:13 |
| fenn | yeah, the yaml will get boiled down into longer and longer sequences of lower level data, until it reaches basic 'move motor' type information | 01:14 |
| kanzure | or "move your ass, human!" | 01:14 |
| fenn | well, you'd want to give the human higher level data | 01:14 |
| fenn | depending on their skill level, of course | 01:14 |
| fenn | you can say 'now solder r1. now solder r2' to an electronics pro | 01:15 |
| kanzure | I was thinking something more subtle; electric shocks? joking aside, yes | 01:15 |
| fenn | er, can't | 01:15 |
| fenn | http://ladyada.net/make/usbtinyisp/solder.html | 01:15 |
| fenn | it would be neat if we could automatically generate instructions like that | 01:16 |
| kanzure | she's a classic | 01:17 |
| kanzure | how many times has she been on Slashdot now? three? four? | 01:17 |
| fenn | dunno, slashdot is so crap anymore i dont read it | 01:17 |
| kanzure | especially after the electronics "SHE'S A WOMAN WITH ELECTRONS BUZZING AROUND, A TERRORIST! JUMP HER!" scare | 01:17 |
| -!- Splicer2 [n=p@h174n1c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 01:18 | |
| fenn | could you explain what this line means? To continue: just as you cannot search for software based on functionality, you cannot search for other tools based on functionality, except in as much as the language and social context allows you to express similar ideas that others have had before (natural languages have allowed this to some extent) | 01:21 |
| kanzure | yes | 01:22 |
| kanzure | suppose you 'know' that something must exist out there in the world | 01:22 |
| kanzure | try to find it on Google without having verbal grounding in reality | 01:22 |
| kanzure | knowing how other people would word it. | 01:22 |
| fenn | sure i had that problem the other day with eval() | 01:22 |
| kanzure | same thing with software ... you don't search for "if then if then if then" (since that'd be programming the thing you want!) | 01:22 |
| kanzure | you search for tags and descriptions etc., there is a lot of social context that programmers usually tend to forget ... they are not intuiting their programs from out of the blue :) | 01:23 |
| kanzure | as much as I'd like to think I am ... | 01:23 |
| fenn | you read too much for that to be true | 01:23 |
| kanzure | <-- tainted | 01:24 |
| kanzure | but I started off a total moron, so perhaps not? | 01:24 |
| fenn | back before the internet, i came up with the idea that we only really have three or four original thoughts in a year (actually i was a bit more harsh, saying all thoughts were derived from someone else) | 01:24 |
| kanzure | sure, aggregation | 01:25 |
| fenn | but see, we dont realize it, we think they're all new ideas and belong to us | 01:25 |
| kanzure | permutation/mutation/insight v. recombination | 01:25 |
| kanzure | s/recombination/recombination+aggregation/ | 01:25 |
| kanzure | perhaps permutation and mutation would go along with refactoring and documenting | 01:26 |
| fenn | no, refactoring is generalizing | 01:26 |
| fenn | or categorizing sometimes | 01:26 |
| kanzure | hm. | 01:26 |
| fenn | the point of refactoring is that it's not a creative act | 01:27 |
| fenn | you keep the same functionality but make a program easier to modify or improve upon | 01:27 |
| kanzure | hm, yes, so therefore it doesn't fit on that spectrum I was using above | 01:27 |
| fenn | epicycles -> kepler's law | 01:27 |
| fenn | can you imagine epicycles with relativity thrown in? | 01:28 |
| kanzure | epicycles were simply the rotations around celestial bodies, right? | 01:28 |
| kanzure | relativity is then just observer bias factored in, yes? | 01:28 |
| fenn | sorta. they said spheres are perfect so therefore planets must move in circles, or at least circles on circles | 01:29 |
| fenn | then they had circles on circles on circles to make it fit the data | 01:29 |
| fenn | so kepler came along and refactored it, 'its an ellipse you dummies' | 01:29 |
| kanzure | I am impressed + baffled by astronomers | 01:29 |
| kanzure | I keep on looking at star charts and so on | 01:29 |
| kanzure | and I have *NO* clue how these things work | 01:29 |
| kanzure | the planet is freaking *rotating* | 01:29 |
| kanzure | and all of the star charts don't seem to take that into account | 01:30 |
| kanzure | how the hell can they predict where the stars will be no matter where you are in the world? | 01:30 |
| fenn | oh, you can 'factor out' the planets rotation by taking an observation at the same time each day | 01:30 |
| kanzure | doesn't it depend on your location of observin? | 01:30 |
| fenn | or you can generalize the relationship between time and rotation of the sky | 01:30 |
| kanzure | altitude, direction, coordinates, etc.? | 01:30 |
| kanzure | angle of inclination, ... | 01:31 |
| fenn | well, most observatories dont travel around :) | 01:31 |
| kanzure | ah | 01:31 |
| fenn | angle of inclination is what you're measuring | 01:31 |
| kanzure | but wouldn't that depend on your xyz coords on the surface topology of the planet manifold? | 01:31 |
| kanzure | remove the word 'manifold' from that sentence | 01:31 |
| fenn | er.. angle of inclination relative to the horizon depends on the local topology, yes | 01:32 |
| fenn | so usually they fix on the north star or something (i'm not an astronomer) | 01:32 |
| kanzure | ohh | 01:33 |
| kanzure | and then they can use time synchronization via watches and clocks | 01:33 |
| kanzure | thus you should be able to determine what's going on based off of your watch | 01:33 |
| kanzure | and the location of some main orienting ref object | 01:33 |
| kanzure | and then extract your relative differences or something | 01:33 |
| fenn | yep | 01:33 |
| kanzure | never heard any talk of that, though | 01:33 |
| fenn | the other half of astronomy is emission spectrography (or absorption if its a planet) | 01:34 |
| kanzure | right, I can understand that | 01:35 |
| kanzure | easy stuff, that's just astrophysics, photons and red shifts and shifts due to movements etc. | 01:35 |
| fenn | oh, sure, easy | 01:35 |
| fenn | ;) | 01:35 |
| kanzure | much easier when you can do it from orbit | 01:35 |
| fenn | nevermind that it's all LIES | 01:36 |
| kanzure | without atmospheric interference | 01:36 |
| kanzure | is it? | 01:36 |
| fenn | it all seems rather tenuous to me | 01:36 |
| fenn | this has redshift, therefore it must be moving away from us .. well, what if space is non-uniform? | 01:36 |
| kanzure | 'space' itself is tricky stuff | 01:37 |
| fenn | ok, i'll state it differently. what if the speed of light changes depending on where you are | 01:37 |
| kanzure | while I'm not about to suggest the ancient aether theories, I'm not about to suggest that the 'vacuum of space' is a vacuum ... not with the nonzero energy content of the vacuum of space. It's 4 K. | 01:38 |
| kanzure | sure, if it's not truly a vacuum then the speed of light is relative to the environment | 01:38 |
| kanzure | just like we can 'slow' light down in ultracold environments | 01:38 |
| fenn | well, i dont have a telescope so i basically have to take these peoples' word for it | 01:41 |
| kanzure | launching a satellite is peanuts in comparison to most of my ideas, eh? | 01:42 |
| kanzure | the guys in #space are building one | 01:42 |
| kanzure | or maybe it's #astronomy | 01:42 |
| fenn | launching a satellite is 1957 technology | 01:42 |
| kanzure | then why the hell do people complain so much about doingi t | 01:42 |
| kanzure | jackasses | 01:42 |
| fenn | because they dont have the funding of the entire post-war soviet union | 01:43 |
| kanzure | please, with your interest in gingery? | 01:44 |
| kanzure | many rocket enthusiasts do metal shopping and make their own damn rockets by hand | 01:44 |
| fenn | many rocket enthusiasts dont understand the extreme differences between a sounding rocket and an orbital rocket | 01:44 |
| fenn | sure, what's another few decimal places | 01:45 |
| kanzure | heh | 01:45 |
| fenn | and its not just quantitative, it's qualitative because the costs go up exponentially | 01:45 |
| kanzure | costs? | 01:45 |
| fenn | yes, costs: time, energy, motivation | 01:46 |
| kanzure | motivation ... should never cost. | 01:46 |
| fenn | bah you can say that about anything | 01:46 |
| kanzure | "As I move, so I move the universe." - Danlo | 01:46 |
| kanzure | it's up on my whiteboard | 01:46 |
| fenn | time is nature's way of making sure everything doesn't happen at once - what is motivation for then? | 01:47 |
| kanzure | Lion Kimbro would argue time is bullshit | 01:47 |
| kanzure | http://www.speakeasy.org/~lion/ | 01:47 |
| fenn | ask him what we should do next :) | 01:48 |
| kanzure | shit shit shit shit shit | 01:48 |
| kanzure | http://www.davidzindell.com/ | 01:48 |
| fenn | Time is an Invention of Awareness | 01:49 |
| fenn | sounds like it exists then? | 01:49 |
| kanzure | but I was looking for http://hometown.aol.com/pylebc/zindell.htm | 01:49 |
| kanzure | actually another page | 01:49 |
| fenn | bleh.. that pseudo-spiritual stuff doesnt motivate me, unfortunately | 01:50 |
| kanzure | ignore it | 01:51 |
| kanzure | he's otherwise a high signal to noise author | 01:51 |
| kanzure | on which Orion's Arm is mostly based | 01:51 |
| kanzure | ah, here we go | 01:53 |
| kanzure | http://home.austin.rr.com/lperson/fireser.html | 01:53 |
| fenn | oh well, at least he doesnt trivialize a post-human intelligence the way charles stross does in accelerando | 01:58 |
| kanzure | heh | 01:58 |
| fenn | "look its a girl in a box" | 01:58 |
| kanzure | hurray | 01:58 |
| kanzure | hehe | 01:58 |
| kanzure | yeah, I wouldn't mind writing some intense scifi | 01:58 |
| kanzure | I certainly know I can read enough and do a massive review to aggregate enough future tech and philosophy in one gulp | 01:59 |
| fenn | writing is hard, really | 01:59 |
| fenn | not worth it | 01:59 |
| fenn | why do you think there's so much crap on TV | 01:59 |
| kanzure | I find writing to be easy. | 02:00 |
| kanzure | refactoring is what's hard | 02:00 |
| kanzure | and I much rather prefer my stories be told in physical constructions (==> skdb) or functionality rather than empty promises | 02:01 |
| fenn | "IMPRINTING - Changing the metabolic pathways and neural network; redefining the synapses of the brain" this one happens to be more real than most people know | 02:04 |
| kanzure | location? | 02:04 |
| fenn | http://hometown.aol.com/pylebc/zgloss.htm | 02:04 |
| kanzure | just seems to be out of the blue-- ah | 02:04 |
| kanzure | yep | 02:04 |
| kanzure | your thoughts are changing metabolism, technically | 02:04 |
| kanzure | because your thoughts require energy, no? | 02:05 |
| kanzure | the mapping between thoughts and precise, discrete changes in cellular metabolism is interesting | 02:05 |
| kanzure | I think this is where Gary Lynch did his LTP studies, but it wasn't from a subjective POV of thinking of course | 02:05 |
| kanzure | (it was rat training + mem engrams) | 02:05 |
| fenn | its not so much a change in the neural network as a change in the weight-assigning/backpropagation algorithm | 02:06 |
| fenn | they never should have changed it from 'long lasting potentiation' | 02:07 |
| fenn | one of these years i'll learn all the silly neuro jargon | 02:08 |
| -!- Splicer2 [n=p@h174n1c1o261.bredband.skanova.com] has joined #hplusroadmap | 02:11 | |
| kanzure | fenn: I'll write up that list of progs for skdb and so on tomorrow; I've been toying around with ikiwiki tonight and will be getting to work in the morning | 02:12 |
| kanzure | tonight officially has sucked - inattentive + hyperaction => circles | 02:13 |
| fenn | i'm processing the skdb article into something resembling prose | 02:16 |
| kanzure | thank you | 02:20 |
| kanzure | I'll probably join you on that sometime tomorrow | 02:20 |
| -!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has quit ["Leaving."] | 02:21 | |
| -!- Splicer [n=p@h174n1c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 02:25 | |
| -!- Splicer2 [n=p@h174n1c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 03:21 | |
| -!- Splicer2 [n=p@h210n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 03:21 | |
| -!- Splicer [n=p@h210n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 03:42 | |
| -!- Splicer2 [n=p@h210n2c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 04:01 | |
| -!- Splicer2 [n=p@h210n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 04:26 | |
| -!- Netsplit leguin.freenode.net <-> irc.freenode.net quits: Splicer | 04:28 | |
| -!- Splicer [n=p@h210n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 05:40 | |
| -!- Splicer2 [n=p@h210n2c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 05:55 | |
| -!- Splicer2 [n=p@h70n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 05:55 | |
| -!- Splicer [n=p@h210n2c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 06:05 | |
| -!- Splicer [n=p@h62n3c1o261.bredband.skanova.com] has joined #hplusroadmap | 06:51 | |
| -!- Splicer2 [n=p@h70n2c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 07:06 | |
| -!- Splicer2 [n=p@h28n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 10:07 | |
| -!- Splicer [n=p@h62n3c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 10:19 | |
| -!- Splicer [n=p@h104n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 10:29 | |
| -!- Splicer2 [n=p@h28n2c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 10:41 | |
| -!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has joined #hplusroadmap | 10:44 | |
| kanzure | http://www.facebook.com/event.php?eid=17728995118 <-- Austin RepRap group with a n | 10:57 |
| kanzure | http://www.actlab.utexas.edu/ | 10:57 |
| kanzure | Okay. Back to work. | 11:00 |
| kanzure | http://math.utexas.edu/users/henrys/waystogowrong.html <-- depressing | 11:11 |
| kanzure | http://heybryan.org/mediawiki/index.php/Societal_engineering_knowledge_database#todo | 11:15 |
| -!- Splicer2 [n=p@h104n2c1o261.bredband.skanova.com] has joined #hplusroadmap | 11:33 | |
| kanzure | Hey Splicer2. | 11:44 |
| -!- Splicer [n=p@h104n2c1o261.bredband.skanova.com] has quit [Read error: 110 (Connection timed out)] | 11:45 | |
| kanzure | fenn: I think we need to split agx-make into some subprograms. | 11:53 |
| kanzure | fenn: agx-make is going to be somewhat of a nightmare. Do we want scripts to do API calls that would interface with the machinery? Or do we want to compile a final script/program to be executed over the particular fablab machinery network? | 12:19 |
| kanzure | the compilation case does not allow the 'human readable output' alternatives (except if there's a terminal/printer that could be programmed to spit out the instructions ... hrm.) | 12:20 |
| kanzure | I asked in #robotics for some help with the ideas of configuration of a local setup ... yeah. They just went off on random rants. | 13:31 |
| kanzure | I guess we can enforce configurations upon users. "To execute this script, the lab has to be moved to position-style-31410, proceed? WARNING: THIS WILL MAKE YOUR MACHINERY MOVE!". And then if they are motor-powered, they will wheel around into a particular config. | 13:37 |
| kanzure | As for the API calls. The 'api layer' thing. Conveniently, all of your drivers *should* be up on skdb for others to download, or to at least include in revisions of the 'api layer' software for interfacing with the physical network. How would it work? How does the printing server usually work? | 13:39 |
| kanzure | http://cups.org/ | 13:42 |
| kanzure | 'CUPS uses the Internet Printing Protocol ("IPP") as the basis for managing print jobs and queues and adds network printer browsing and PostScript Printer Description ("PPD") based printing options to support real-world printing.' | 13:42 |
| kanzure | heh, they describe it as real-world printing. | 13:42 |
| kanzure | http://www.cups.org/documentation.php/overview.html | 13:44 |
| kanzure | http://www.cups.org/documentation.php/spec-design.html even better | 13:45 |
| kanzure | hm, ##cups is a pretty small group. | 13:49 |
| kanzure | How does cups interface with drivers? | 13:55 |
| kanzure | Linux kernel 2.6 added 'Vendor neutral virtualisation API' - what does this mean? | 14:05 |
| kanzure | and why did bluetooth support have to be released with an entirely new version of the kernel (2.4)? that doesn't sound right | 14:06 |
| kanzure | `Every version of a device, such as a printer, requires its own specialized commands. In contrast, most applications access devices (such as sending a file to a printer) by using high-level, generic commands, such as PRINTLN. The driver accepts these generic statements and converts them into the low-level commands required by the device.` | 14:08 |
| kanzure | http://devresources.linux-foundation.org/dev/opendrivers/wiki/index.php/Documentation | 14:15 |
| kanzure | "During startup, a driver registers with its bus, and the bus inserts the driver | 14:21 |
| kanzure | into its internal list of drivers. The bus then attempts to bind the driver to the | 14:21 |
| kanzure | bus's list of devices. A bus compares a list of bus-specific device ID numbers that it supports with the ID numbers | 14:21 |
| kanzure | of the devices on the bus. If a match is made, the device is 'attached' to the driver. The driver allocates a driver-specific | 14:21 |
| kanzure | object to describe the device. This usually includes a class-specific object as well, | 14:21 |
| kanzure | which the driver uses to register the device with it's class" | 14:21 |
| kanzure | http://wiki.kernelnewbies.org/KernelMentors | 14:23 |
| kanzure | the guys in #kernelnewbies on oftc recommended http://www.lantronix.com/device-networking/embedded-device-servers/xport.html | 15:10 |
| kanzure | but otherwise sat quietly. | 15:10 |
| kanzure | how do you interface with the hardware? let's say a driver is written, then what? do you sit on /dev/ and listen with a pipeline? | 15:11 |
| kanzure | looks like cupsd is based around the idea of maintaining the status quo of functionality (I am not using 'status quo' in the negative connotation here) | 15:13 |
| kanzure | huh, looks like you just echo > /dev/you and cat /dev/you to get information. http://www.freesoftwaremagazine.com/articles/drivers_linux?page=0%2C6 | 15:19 |
| kanzure | I guess you can then have a message passing protocol to be implemented | 15:20 |
| kanzure | userpsace example - http://www.freesoftwaremagazine.com/articles/drivers_linux?page=0%2C10 | 15:26 |
| kanzure | yep, just reading back and forth from /dev/deviceshere | 15:26 |
| kanzure | hm | 15:26 |
| kanzure | (the guys in #kernelnewbies suggested that it be all done over ethernet or usb, so that it's already in userspace -a fine idea - but it needs to be generally extensible) | 15:27 |
| kanzure | so I think it might be a good idea to go back to the 'API layer' idea, having a 'cupsd' equivalent to run things, with extendibility such that new 'protocols' can be implemented when necessary, and then we just hold general tools to a standard with an iron fist | 15:28 |
| kanzure | On Saturday 19 April 2008, Stefano Vaj wrote: | 15:30 |
| kanzure | > While I may have no final view on the subject, it should be | 15:30 |
| kanzure | > underlined that *both* points of view are equally "eugenetic". That | 15:30 |
| kanzure | > of the deaf community and parents who want to ensure conformity of | 15:30 |
| kanzure | > their offspring to their own Idealtypus, and that of the broader | 15:30 |
| kanzure | > community who feels that would rather have it conform with a norm | 15:30 |
| kanzure | > where deafness is fundamentally a disability, and little else. And I | 15:30 |
| kanzure | > do not believe that a *political* and *aesthetic* choice between | 15:30 |
| kanzure | > these two conflicting views can really be avoided through reference | 15:30 |
| kanzure | > to some general governing principles. | 15:30 |
| kanzure | re: the iron fist; protocol-software can be stored in skdb as packages themselves (just no physical expression). I wonder, to what extent do the tools get to control their own state, versus management of state on the server? Part and part? Not all tools are going to manage their state, and likewise some tools will manage their state more thoroughly than others; so cupsd has to take this into account. | 15:33 |
| kanzure | the abstraction that cupsd facilitates is very nifty - you just press the 'print' button and it takes care of everything else. | 15:34 |
| kanzure | kind of hard to imagine how the /dev interface is supposed to work when I don't have any tools off the top of my head | 15:39 |
| kanzure | and we get to design the tools too, that's always useful | 15:39 |
| kanzure | but on the other hand, interfacing with other, older machinery would be useful as well | 15:39 |
| kanzure | this isn't exactly the most standardized project though - many factories have completely custom-made solutions | 15:40 |
| kanzure | ah, the drivers manage the state information | 15:44 |
| kanzure | and then the userspace programs just get data back and forth | 15:44 |
| kanzure | maybe. distinction of where to put what for userspace/kernelspace there is hazy. | 15:44 |
| kanzure | okay, so we'd have code in skdb that could be used to interact with particular machines, or the user can code up his own driver and userspace program that can be interfaced with the 'cupsd equivalent' | 15:46 |
| kanzure | now, agx-make and cupsd have to know about the local configuration - what machinery is connected related to the fablab, and what isn't? When there isn't something available, known to it, then it should default to 'human instruction mode' - where should it dump the instructions? Should agx-make dump the instructions state to the terminal screen? | 15:46 |
| fenn | hi. still reading backlog | 15:52 |
| fenn | printer analogy can only go so far. a printer will always use ink cartridges, will always print on paper, will always print 2d documents that eventually end up as a high-res bitmap | 15:54 |
| fenn | all printers do this | 15:54 |
| fenn | manufacturing technology is a lot more complex | 15:55 |
| fenn | it might work for DNA reader/writer hardware | 15:55 |
| fenn | i dont want to require any specific hardware like xport or even x86 processors | 15:56 |
| fenn | that's part of the hardware 'infrastructure' that is unique to each user | 15:56 |
| kanzure | sure | 15:57 |
| fenn | state changes have an associated cost, which influences the optimization path | 15:58 |
| fenn | buying a new machine is a change in state | 15:59 |
| kanzure | no, I was talking about 'state change' in terms of what the server knows about the hardware, and what the hardware knows about itself (what position it is, for example) | 15:59 |
| fenn | you might be able to hack the drill press into a milling machine, but you could also just buy a milling machine | 15:59 |
| kanzure | or like the linux device drivers - the driver can maintain an internal state so that it knows what to expect from the port on the next interrupt or whatever | 15:59 |
| fenn | oh, some printers are smarter than others, lets leave it at that | 16:00 |
| kanzure | I see, so how does cupsd manage that? | 16:01 |
| kanzure | I guess each new driver module just hides that 'smartness' away and lets cupsd work with it like any other driver | 16:01 |
| fenn | judging by the vast array of printer drivers, each printer has its own driver | 16:01 |
| kanzure | so it might be hiding or adding fake functionality | 16:01 |
| fenn | right | 16:01 |
| fenn | not necessarily fake functionality, just an abstraction layer | 16:02 |
| fenn | wherever you have an interface there's an abstraction layer associated with it | 16:02 |
| fenn | this applies to physical objects too? | 16:02 |
| kanzure | eh? | 16:02 |
| fenn | nevermind, too philosophical for me | 16:03 |
| kanzure | okay, so when I was talking about my 'api layer' earlier, I guess I can convert that stuff into saying that it is the job of the driver to meet the standard specs for cupsd to work with ... then, all skdb scripts just implement one unified language for manufacturing | 16:03 |
| kanzure | 'one unified language for manufacturing' or api calls | 16:03 |
| kanzure | somebody suggested | 16:03 |
| kanzure | tool.arm.moveLeft(90) | 16:03 |
| kanzure | but that doesn't sound like a smart idea | 16:03 |
| kanzure | that's very very messy | 16:03 |
| fenn | no, that doesnt work, you try to make 'one unified language for manufacturing' and end up with a monster that nobody can use | 16:04 |
| kanzure | (also, 3D object interpretation programs should be able to generate a list of movements or whatever) | 16:04 |
| kanzure | right | 16:04 |
| kanzure | so it has to be something else | 16:04 |
| kanzure | you have much more experience in this area | 16:04 |
| fenn | the STEP file format is a real-world example of trying to make one unified language for manufacturing | 16:04 |
| fenn | i like the 'rhizopod' idea | 16:05 |
| fenn | http://www.lantronix.com/device-networki | 16:05 |
| fenn | not that | 16:06 |
| fenn | http://scratchcomputing.com/projects/vectorsection/ | 16:06 |
| fenn | another graph :) | 16:06 |
| kanzure | hm | 16:07 |
| kanzure | I see many ways that we can implement reprogrammable factories | 16:07 |
| kanzure | most of the tools will have microcontrollers, thus we can do a massive recompiling of everything down to that code, from parallel code down to the physical substrates | 16:07 |
| kanzure | well | 16:07 |
| kanzure | screw it, I guess we're going to have to demand standardization of shop setups | 16:08 |
| kanzure | just like the 'debian definition' is very specific | 16:08 |
| kanzure | I'd like to avoid that ... but I don't see another way. | 16:08 |
| fenn | there is a subset of g-code that is conserved across all industry, that's the g1 command (and possibly g2/g3) | 16:08 |
| kanzure | this means that "you *must* have an arm that passes these standardization codes, and it *must* be in this location, and it must be able to move at this rate and with this precision, etc." | 16:09 |
| fenn | G1 X0 F60 means goto x 0 at 60 inches or mm per minute | 16:09 |
| kanzure | uh-huh | 16:09 |
| fenn | thats a lot of musts.. i only know one person with a functioning robot arm | 16:10 |
| kanzure | oh, I mean a general arm | 16:10 |
| kanzure | like the ones we see in many industrial manufacturing operations, esp. automobile industry | 16:10 |
| fenn | yea, a robot arm | 16:11 |
| fenn | oh wait, two | 16:11 |
| fenn | they would both laugh at our idea, so you see, um.. i dont know what i'm saying | 16:11 |
| fenn | industry is very conservative | 16:12 |
| fenn | it's why people still use g-code | 16:12 |
| kanzure | I'm guessing the real reason is because they don't have young kids coming in to redesign the factories | 16:12 |
| kanzure | I bet they think it's impossible | 16:12 |
| kanzure | + there's not enough people wanting completely new factories from the ground up | 16:12 |
| fenn | the reason japan and germany are so successful is because we bombed the shit out of all their old factories and they were forced to build new ones | 16:13 |
| kanzure | and open source solutions are long-term, reconfiguration being needed for updates | 16:13 |
| kanzure | FPGAs for factories | 16:14 |
| fenn | indeed | 16:14 |
| kanzure | so. | 16:15 |
| kanzure | consensus? | 16:15 |
| kanzure | oh | 16:15 |
| fenn | its not going to be a viral meme that spreads through industry, they're going to laugh at the whole idea of autogenix until they're forced to scramble and catch up | 16:16 |
| kanzure | sure | 16:16 |
| kanzure | I don't care about industry | 16:16 |
| fenn | ok, good | 16:16 |
| kanzure | (I mean, it would be great if they did want to contribute, of course) | 16:16 |
| kanzure | but I was asking for a consensus on what to do with our cupsd clone/hack | 16:17 |
| fenn | consensus means everyone agrees | 16:17 |
| kanzure | do we want to force 'standardization' (with all of those musts) to make distribution variants for shops? "Yep, I'm running an xyz flavor of shop." | 16:17 |
| kanzure | else, how do we make this reconfigurable such that any package could specify instructions for no particular implementation | 16:18 |
| fenn | standardization is just another word for black box | 16:18 |
| kanzure | distributions would have very specific rules | 16:18 |
| kanzure | "there is always a type-354r14 class 1294 arm at coordinates 341, 414" | 16:19 |
| fenn | that sounds too fragile, all DLL-hell and no duck-typing | 16:19 |
| kanzure | right, but I don't know how to do the alternative quite yet | 16:19 |
| kanzure | oh | 16:19 |
| kanzure | I guess any machinery that you *do* add, must be in skdb and fully specced out | 16:19 |
| fenn | first of all, why does the arm have to be at 341,414 | 16:19 |
| kanzure | needs to be out of the way and have its own spot | 16:20 |
| kanzure | then the skdb program would run | 16:20 |
| kanzure | and move it around to do its thing for that particular project | 16:20 |
| kanzure | *particular package | 16:20 |
| kanzure | (I don't like it either.) | 16:20 |
| fenn | neither of those things have anything to do with the function of an arm, which is to position objects | 16:20 |
| fenn | the arm has to be relative to something else, (or be not in the way, relative to seomething else) | 16:21 |
| kanzure | right, ideally that would be a movable arm, so that you could get cargo and move it from one area to the next | 16:21 |
| kanzure | wheels or something, I don't care | 16:21 |
| fenn | that's more than an arm | 16:21 |
| kanzure | anyway, we both know we hate this implementation | 16:21 |
| kanzure | so back to the reconfigurable one | 16:21 |
| kanzure | like I said - any machine you do add to your setup, you have to add an entry into skdb and also tell cupsd that you have it (part of the install / setup process, easy) | 16:22 |
| kanzure | then, you have to find a 'close enough class' that is already implemented | 16:22 |
| fenn | skdb is social knowledge, so you'd only add an entry if nobody else had one quite like it | 16:22 |
| kanzure | correct | 16:22 |
| kanzure | if it's already there, you'd take it | 16:22 |
| fenn | ok | 16:22 |
| kanzure | which would download a driver for you | 16:23 |
| kanzure | but if you have a proprietary instrument, though | 16:23 |
| kanzure | things are going to be different | 16:23 |
| fenn | define proprietary | 16:23 |
| kanzure | not open source, you have to consider the driver a blackbox | 16:23 |
| fenn | it does have some specifications though | 16:23 |
| kanzure | how so? | 16:23 |
| fenn | you know, 50 pages/min, 600dpi, requires 110VAC, etc | 16:24 |
| fenn | and presumably the company provides instrutions for how to use the software | 16:24 |
| fenn | (or else we're in an nvidia mess) | 16:24 |
| kanzure | if you get those specs from the company, then you have to hack together the interface with cupsd to that driver | 16:26 |
| kanzure | now, there's one part missing I think | 16:26 |
| fenn | ok, so besides not knowing how to build the thing, what else do we have to worry about? | 16:26 |
| kanzure | you have to be able to interpret the code in (skdb) relative to the new tool you have | 16:27 |
| kanzure | this would be a plugin to agx-make, right? | 16:27 |
| kanzure | erm | 16:27 |
| kanzure | your new tool exposures some general functionality | 16:27 |
| kanzure | once you tell agx-make how to access that functionality, | 16:27 |
| kanzure | you need to tell your computer how to interpret requests for that functionality from (.skdb) packages that are running their code | 16:28 |
| kanzure | does that make sense? | 16:28 |
| fenn | there would be some server running, listening for requests to make stuff | 16:28 |
| fenn | not sure i see the problem? | 16:29 |
| fenn | you to describe the capabilities of the machine? | 16:29 |
| fenn | s/you/how/ | 16:29 |
| kanzure | sigh | 16:29 |
| kanzure | example: skdb package chair | 16:30 |
| kanzure | chair has some standardized code to make this chair, it's in the file make_chair.py or something | 16:30 |
| kanzure | (the file name was mentioned in the metadata file / skdb file or whatever) | 16:30 |
| kanzure | make_chair.py has to tell it how to make the chair - it can say "here's this 3D file, make this" | 16:30 |
| kanzure | or it can say other things too like "here's how to specifically maneuver about to make this" | 16:30 |
| kanzure | now, the specific maneuvering is *general* and *abstract* since it shouldn't ever assume you have specific hardware | 16:31 |
| fenn | 3d file = cad, specific maneuvers = cam output | 16:31 |
| kanzure | yeah, but what about the overall manufacturing circuit? | 16:31 |
| fenn | cam is the process of taking a cad file, looking at the machine you have, and making specific maneuvers | 16:31 |
| fenn | traditional cam software is very very stupid | 16:32 |
| kanzure | okay, so cam would part of cupsd | 16:32 |
| kanzure | *would be part of | 16:32 |
| kanzure | also, it has to know what the current state of the manufacturing process is at | 16:32 |
| fenn | it usually doesnt know *anything* and you have to tell it how to do everything, one step at a time | 16:32 |
| kanzure | so it would have to know that 'we have a partial chair sitting on platform A223' | 16:32 |
| kanzure | but you can't do that though | 16:32 |
| kanzure | you have various hardware implementations | 16:32 |
| kanzure | otherwise we get to those 'too many musts' | 16:32 |
| fenn | why would it have to know we have a partial chair? cant it just know that there was a job running and got interrupted? | 16:33 |
| kanzure | sure | 16:36 |
| fenn | traditional robots are blind and forgetful | 16:36 |
| kanzure | sigh | 16:36 |
| kanzure | you dont' seem to get it thoguh | 16:36 |
| fenn | you're right | 16:36 |
| kanzure | you keep on saying that they are blind and stupid and that is true | 16:36 |
| kanzure | but the issue is that the hardware is different | 16:36 |
| kanzure | my setup is not necessarily the same as your setup (we have established this -- otherwise "too many musts" etc) | 16:37 |
| fenn | the reason they are blind and stupid is that the software to make them otherwise is hard to write | 16:37 |
| kanzure | so the hardware implementation of all of the tools and machinery has different protocols | 16:37 |
| kanzure | and the simplification layer doesnt' necessarily know how to connect the (skdb) calls to the driver calls for the machinery | 16:37 |
| fenn | it doesnt? | 16:37 |
| kanzure | no, how would it? | 16:38 |
| fenn | simplification is 3d file => sequence of movements, right? | 16:38 |
| fenn | yes/no? | 16:38 |
| fenn | assuming it is, then each piece of hardware has some metadata provided by a skdb package that is installed on the user's system | 16:39 |
| fenn | and that metadata describes the functionality of each machine | 16:39 |
| fenn | so there are rules coded in other skdb packages explaining how to go from raw lumber to radially symmetrical objects such as chair legs | 16:40 |
| fenn | and other rules explaining assembly of chair legs into chairs | 16:40 |
| fenn | you dont necessarily need a lathe to make a chair leg, but it's the most efficient method (in terms of time, cutter wear, surface finish) | 16:41 |
| fenn | say you had a cnc milling machine, you could mill a bas-relief chair leg into a block of wood, then flip it over and do the other side | 16:42 |
| fenn | net result: radially symmetric object | 16:42 |
| fenn | clamping a round shape is a little tricky but it will work | 16:43 |
| fenn | figuring out the clamping is usually the most costly and time-consuming part of a cnc job | 16:44 |
| fenn | so its really strange that there is no automated software for it | 16:44 |
| fenn | mental map of workpiece holding systems: http://www.emeraldinsight.com/fig/0330220201017.png | 16:52 |
| kanzure | "and that metadata describes the functionality of each machine" | 16:53 |
| kanzure | huh | 16:53 |
| kanzure | I thought we agreed that you can't formally describe functionality | 16:53 |
| fenn | i never said that, you keep saying it and i disagree | 16:56 |
| fenn | that's why i say each package has a provides: section | 16:56 |
| kanzure | how would that be possible | 16:58 |
| kanzure | klfajsfdkjasl;fjmal;dkfjaskj | 16:58 |
| kanzure | provides is not the same thing as functionality | 16:58 |
| kanzure | the program itself is the functionality | 16:58 |
| kanzure | and is explicitly defined by the code | 16:58 |
| kanzure | you have never disagreed with me on the inability to search for software except by natural languages | 16:59 |
| fenn | there's no classification system for software functionality | 16:59 |
| fenn | however, there are classification systems for physical functionality | 17:00 |
| fenn | example: transportation | 17:00 |
| fenn | if there were no classification, you'd have to say, "find me a thing.. that can move stuff from one location to another" | 17:00 |
| fenn | then you'd have to go through all the ambiguities of that statement | 17:01 |
| fenn | define 'stuff' and 'one location' etc | 17:01 |
| kanzure | I don't understand how that helps the situation | 17:01 |
| fenn | because cars and trains and airplanes are classified as transportation | 17:01 |
| kanzure | so what? | 17:02 |
| fenn | so you can search for transportation | 17:02 |
| kanzure | that's social knowledge | 17:02 |
| fenn | yes | 17:02 |
| kanzure | that's the other part of skdb | 17:02 |
| kanzure | we're discussing the manufacturing code | 17:02 |
| fenn | ok, there's also "3 axis subtractive machining" | 17:02 |
| fenn | is that not a functionality? | 17:03 |
| fenn | a lathe is 2-axis machining, or single-point turning | 17:04 |
| kanzure | you are hurting me :( | 17:04 |
| kanzure | please explain what the .skdb files would have for their manufacturing code in that scenario | 17:04 |
| kanzure | and how would that interface with the random-ass hardware that you happen to have in your shop environment | 17:04 |
| fenn | chair would have 3d cad files showing a radially symmetrical object and describing its material (wood) | 17:06 |
| fenn | algorithm sees that you have lathe installed, which provides 2-axis-machining, it consults the libmachining to find that yes, you can make radially symmetrical objects from wood with 2-axis-machining | 17:07 |
| fenn | algorithm also sees that you have mill installed, with provides 3-axis-machining | 17:08 |
| fenn | this can also make radially symmetrical objects, but the cost is higher | 17:08 |
| kanzure | what is 'algorithm' - is it agx-make? the python scripts for that particular skdb package? | 17:09 |
| fenn | so the algorithm chooses lathe and invokes CAM routines to cut the part on a lathe, then passes the job off to the cupsd layer | 17:09 |
| kanzure | wait, what? | 17:09 |
| fenn | its the program that traverses the big dependency tree describing everything | 17:09 |
| kanzure | okay, so this is 'job-creation routine' | 17:09 |
| kanzure | so agx-make has two parts: (1) job maker/builder, and (2) the scheduler/head-of-the-octopus | 17:10 |
| fenn | you have to pretend to build the job before you can decide if it's worth making (unless there are heuristics available) | 17:11 |
| fenn | for all intents and purposes pretending to build it is the same as building it | 17:11 |
| kanzure | pretending to build == simulation? | 17:11 |
| fenn | yes | 17:11 |
| kanzure | or do you mean 'in your head, does it make sense'? | 17:11 |
| kanzure | okay | 17:11 |
| kanzure | so | 17:12 |
| fenn | i mean, you'd invoke cam routines, see how long it takes by adding up all the movements and feed rates | 17:12 |
| kanzure | how would the algorithm see that you have lathe installed? you need a local description of everything you have in your shop, obviously, and then cross-referenced to skdb out on the internet? | 17:12 |
| fenn | yes, except for the internet part | 17:13 |
| fenn | its a local 'installation' | 17:13 |
| fenn | metadata for every piece of hardware in your shop | 17:13 |
| fenn | maybe you need cad schematics too | 17:14 |
| fenn | models, not schematics | 17:14 |
| kanzure | it's a local installation, but ideally your parts are described on the internet, no? the only reason they would not be would be if (1) they are proprietary and to some extent closed, or (2) you haven't gotten around to doing a writeup ;) | 17:15 |
| fenn | its not as hard as it sounds, most lathes are more or less identical, but with a few dimensions different | 17:15 |
| fenn | the same with drill bits, bolts, etc | 17:15 |
| fenn | sometimes it can cost less to buy something standardized than to try to model some old piece of junk | 17:16 |
| fenn | but if people want to model junk, that's fine too | 17:17 |
| fenn | depends on how much you value your time | 17:17 |
| fenn | so, do you agree that it's possible to describe functionality? | 17:18 |
| kanzure | the description is only 'social' in nature | 17:19 |
| kanzure | and I'm fine with that, as long as we keep that straight | 17:19 |
| kanzure | it's just kind of unfortunate | 17:20 |
| kanzure | oh, I get it now | 17:20 |
| kanzure | you have to know what you have built | 17:20 |
| kanzure | heh | 17:20 |
| fenn | the description '2 axis machining' is a kind of black-box | 17:20 |
| kanzure | in terms of what others have previously built (so that you can classify it properly) | 17:20 |
| kanzure | yeah, if you built a '2 axis machining' thing and didn't know it, good luck trying to hook up with the general api or whatever | 17:20 |
| kanzure | although you seem to be arguing that skdb packages would not specify their own 'agx-make' code | 17:20 |
| fenn | you can build a physics simulation of the cutting process, but it would take huge amounts of computer power to run that on every package | 17:21 |
| fenn | there's got to be some way to automatically generalize, but i dont know how to do it | 17:21 |
| fenn | some skdb packages would be processes only, not physical objects | 17:22 |
| fenn | these would be mostly code | 17:22 |
| fenn | so the chair package could specify 2-axis-machining, or they might be lazy and let the computer figure it out | 17:23 |
| fenn | i think i'd prefer the lazy route since it should reveal the most optimal path | 17:24 |
| fenn | you dont _need_ a lathe to make a chair | 17:24 |
| fenn | there are a lot of inflexible ways of thought.. we call them culture :) | 17:25 |
| kanzure | huh? | 17:25 |
| fenn | "everyone knows that chair legs are made on a lathe" | 17:26 |
| fenn | then you look at my extruded plastic thing and dont know what it is | 17:26 |
| fenn | or something like this http://www.n55.dk/MANUALS/DYNAMIC_CHAIR/CHAIR.html | 17:27 |
| fenn | the dynamic chair is very efficient in its manufacturing process because it uses the same tools they use to build everything else | 17:28 |
| fenn | it's probably not a good choice if you dont have access to a sheet metal brake | 17:28 |
| fenn | now, i'm not saying the computer will know how to come up with this innovative way of making a chair, because that requires generalizing the concept of what a chair does | 17:30 |
| fenn | but you could program a 'fitness function' for 'chairness' and let a genetic algorithm pull bits and pieces from skdb | 17:31 |
| kanzure | huh?? | 17:32 |
| fenn | have you seen this page? http://ccsl.mae.cornell.edu/ | 17:32 |
| kanzure | so | 17:33 |
| kanzure | even though we have an exact chair description and so on | 17:33 |
| kanzure | we still use a genetic algo?? | 17:33 |
| fenn | an exact chair description is brittle | 17:34 |
| fenn | if you dont have all the pieces required for building the chair exactly the way it's specified, you're stuck | 17:34 |
| kanzure | why do you need a genetic algorithm though | 17:35 |
| fenn | if you have a genetic algo for building chairs, you can automatically make the chair out of whatever you have (maybe) | 17:35 |
| kanzure | you've completed destroyed my entire concept of wtf this thing is going to be doing | 17:35 |
| kanzure | :) | 17:35 |
| fenn | i know :( | 17:35 |
| kanzure | let's just assume that you have the right materials | 17:35 |
| fenn | if you have the exact tools and the exact materials you can make the exact replica | 17:36 |
| fenn | i just dont think black boxing is all that useful | 17:36 |
| fenn | it leads to DLL-hell | 17:36 |
| fenn | it leads to everything we associate with computers | 17:36 |
| kanzure | so I don't know what's going on now | 17:37 |
| kanzure | what's the point of adding .skdb packages if they don't tell you anything | 17:37 |
| fenn | but they do tell you things | 17:37 |
| fenn | say we have a package gen-chair | 17:38 |
| fenn | gen could stand for generic or genetic algorithm | 17:38 |
| fenn | it's a generalized chair | 17:38 |
| kanzure | right? | 17:38 |
| kanzure | so it has to tell you how to make it still | 17:39 |
| fenn | you feed it _specific_ things you have, it pieces them together in random ways and tries to improve on its design, where improve is based on the fitness function | 17:39 |
| fenn | then it dumps its generated model into some manufacturable format | 17:39 |
| kanzure | oh, right | 17:39 |
| kanzure | because the items in skdb aren't necessarily the ones you locally have | 17:39 |
| fenn | the manufacturability would be part of the fitness function i guess | 17:39 |
| kanzure | so the piecing-together and scoring function would involve autospec again | 17:39 |
| kanzure | to make sure the units resolve | 17:40 |
| kanzure | and that you have the local resources etc. | 17:40 |
| kanzure | (inventory) | 17:40 |
| kanzure | then the manufacturing format - this still needs some work though ... | 17:40 |
| fenn | yep | 17:40 |
| fenn | i'm sure there's a manufacturing ontology out there already | 17:41 |
| fenn | like http://efunda.com/ | 17:41 |
| fenn | unfortunately they have no quantitative analysis of each process | 17:43 |
| kanzure | the manufacturing format would the 'job' | 17:44 |
| kanzure | which cupsd/equivalent would schedule, queue up and execute | 17:45 |
| fenn | cups has it easy for printing because there's no optimization possible | 17:45 |
| fenn | its the same thing every time | 17:46 |
| kanzure | all optimization is done during the job-production step | 17:46 |
| kanzure | so by the time it gets to cupsd, no optimization is needed | 17:46 |
| fenn | yeah but there is inter-job optimization too | 17:46 |
| kanzure | hm | 17:46 |
| kanzure | oh, that's true | 17:46 |
| fenn | like, while the furnace is still hot, lets do these other castings quick | 17:46 |
| kanzure | okay, so the manufacturing-format has what optimized? | 17:46 |
| kanzure | it's optimized to local tools, machinery, etc., right? | 17:46 |
| fenn | optimized to the end user | 17:46 |
| kanzure | 'tailored' for the end user | 17:47 |
| kanzure | is more like ti | 17:47 |
| kanzure | and then cupsd would do job-optimization, then job-scheduling/queuing. | 17:47 |
| kanzure | job optimization sounds kind of hard | 17:47 |
| kanzure | I am imagining a yaml format for graphing it out physically | 17:47 |
| kanzure | where nodes are processes in the manufacturing steps | 17:47 |
| kanzure | or something? | 17:47 |
| kanzure | and then it would optimize the movement of the object from one place to another? | 17:47 |
| kanzure | does that fit what you were thinking? | 17:47 |
| fenn | optimization would occur at each level of blackboxing (abstraction) | 17:48 |
| kanzure | http://www.bathsheba.com/ artist doing 3d metal printing | 17:48 |
| kanzure | what does that mena | 17:48 |
| kanzure | *mean | 17:48 |
| kanzure | you said basically nothing to me there :) | 17:48 |
| fenn | so as you get closer to sequences of movements, you shuffle things around after each expansion step | 17:48 |
| fenn | first you knead the bread, then you bake the bread, then you eat the bread | 17:49 |
| kanzure | oh, so you mean 'optimize the kneading process' ? | 17:49 |
| kanzure | how do you know that it does not violate something that has to occur during the kneading? | 17:49 |
| kanzure | http://www.bathsheba.com/sculpt/process/index.html#3dprint | 17:49 |
| fenn | there are rules explaining what you can and cant do | 17:50 |
| fenn | or, you could define a fitness function for 'kneaded bread' and let it try all the permutations :P | 17:51 |
| Splicer2 | i'm guessing you've seen the open source CAM portion of this: http://www.editthis.info/opencam/Main_Page | 17:51 |
| -!- Splicer2 is now known as Splicer | 17:52 | |
| fenn | yes i wrote half of that | 17:52 |
| Splicer | hehe | 17:52 |
| kanzure | fenn: how depressing (re: last msg) | 17:52 |
| kanzure | I mean, that it's such a small number of people that might have a clue of what's going on | 17:53 |
| kanzure | fenn: how would you describe the rules for allowed/unallowed edits to the process? I guess you can have a 'no rewrite zone' and then stuff that you don't care about, but ... | 17:54 |
| kanzure | give me a reason why we want 'job optimization' and evidence that it is possible | 17:54 |
| fenn | uhhh.. what? | 17:54 |
| kanzure | it's already tailored to your machines, your layout, your resources, ... (from the other stage) | 17:54 |
| kanzure | (from the manufacturing-format generator) | 17:55 |
| fenn | because otherwise the computer will do things in what appears to be the stupidest order possible | 17:55 |
| kanzure | wasn't that programmed originally by the skdb package maintainer? | 17:55 |
| kanzure | oh | 17:57 |
| kanzure | no, I suppose not | 17:57 |
| kanzure | since your configuration/layout for your machines is different | 17:57 |
| fenn | what OS are you running, debian or ubuntu? | 17:57 |
| kanzure | deb | 17:57 |
| kanzure | so you said to blackbox it all, and then just optimize each step really | 17:57 |
| kanzure | so I'd call this layout optimization | 17:57 |
| fenn | would you mind installing emc2-sim so i can demonstrate some stuff? | 17:57 |
| kanzure | or optimization due to layout | 17:57 |
| kanzure | let me check | 17:58 |
| fenn | its not in the repo's you have to add a repo to /etc/apt/sources.list | 17:58 |
| kanzure | ugh | 17:58 |
| kanzure | can we do it later? I'm completely willing, but going off to do some random stuff in a few moments | 17:58 |
| kanzure | re: optimization at the cupsd stage - what is there beyond 'layout optimization' (obviously you want to choose the shortest possible path, don't be criss-crossing all over the shop) | 17:59 |
| kanzure | and I'm gone for a while or something | 17:59 |
| fenn | ok, emc stuff later | 18:00 |
| fenn | layout optimization is really about conserving time and energy | 18:00 |
| fenn | keeping the oven warm for the next batch is conserving energy | 18:01 |
| fenn | but if you try to do everything in one big batch, you might end up missing some deadline (or just take 'too long') | 18:01 |
| kanzure | huh? | 18:02 |
| fenn | you can boil time and energy and materials down into dollars, and then just optimize for dollars | 18:04 |
| fenn | but there are constraints, you might not have enough (time energy materials) | 18:04 |
| fenn | there may be no solution | 18:04 |
| fenn | usually though, there are too many solutions | 18:05 |
| kanzure | so it's another GA layer of optimization? | 18:05 |
| kanzure | what are the ways of permutation and combination for this layer of GAs? | 18:05 |
| kanzure | obviously it's not manufacturing-formt | 18:05 |
| kanzure | *format | 18:05 |
| kanzure | but instead it's the options specific to using the machines | 18:05 |
| kanzure | or something like that? | 18:05 |
| kanzure | you mentioned 'culture' earlier | 18:05 |
| fenn | i was imagining it as a highly structured yaml file that gets boiled down into a flat sequence of movements | 18:05 |
| fenn | and at each stage of 'boiling' you rearrange the pieces | 18:06 |
| fenn | you could put multiple jobs into the same big yaml file at the start | 18:06 |
| fenn | or maybe just keep dumping them in at the bottom of the file, like a queue | 18:07 |
| fenn | i like files because i can interrupt the system and easily see what is going on | 18:09 |
| fenn | there are lots of tools for inspecting and manipulating files and i already know how to use them | 18:10 |
| fenn | anyway, 'culture' is not just what you have, its what you have to think with | 18:11 |
| fenn | you can build a new machine if you can think it | 18:11 |
| Splicer | I was thinking about that today... concerning this bioDIY thing, I think people view it differently in different countries. | 18:26 |
| Splicer | I think it doesn´t feel as threatening in Canada for instance, that it lacks that tint. | 18:28 |
| fenn | sure, they didnt have to listen to endless news stories about anthrax | 18:29 |
| fenn | and a 'diy chemist' is not necessarily someone making bombs or drugs | 18:29 |
| fenn | (if you're canadian) | 18:30 |
| fenn | that's not what i mean by culture though | 18:30 |
| Splicer | people should probably be more concerned with a computer somewhere accidentally doing something.. but people are used to those now | 18:30 |
| fenn | i want you to worry less, live more | 18:31 |
| fenn | srsly.. you're more likely to die in a car wreck | 18:31 |
| Splicer | yeah | 18:32 |
| Splicer | it puts a cap on things | 18:33 |
| kanzure | fenn: so, split it up into microjobs, and then optimize each of the microjobs so that you don't unheat the oven and other stuff like that | 18:35 |
| kanzure | it would be something like 'trend analysis' across the microjobs perhaps | 18:35 |
| Splicer | you´re talking about building a unified CAM process and manufacturing format right? Which would take a recepie as input and optimize itself to te tools at hand? | 18:42 |
| fenn | yep | 18:43 |
| fenn | Splicer: congratulations on actually 'getting it' unlike epitron :) | 18:43 |
| fenn | Splicer: here's a pretty picture i drew showing the overall structure: http://heybryan.org/mediawiki/index.php/skdb | 18:44 |
| kanzure | fenn: worst case scenario: it doesn't auto-optimize. fix: somebody adds it in. I think that if it just *flat out works* in the first place, then that's fine. So let's just code in an optimize() method (for this job-optimization, not manufacturing-format optimization (where we do the GA re: local tools and resources) | 18:49 |
| Splicer | because the skdb knows everything it's ideal to automatically optimize the solution. If it is fed a welldefined problem, that's the purpose of the system? | 19:02 |
| fenn | it's also useful for determining what to do next | 19:03 |
| fenn | say you have a goal in mind, if you can define what your present capabilities are it can find the best path ( i guess that's the same as optimizing) | 19:03 |
| Splicer | yeah | 19:04 |
| Splicer | it's a very cool idea... but huge | 19:05 |
| fenn | i expect it to take a long time to get to that point, so we'll be making useful tools in the meanwhile | 19:06 |
| fenn | non-optimized is fine i guess, just frustrating to watch: http://video.google.com/videoplay?docid=5506285167473738661&q=&hl=en | 19:12 |
| fenn | same brazilian guys | 19:12 |
| kanzure | fenn: so you agree it's okay if we ignore that 2nd layer of optimization for now? otherwise I think the architecture is good to go | 19:20 |
| kanzure | Splicer: it's actually quite small | 19:37 |
| kanzure | the infrastructure/architecture is pretty simple | 19:37 |
| kanzure | now that we know what it is ;) | 19:37 |
| Splicer | I don´t know, the more one thinks about it the bigger it gets don´t you find? | 20:24 |
| Splicer | When I look at your pic, I can´t even figure out what a question would look like that would optimize the connectors in it... it sounds like the kind of problem people have been struggling with for a long time, and the answer would border to somekind of general intelligence. | 20:30 |
| Splicer | "Build me a radio" | 20:30 |
| fenn | Splicer: two two hexagons are nodes in the tree, you want to find the optimum path by hopping from hexagon to hexagon | 20:32 |
| Splicer | what? | 20:33 |
| fenn | the top to bottom stuff happens more or less automatically | 20:34 |
| fenn | going from start to finish is left to right though | 20:34 |
| Splicer | you´re saying; everything is already defined and because of the structure it's defined in, the solution is just following a straight path? | 20:36 |
| fenn | the solution is finding the shortest path | 20:39 |
| Splicer | it sounds a bit like the problem of trying to get a computer to understand natural language, it's not as trivial as it sounds. | 20:42 |
| kanzure | haha | 20:44 |
| kanzure | that's why it's not natural language | 20:44 |
| fenn | fortunately most engineering languages are formal notations | 20:45 |
| fenn | there is a lot of 'common sense' to be dealt with though | 20:46 |
| kanzure | it would suck if you had to speak to your mechanical conveyor belt instead of formal specs | 20:46 |
| kanzure | or formal voltages | 20:46 |
| fenn | transformal voltages? :) | 20:46 |
| kanzure | <-- clueless | 20:47 |
| Splicer | ambiguities aside, formal language is also jumping between hexagons and evaluating paths. | 20:48 |
| kanzure | huh? | 20:48 |
| kanzure | 'jumping' makes your sentence not compile in my brain | 20:48 |
| fenn | traversing trees | 20:49 |
| Splicer | yeah | 20:49 |
| kanzure | Splicer: so do you see how it's not like trying to get a computer to parse natural language? | 20:53 |
| kanzure | How weird. A while ago I was clicking through some facebook pages (repetitions ... clicked through 40 pages because their interface sucks; I probably should have gone installed a userscript for facebook to autoload the next page and attach to the end of the screen, but whatever), | 20:56 |
| kanzure | and at one point I got an idea for a next step in one fo the tasks I was working on, but I didn't do it *because I was not sure which xterm session was the one I wanted*. How weird is that. | 20:56 |
| Splicer | I don´t know if I understood everyting.. but isn´t it almost exactly the same? The languagespace is defined in the skdb in some formal notation and getting an answer to a question is just traversing the tree? | 20:57 |
| kanzure | what question? | 20:57 |
| kanzure | yeah, okay, I see what you mean | 20:57 |
| kanzure | the thing is that we can only ask very exact questions | 20:57 |
| Splicer | "using this input, optimize it" | 20:57 |
| kanzure | in the case of a self-replicating machine, we hypothesize that it's simply where all of the packages completely loop together | 20:57 |
| kanzure | ah, well, the optimization stuff is two-fold | 20:58 |
| fenn | the idea is to 'make no assumptions' | 20:58 |
| kanzure | it's all on the userend | 20:58 |
| kanzure | (1) optimize it to the materials and resources available, plus the tools available - this is just a GA and is simple, plus a scoring function | 20:58 |
| kanzure | (2) possibly optimize it to the microjobs or whatever, so that you conserve power and don't waste heat, or don't go in a zigzag when a line would work | 20:58 |
| fenn | sloppy questions are fine, and preferred if they reveal a better solution that a more strict question might have prevented (inadvertently) | 20:58 |
| kanzure | so the optimization is well-defined | 20:58 |
| kanzure | and more optimization algos can be added of course, and users can download 'em to run them | 20:59 |
| kanzure | hm, sloppy questions might be harder to ask though | 20:59 |
| kanzure | you can't just search for the d-loops | 20:59 |
| kanzure | I guess I need an example of a sloppy question | 20:59 |
| kanzure | "find me a toaster" - that's sloppy, right? you would get everything tagged/described-as toaster | 20:59 |
| kanzure | or any aliased with the word toaster | 20:59 |
| fenn | yes, its hard for humans to make just the right amount of uncertainty | 20:59 |
| kanzure | *anything | 20:59 |
| Splicer | those are the oldest games in computer science i think | 21:00 |
| fenn | we want to run our neural nets in reverse and generate a plausible situation, then ask the machine to do it | 21:00 |
| kanzure | yes, they are the oldest games - and we are damn good at them too | 21:00 |
| kanzure | fenn: huh? | 21:01 |
| kanzure | I don't think that's an accurate description really | 21:01 |
| fenn | that's what people want to do | 21:01 |
| Splicer | the best example i know is the google translator | 21:01 |
| kanzure | oh | 21:01 |
| kanzure | fenn: yes | 21:01 |
| kanzure | Splicer: how is that an example of what we are describing? | 21:01 |
| kanzure | Google Translator is natural language processing (actually, it's a lot of lookups and replaces, plus grammar rules and so on) | 21:01 |
| kanzure | actually, Google search fails in the multilingual domains, eh? | 21:02 |
| fenn | google translator sucks, it has no sense of context | 21:02 |
| kanzure | I was suggesting to Jimmy Wales something called MPOV search. | 21:02 |
| kanzure | Jimmy was being bashed to death in the news a few months ago | 21:02 |
| kanzure | he released Wikia Search, and had barely crawled a few hundred thousand sites | 21:02 |
| kanzure | so everybody was very bitter at him | 21:03 |
| kanzure | I jumped on the opportunity and started corresponding with him | 21:03 |
| kanzure | Wikia Search is basically built on lucene, nutch, and a few social networking + wiki frontends for their website (yikes!) | 21:03 |
| kanzure | all open source, so I started thinking about that | 21:03 |
| kanzure | and one of the ideas that I came up with, along with the help of a few others on the search-l mailing list over at wikia.com, was the idea of MPOV search, or the concept that people are searching from a certain context | 21:04 |
| kanzure | he hated it because it was point of view search | 21:04 |
| kanzure | and he wants to be "Objective" | 21:04 |
| kanzure | can anybody spot anything wrong with that when it comes to searching? ;) | 21:04 |
| kanzure | the user is the ultimate evaluator in terms of search result quality -- attention is the primary scarcity, and they have a point of view within which they are searching. So if you could load up a specialized "ranking algorithm" for certain specialized searches, you'd sure as hell use it. | 21:05 |
| kanzure | which fits into the Wikia/ATLAS-L architecture of the internet | 21:05 |
| fenn | there's no such thing as NPOV | 21:05 |
| kanzure | producers, aggregators, brokers of 'information nuggets' | 21:05 |
| kanzure | fenn: correct | 21:05 |
| fenn | language automatically induces a bias, you're imposing your ontology on the undescribable 'objective reality' | 21:06 |
| kanzure | neat, Machinae supremacy just clicked on for demoscene radio. ('Gianna sisters'). Been forever since I've heard these guys. They had good web presence back in the day. | 21:06 |
| kanzure | fenn: seems obvious to us, but Jimmy was probably feeling attacked | 21:06 |
| kanzure | since he upholds his Wikipedia editing policies as the highest morality or something | 21:06 |
| fenn | sure, NPOV is how you keep people from fighting | 21:07 |
| kanzure | heh - how's that working out for them? (edit wars ...) | 21:07 |
| fenn | take a little from here, and a little from here, and everyone gets their say in | 21:07 |
| fenn | well, arbitration does .. something | 21:07 |
| fenn | i dont pay much attention to wikipedia politics (maybe i should) | 21:07 |
| kanzure | don't know if you should | 21:08 |
| kanzure | I certainly don't, except what I get in my inbox or see in the news | 21:08 |
| fenn | they generate a crapload of text to sift through | 21:08 |
| kanzure | or when I try to edit an article | 21:08 |
| kanzure | yes | 21:08 |
| fenn | and all the silly arguments are archived because nobody can come up with a consensus statement | 21:08 |
| kanzure | have I told you about ##wikipeding ? | 21:08 |
| kanzure | the idea is to do community depth-N reads of wikipedia in realtime | 21:08 |
| fenn | http://fennetic.net/milesaway/decision_making | 21:08 |
| fenn | ^^consensus appeals to people but it doesn't work in groups > 3 | 21:09 |
| kanzure | I wonder what would happen if I tried pathologically avoiding decisions | 21:09 |
| kanzure | either I blank everything out and claim it's not a decision | 21:09 |
| kanzure | or I'd retreat and be doing nothing (but isn't that a decision itself?) | 21:10 |
| kanzure | hehe | 21:10 |
| fenn | you'd get paralyzed and accomplish nothing | 21:10 |
| kanzure | perhaps | 21:10 |
| fenn | trying to follow too many paths at once | 21:10 |
| kanzure | which is the core problem of ADHD | 21:10 |
| fenn | yes | 21:10 |
| fenn | limited bandwidth | 21:10 |
| kanzure | wild, disruptive microcolumns that have to be locked down by avoiding stimulation | 21:10 |
| fenn | oh, if you think the bandwidth is "just the way things are" then the problem is you | 21:11 |
| fenn | i'm an unreasonable person though | 21:11 |
| kanzure | 'We discuss how excessive neuronal processing may render the world painfully intense when the neocortex is affected and even aversive when the amygdala is affected, leading to social and environmental withdrawal. Excessive neuronal learning is also hypothesized to rapidly lock down the individual into a small repertoire of secure behavioral routines that are obsessively repeated.' | 21:11 |
| kanzure | what? (re: your oh line) | 21:11 |
| fenn | wild disruptive microcolumns is ok if you can do something with their output (it is useful, in the proper context) | 21:12 |
| fenn | but if there's too much output fighting over the same limited i/o device (eyes ears mouth) then none of them are happy | 21:12 |
| Splicer | like the mad stock guy? | 21:12 |
| kanzure | fenn: output is constrained though, only so many fingers and toes | 21:13 |
| kanzure | only so many wpm typed/spoken | 21:13 |
| fenn | well, fix it :) | 21:13 |
| kanzure | and I tend to do 150 wpm typed when really hyperactive and on a roll, it's so that I can try to record each thought as it occurs, but I still lose so much information ... | 21:13 |
| kanzure | right | 21:13 |
| kanzure | thus my projects | 21:13 |
| kanzure | ah, we were going to do something with emc weren't we? | 21:14 |
| fenn | i was going to show you the difference between optimized and unoptimized paths | 21:15 |
| fenn | if i can figure out how to do that | 21:16 |
| kanzure | paths would be ways of editing the object? | 21:16 |
| fenn | machining the object | 21:16 |
| kanzure | hunk of metal -> tool to drill all over it or whatever -> yeah | 21:16 |
| kanzure | okay | 21:16 |
| kanzure | so some motions would be optimal, others would be stupid for a certain form | 21:16 |
| kanzure | it's certainly an art, but also an art that gets to be automated for obvious reasons (the computer precision) | 21:16 |
| fenn | so, this script will do stuff to download the simulator: http://www.linuxcnc.org/dapper/emc2-install-sim.sh | 21:17 |
| fenn | you can look and see what its doing of course | 21:18 |
| fenn | um, i think it will work. there was an issue with libpth2 vs libpth20 yadda yadda | 21:18 |
| Splicer | time to sleep.. cu | 21:18 |
| -!- Splicer [n=p@h104n2c1o261.bredband.skanova.com] has quit [] | 21:19 | |
| kanzure | fenn: please go on - I'm doing a few things at the moment but am following and see the script, yes | 21:19 |
| kanzure | also, what's with the pgp stuff? | 21:19 |
| kanzure | I recently noticed that APT has been asking me to continue without signatures, but I didn't make much of it | 21:19 |
| fenn | its because it's a different repository that is signed by an independent developer | 21:19 |
| kanzure | (don't know why - wasn't thinking) | 21:19 |
| fenn | so you have to say 'i trust this developer' | 21:20 |
| kanzure | well yeah, but apt has been doing it just in general even when downloading from kernel.org | 21:20 |
| fenn | hmm.. debian etch shipped with some outdated keys i think.. | 21:20 |
| kanzure | weird | 21:20 |
| kanzure | the fact that security contexts are socially diffused illustrates a good reason for skdb, for example | 21:21 |
| kanzure | the whole 'reason why circuit ideas are hidden' | 21:21 |
| kanzure | (it's not that they are hidden, but that they are social knowledge and contextual, etc. etc.) | 21:21 |
| kanzure | erm. I have concluded that I suck at verbalizing. | 21:21 |
| kanzure | hrm, my ~/cache/ is 7 GB. | 21:22 |
| kanzure | 3.6 GB of CFD software. | 21:23 |
| kanzure | http://heybryan.org/wiki/ ikiwiki works | 21:30 |
| kanzure | but it does not look like it allows editing from the web | 21:30 |
| kanzure | is this a problem? | 21:30 |
| kanzure | adding a few scripts to allow editing of the flatfiles is easy, but how would this be interfaced to the versioning system? | 21:31 |
| kanzure | would I be able to just dump the new edit of a page into the repository (dump it into git, actively), if I do this, would git know to check for updates and so on? | 21:32 |
| kanzure | (so it'd be a web frontend to sending updates to the git repository and 'commiting' them) | 21:32 |
| kanzure | doesn't sound too safe ... hrm. | 21:32 |
| fenn | duh.. what? ikiwiki is supposed to commit to its own git repository whenever a web edit is made | 21:33 |
| kanzure | ah, nevermind | 21:34 |
| kanzure | it's extra config options that I need. | 21:34 |
| kanzure | sorry, didn't read far enough | 21:34 |
| kanzure | over at http://ikiwiki.info/setup/ | 21:34 |
| fenn | i dont really get what srcdir is | 21:38 |
| kanzure | the files? | 21:39 |
| kanzure | it's a wiki compiler, not a straight up wiki | 21:39 |
| kanzure | so /var/www/wiki/ is the presentable files, while wikiwc are the plaintext contents | 21:39 |
| kanzure | or the 'working copy' | 21:39 |
| kanzure | dunno | 21:39 |
| kanzure | nevermind | 21:39 |
| fenn | oh i see | 21:40 |
| fenn | woah, interesting collapsible text on the ikiwik setup page (click on Git) | 21:41 |
| kanzure | yep | 21:46 |
| kanzure | it's probably just a div visibility=hidden | 21:46 |
| kanzure | div id=5 onmousedown="javascript:document.getElementByID(5).visibility=show;" style="visibility:hidden;" | 21:47 |
| kanzure | oh, well, the onmousedown is on the link obviously | 21:47 |
| kanzure | and it'd be a toggle instead of a straight up 'only show it' method | 21:47 |
| kanzure | http://heybryan.org/wiki/ adding git at the moment | 21:48 |
| kanzure | gah, ikiwiki-makerepo does not exist | 21:50 |
| fenn | /usr/bin/ikiwiki-makerepo ? | 21:50 |
| kanzure | nope | 21:51 |
| kanzure | only ikiwiki and ikiwiki-mass-rebuild | 21:51 |
| kanzure | http://ikiwiki.info/plugins/ awesome, they have an hnb plugin (hnb is what I used to make http://heybryan.org/todo.html back in the day) | 21:51 |
| kanzure | so in 2007-12 ikiwiki-makerepo was announced on the debian developer mailing list - http://lists.debian.org/debian-devel-changes/2007/12/msg00054.html | 21:53 |
| kanzure | so. | 21:53 |
| fenn | i have ikiwiki 2.41 | 21:54 |
| fenn | how did you get the edit buttons? | 21:55 |
| kanzure | 1.33 wtf | 21:55 |
| kanzure | fenn: have to enable cgi | 21:55 |
| kanzure | 1.33.3 in particular | 21:55 |
| kanzure | this sucks. | 21:55 |
| kanzure | are my sources old? | 21:55 |
| fenn | i'm using etch/testing | 21:56 |
| kanzure | apt-get update; now I got 1.33.4 ... | 21:56 |
| fenn | oh wait, theres an unstable so i guess not | 21:56 |
| kanzure | I thought I installed testing back in February. | 21:56 |
| kanzure | I'm using etch. | 21:58 |
| kanzure | added to my sources.list: deb http://mirrors.kernel.org/debian/ etch main contrib non-free | 22:00 |
| kanzure | deb-src http://mirrors.kernel.org/debian/ etch main contrib non-free | 22:00 |
| kanzure | apt-get update | 22:01 |
| kanzure | 1.33.4 is still the latest. | 22:01 |
| kanzure | did you install by hand? | 22:01 |
| kanzure | ah, unstable | 22:03 |
| kanzure | do I dare go unstable. | 22:03 |
| kanzure | something must be wrong- "make: command not found" | 22:09 |
| fenn | that's definitely not right | 22:09 |
| fenn | oh joy how i missed perl CGI errors | 22:17 |
| kanzure | http://www.google.com/codesearch?client=opera&rls=en&q=msgfmt&sourceid=opera&ie=UTF-8&oe=utf-8&um=1&sa=X&oi=codesearch_group&resnum=4&ct=title | 22:20 |
| kanzure | got my first 'Google Code' result in my web search results | 22:20 |
| kanzure | hm | 22:20 |
| fenn | ffs | 22:21 |
| fenn | i think ikiwiki doesnt have all the dependencies in its debian package | 22:21 |
| kanzure | msgfmt is in package 'gettext' | 22:22 |
| kanzure | http://en.wikipedia.org/wiki/SPARQL thoughts? | 22:28 |
| kanzure | it looks like a way to work with structured documents (RDFs) | 22:28 |
| kanzure | I think the yaml equivalent is http://yaml.kwiki.org/index.cgi?YPathBrainstorm 'ypath' | 22:30 |
| kanzure | so is SPARQL to RDF as xpath is to XML? | 22:31 |
| fenn | cool it works | 22:35 |
| fenn | um, i dont know how exactly :( | 22:35 |
| kanzure | git: 'init' is not a git command | 22:45 |
| kanzure | hrm | 22:45 |
| fenn | here's what i did: http://fennetic.net/pub/irc/ikiwiki | 22:45 |
| fenn | git init works here | 22:45 |
| fenn | git-core 1:1.5.3.7-1 | 22:46 |
| kanzure | 1:1.4.4.4-2 -- it keeps on talking about how I should port over to gitfm, but then gitfm isn't used by ikiwiki, now is it? | 22:47 |
| fenn | the package 'git' is the wrong git, that's some other bullshit that happens to have the same name | 22:47 |
| fenn | apt-get remove git; apt-get install git-core | 22:48 |
| kanzure | usr/bin/git: no such file or directory | 22:49 |
| kanzure | so. | 22:49 |
| fenn | zowee, there are 138 git-* commands | 22:49 |
| kanzure | git-core - content addressable filesystem | 22:49 |
| kanzure | git-cvs - content addressable fs (cvs interoperability) | 22:50 |
| kanzure | but I have git-core installed, but still no git command | 22:50 |
| fenn | dpkg -S /usr/bin/git | 22:50 |
| fenn | git-core: /usr/bin/git | 22:50 |
| kanzure | not found | 22:50 |
| fenn | wtf | 22:50 |
| kanzure | git-core version 1:1.4.4.4-2 | 22:50 |
| kanzure | replaces: cogito | 22:50 |
| kanzure | so, what's going on? | 22:52 |
| fenn | this is what dpkg -L git-core shows: http://pastebin.ca/990350 | 22:54 |
| fenn | are you getting something similar? | 22:55 |
| fenn | i see this in the changelog (git-core 1.5.2.1-1) debian/control: git-core: Conflicts: git (<< 4.3.20-11) (/usr/bin/git transition). | 22:56 |
| fenn | so looks like the package 'git' had /usr/bin/git and they switched it around so git-core was the 'real' git | 22:57 |
| fenn | what a fubar | 22:57 |
| kanzure | I don't have git-submodule, among others like 'git' | 22:57 |
| kanzure | hm? | 22:57 |
| kanzure | so then what am I to do? | 22:57 |
| fenn | /usr/bin/git-scm | 22:59 |
| fenn | i think that provides the same funcitonality as /usr/bin/git | 22:59 |
| kanzure | I'll ln it then. | 23:00 |
| fenn | or /usr/bin/git.transition | 23:00 |
| kanzure | no such file | 23:01 |
| kanzure | git-scm | 23:03 |
| kanzure | fatal: cannot handle scm internally | 23:03 |
| fenn | ./usr/bin/git-scm link to ./usr/bin/git-format-patch | 23:06 |
| fenn | looks like its just a link | 23:06 |
| kanzure | so should we go for svn? | 23:06 |
| fenn | ugh no | 23:06 |
| kanzure | cvs? | 23:06 |
| fenn | nooooo!! | 23:06 |
| fenn | this is a debian etch problem | 23:06 |
| kanzure | fenn, your precious git package does not exist | 23:06 |
| kanzure | eh | 23:06 |
| fenn | GIT Transition Plans. Ian Beckwith explained that /usr/bin/git will be maintained via alternatives. Both GNU Interactive Tools and Linus Torvalds' source code management system git provide the same program name. The GNU interactive tools have been renamed into gitfm to resolve this conflict and in etch will contain a wrapper that is able to execute the other program. | 23:08 |
| kanzure | hm | 23:08 |
| fenn | from http://lists.debian.org/debian-release/2006/08/msg00351.html | 23:09 |
| kanzure | how does that resolve the conflict | 23:09 |
| kanzure | update-alternatives --config git | 23:09 |
| kanzure | There is only 1 program which provides git \n (/usr/bin/git-scm). Nothing to configure. | 23:10 |
| fenn | you have no file /usr/bin/git-scm? | 23:10 |
| kanzure | fatal: cannot handle scm internally | 23:10 |
| fenn | the fatal: is at least a git error :) | 23:13 |
| fenn | i think git 1.5 is a lot different from 1.4 | 23:15 |
| kanzure | there is a way to specify which version to get via apt | 23:16 |
| kanzure | apt-get install git-core <something here> | 23:16 |
| fenn | git-core=1.5.123123123 | 23:16 |
| fenn | but its not in the etch repo's | 23:16 |
| fenn | because etch is old | 23:16 |
| kanzure | so, should I try apt-get dist-upgrade? | 23:16 |
| fenn | if you dont mind things being broken for a while :) | 23:16 |
| kanzure | eh, but I do | 23:17 |
| kanzure | can I install it by hand? | 23:17 |
| fenn | yes | 23:17 |
| kanzure | you recommend 1.4? | 23:18 |
| fenn | no, 1.5 | 23:19 |
| fenn | http://packages.debian.org/testing/devel/git-core | 23:19 |
| fenn | or from source: http://kernel.org/pub/software/scm/git/git-1.5.5.tar.bz2 | 23:20 |
| kanzure | ./configure was successful | 23:23 |
| kanzure | make | 23:23 |
| kanzure | in file included from daemon.c:1, cache.h:9:18: error: zlib.h: No such file or directory | 23:23 |
| kanzure | hrm | 23:23 |
| fenn | sudo apt-get build-dep git-core | 23:26 |
| fenn | should get a lot of -dev packages | 23:26 |
| fenn | hrm no source package.. lame | 23:27 |
| kanzure | I got the dependencies from http://ftp.de.debian.org/debian/pool/main/g/git-core/git-core_1.5.5-1.dsc | 23:27 |
| fenn | ok | 23:28 |
| kanzure | "Yes, and how do you envision that working? I think a great deal of time and effort would need to be given to societal structure on Mars. Dare I say that it would be much more useful for a Virgle colony to leave social planning to social scholars... I would much rather live in a Paul Fernhout society than a Bryan Bishop society. Stick to engineering or physics, Bryan, haha." - some random guy from the Virgle mailing list re: 'some | 23:32 |
| kanzure | sigh | 23:32 |
| fenn | fernhout is a social scholar now? | 23:34 |
| fenn | cant we all just get a long | 23:35 |
| kanzure | the virgle group is a big source of hostility | 23:35 |
| kanzure | it's really weird | 23:35 |
| kanzure | I think it's attracted the wrong people or something | 23:35 |
| kanzure | very angry people | 23:35 |
| kanzure | make, make install | 23:36 |
| kanzure | git - still no such file | 23:36 |
| kanzure | this was with 1.5.5 | 23:36 |
| kanzure | (git-core) | 23:36 |
| fenn | /usr/local/bin/git? | 23:38 |
| kanzure | exists | 23:39 |
| fenn | \o/ | 23:39 |
| fenn | so.. does it work? | 23:40 |
| fenn | mkdir testgit; cd testgit; /usr/local/bin/git init | 23:41 |
| kanzure | yes | 23:41 |
| kanzure | but uhh, how do I create the link? | 23:41 |
| kanzure | ln /usr/bin/git /usr/local/bin/git does not work | 23:41 |
| fenn | other way around | 23:41 |
| fenn | ln -s /usr/local/bin/git /usr/bin/git | 23:41 |
| fenn | but it should find /usr/local/bin/git if the command is being called in a halfway decent manner | 23:42 |
| kanzure | 'ln creating symbolic link blah to blah2 : file exists' | 23:42 |
| kanzure | git still does 'no such file or directory' though (for /usr/bin/git) | 23:42 |
| fenn | hmm i thought you just said /usr/bin/git didnt exist | 23:42 |
| kanzure | hm | 23:42 |
| kanzure | right | 23:42 |
| kanzure | bash says that too | 23:42 |
| fenn | ls /usr/bin/git | 23:42 |
| kanzure | responds with /usr/bin/git | 23:42 |
| fenn | uh.. ls -l /usr/bin/git | 23:43 |
| kanzure | /usr/bin/git -> /etc/alternatives/git/ | 23:43 |
| kanzure | erm | 23:43 |
| kanzure | minus the lash / | 23:43 |
| kanzure | *last / | 23:43 |
| fenn | aroo | 23:43 |
| fenn | ls -l /etc/alternatives/git | 23:43 |
| kanzure | it's a link to /usr/bin/git.transition | 23:43 |
| kanzure | which does not exist | 23:43 |
| fenn | i see | 23:43 |
| kanzure | so, remove /usr/bin/git ? | 23:44 |
| fenn | first, uninstall git-core | 23:44 |
| kanzure | yes, git.transition doesn't exist for a good reason | 23:44 |
| kanzure | I did uninstall git-core earlier when there was that conflict | 23:44 |
| fenn | ok | 23:44 |
| kanzure | so it just left the link for some reason | 23:44 |
| fenn | hack away | 23:44 |
| kanzure | okay, let's see now | 23:44 |
| fenn | or you could do update-alternatives git | 23:44 |
| kanzure | http://heybryan.org/wiki/recentchanges/ | 23:47 |
| kanzure | wtf? | 23:47 |
| fenn | yeah its pretty ugly | 23:48 |
| kanzure | but why isn't the [[ wikicamel ]] interpreted? | 23:48 |
| fenn | dont know | 23:48 |
| kanzure | and now the edit page isn't working | 23:48 |
| fenn | eyes burning.. | 23:58 |
Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!
