--- Day changed Mon Apr 06 2009
00:05 -!- cis-action_ [n=cis-acti@146-115-127-170.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com] has joined #hplusroadmap
00:09 -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has joined #hplusroadmap
00:22 -!- cis-action [n=cis-acti@146-115-127-170.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com] has quit [Read error: 110 (Connection timed out)]
00:39 < kanzure> http://code.google.com/p/physim/ "A program that takes an XML document describing a free body diagram and simulates the system over a period of time."
00:59  * kanzure still needs a list of examples of different types of constraints that need to be evaluatable
01:05 < kanzure> the only constraints stuff I ever find when looking at CAD is about geometry, but I know there should (hypothetically?) be more than that.
01:06 < kanzure> in fact, I haven't even seen a system that uses SI units as constraints between two parts in an assembly
01:23 < kanzure> erm, I mean, quantities
01:29 < kanzure> gee, I was talking about this in 2006? http://www.allegro.cc/forums/thread/587715
01:43 < kanzure> implementing physical-quantities (the library) should be good enough for now. then basic chemistry (i.e., for particular materials, concentrations, etc.) these will be 'cookie crumbs' left as a list attached to labeled ports.
01:43 < kanzure> search-shape-similarity-reeb stuff can come later (much later) if necessary
01:43 < kanzure> (and in the mean time, interference checking is probably a better idea anyway)
01:44 < kanzure> a first script that specifies a small object having a list of input and output ports would be nice, and then tie it in to the topsort and pq libraries.
01:45 -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has quit [Remote closed the connection]
01:54 -!- gene [n=chatzill@wireless-128-62-92-192.public.utexas.edu] has quit [Read error: 110 (Connection timed out)]
01:54 -!- cis-action_ [n=cis-acti@146-115-127-170.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com] has quit []
07:13 -!- splicer_ [n=patrik@h187n1c1o261.bredband.skanova.com] has quit ["Ex-Chat"]
07:45 < fenn> topsort? http://pypi.python.org/pypi/topsort/0.9
08:40 -!- gene [n=chatzill@wireless-128-62-96-128.public.utexas.edu] has joined #hplusroadmap
09:48 -!- cis-action [n=cis-acti@c-67-186-135-65.hsd1.ma.comcast.net] has joined #hplusroadmap
09:49 -!- cis-action [n=cis-acti@c-67-186-135-65.hsd1.ma.comcast.net] has quit [Client Quit]
09:57 < kanzure> yes.
10:09 -!- anthonyl [n=duzt@dsl093-216-054.aus1.dsl.speakeasy.net] has joined #hplusroadmap
10:34 < kanzure> http://heybryan.org/books/papers/?C=M;O=D more stuff.
10:42 < anthonyl> stuff :)
10:57 -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has joined #hplusroadmap
10:57 < kanzure> lots of stuff.
11:00 < anthonyl> haha, ya.. your poor server is about to get a burst of traffic i think
11:01 < kanzure> why's that?
11:07 < anthonyl> dunno, from me downloading pdf's but really i'm kinda overexagerating a tad
11:17 -!- nsh- [n=nsh@host217-43-230-27.range217-43.btcentralplus.com] has joined #hplusroadmap
11:21 < willPow3r> really. downloading pdfs from kanzure's connection actually inadvertently DoS's his server
11:31 < anthonyl> wget -m *knowledge*
11:32 < fenn> there's some 30GB there
11:32 < fenn> in /books/
11:32 < fenn> or maybe more
11:33 < UtopiahGHML> "wget -m *knowledge*" makes me think that Talis now offer to store your triple store http://blogs.talis.com/n2/cc for free (50M typles/10Gb restrictions apply, etc)
11:33 < kanzure> btw, for /books/papers/ you can now grab from http://sata.serveftp.org/~bryan/papers/
11:33 < fenn> that's very nice of them
11:34 < kanzure> nice of who? hehe
11:34 < kanzure> I'm the sysadmin
11:34 < fenn> talis
11:34 < kanzure> ?
11:34 < kanzure> oh, sorry, didn't see that message.
11:34 < fenn> free hosting of data sets
11:39 < kanzure> free hosting of your mom
11:39 < kanzure> sorry, my teenager is showing
11:39 < bkero> teenager
11:40 < kanzure> fenn, I sent a request to OM for people to bombard me with types of constraints that a CAD system should have
11:40 < kanzure> you previously mentioned 'protocol', but I've been ignoring it (not because I don't like it, but because I just haven't got around to it)
11:41 < kanzure> it's important, but I think the data structure I proposed in the email to OM can handle it
11:41 < kanzure> it would be something like a simple string comparison evaluator (i.e., the same way that a MIME type is checked :p)
11:42 < fenn> it should support 'backwards compatible' interface
11:42 < fenn> this is easily done with duck typing and class inheritance
11:42 < kanzure> backwards compatible interface?
11:43 < kanzure> "USB 3.0 ports work with USB 2.0" if that's what you mean
11:43 < kanzure> (actually I don't think they do; don't recall the specs.)
11:44 < bkero> USB 3.0 is longer
11:44 < bkero> The extra pins are "behind" the usb2 ports in the slot
11:44 < bkero> http://www.coolest-gadgets.com/wp-content/uploads/usb301.jpg
11:46 < fenn> heh actually i was thinking usb2.0 and 1.1 but i wasnt sure they were actually backwards compatible in the spec or just de-facto
11:46  * fenn shakes head.. why didnt they go with power over ethernet
11:48 < fenn> kanzure: the point of "protocol" is to be able to define any interface spec in terms of physical quantities
11:48 < fenn> and code
11:49 < fenn> so that way you're guaranteed to be able to boil down all the different possible ways of representing something in to the same system
11:49  * fenn grumbles about lack of cheap myvu's on ebay when he actually wants to buy one
12:08 -!- cis-action [n=cis-acti@209.113.205.244] has joined #hplusroadmap
12:08 < fenn> is there any documentation on wearable computers that was NOT written by steve mann?
12:10 -!- xp_prg [n=xp_prg3@99.2.31.217] has joined #hplusroadmap
12:11 < kanzure> fenn: be sure to document teh links you're finding.
12:11 < kanzure> fenn: what? re: the point of 'protocol'. It's not for things like "TCP/IP" vs. "UDP"? I'm not particularly interested in evaluating/analyzing UDP down to the individual electrons being popped off the stack of a current :p
12:11 < kanzure> erm, of a voltage differential
12:11 < kanzure> erm. something like that.
12:12 < fenn> it matters when you're building hardware
12:12 < fenn> not TCP but the actual ethernet signal
12:13 < fenn> (which seems to be rather difficult to track down on teh nets)
12:13 < fenn> fr example USB runs at 10MHz with 5V differential signaling
12:13 < fenn> these are physical quantities
12:14 < fenn> oops 3.3V i mean
12:14 < fenn> but there is code that goes along with these quantities as well, which turns a simple signal into a protocol
12:15 < fenn> analog video signal is a good example to think about
12:16 < fenn> so, wearcam.org is basically just steve mann's homepage
12:16 < fenn> i'm trying to find different implementations of an eyetap (but so far i can't even seem to find just one)
12:20 < kanzure> the pin specs for usb ypu mean/.
12:20 < kanzure> ?
12:21 < fenn> not pins, the electrical activity happening on the bus in response to various situations
12:21 < fenn> for example, in order to get more than 50mA current you have to request it
12:22 < kanzure> okay?
12:22 < fenn> if you don't' request it, the usb port shuts down
12:22 < kanzure> but you just tag it as following the USB standard, no?
12:22 < kanzure> tag tag tag
12:22 < fenn> like i said, i want to be able to represent standards in terms of basic physical quantities
12:23 < fenn> with cheap usb gadgets there is a lock of standards-loophole-hacking going on
12:23 < fenn> s/lock/lot/
12:23 < fenn> they don't necessarily conform to the standard as expected, but it works nevertheless
12:24 < fenn> we can't simply ignore everything that isn't perfectly standards compliant
12:24 < fenn> anyway tagging is a good easy start
12:25 < kanzure> how would you imagine representing specs in terms of basic physical quantities?
12:27 < fenn> some kind of grammar specification?
12:27 < fenn> like ebnf
12:27 < fenn> ideally something that would be computer-executable as is
12:28 < kanzure> ebnf threw me off, apparently it's "e-BNF". what's the e for?
12:28 < fenn> extended i think
12:28 < fenn> the general idea is what's important
12:29 < fenn> you define a set of production rules
12:29 < fenn> i think they are just substitution rules
12:30 < kanzure> hm.
12:31 < kanzure> so what would have the grammar specification? the "usb tag points to the usb grammar spec thingy"?
12:31 < kanzure> and then how would this grammar be used ?
12:31 < kanzure> I just don't know what is being subjected to the substitution rules. the physical quantities?
12:31 < fenn> in simulations
12:32 < fenn> by unit tests
12:32 < fenn> failure mode analysis and so on
12:33 < fenn> honestly i'm not that familiar with computer grammar formalism
12:34 < kanzure> basically it sounds like "partA has xyz port with this stuff, and it is controlled by this program here; partB has a compatible geometry, but with programB"
12:34 < kanzure> and then we want to know whether or not programA and programB will kill each other
12:34 < kanzure> or if they will operate together well?
12:34 < kanzure> is that the deal?
12:35 < kanzure> so I think that it would be reasonable to expect that the port information could be translated into free body diagram information
12:36 < kanzure> and that this information could in general be thrown into an FEM simulator thingy
12:36 < kanzure> but I've never seen an FEM work out two slightly incompatible modes of USB operation, for instance
12:37 < fenn> yes that's' the idea
12:37 < fenn> there are so-called "multi-modal" simulators already
12:38 < fenn> or is it multi-domain
12:38 < kanzure> I've seen multi-domain.
12:38 < kanzure> frankly it seems like the information for those FEM simulations would have to be specified elsewhere
12:39 < fenn> what information?
12:39 < kanzure> in other words I don't see it as much of a block to skdb development
12:39 < kanzure> the information on the scenario: like the two parts interfacing for instance
12:39 < fenn> yeah
12:39 < kanzure> and then the information for simulating USB entirely
12:39 < kanzure> those are some very detailed simulations, and I don't know if skdb should have to provide all of the nasty details
12:39 < kanzure> it should provide boundary conditions, or something
12:39 < kanzure> but that's what's already being tagged with port information and such
12:39 < kanzure> soo..
12:40 < fenn> round and round she goes
12:40 < kanzure> ?
12:40 < fenn> can you imagine if we had to have a committee meeting to decide anything?
12:41 < kanzure> you'll have to submit paperwork to initiate a new imaginary scenario, fenn
12:41 < kanzure> sorry, the office of paperwork submissions is closed for lunch at the moment
12:42 < kanzure> so now. 
12:43 -!- cis-action [n=cis-acti@209.113.205.244] has quit [Read error: 60 (Operation timed out)]
12:44 < kanzure> let's say I write a class for ports, and it has a list of constraint objects which point to their evaluator and a string specifying the constraint expression itself
12:44 < kanzure> the pointer to the evaluator would just be the evaulator class name I guess?
12:45 < kanzure> when checking the compatibility of two given ports (portA, portB), for any constraints between the two with the same type of evaluator, the evaluator is executed
12:45 < kanzure> and then an ok or no-go is returned
12:46 < kanzure> the string expression has to somehow reference the other 'hypothetical object that this port might one day have to mate with', I think, and then set up some expression relating one of its properties to the property of the other
12:46 < kanzure> or something
12:49 < fenn> i wasn't aware of steve's latest 2mm frame design (see page 5) http://wearcam.org/carpe/carpe.pdf
12:50 < fenn> i wonder about the resolution/cost ratio of fiber optic bundles
12:50 -!- xp_prg [n=xp_prg3@99.2.31.217] has quit ["Leaving"]
12:51 < kanzure> (so I imagine multiple evaluators will be written for different domains)
12:51 < kanzure> but the one type that might be the most useful at first would be one that evaluates 'permissibles'
12:51 < kanzure> which would consist of the range of a possible physical quantity going through a port
12:52 < kanzure> for instance, a permissible torque of a port might specify the range of torque that it allows or can provide (depending on the direction of the torque)
12:52 < kanzure> a permissible force, the same thing.
12:52 < fenn> i think you'd want a triple for the constraint expression, not a string
12:53 < fenn> A parallel-to B
12:54 < fenn> pin-Z > 5V
12:54 < fenn> middle part is the evaluation operator
12:55 -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has quit [Remote closed the connection]
12:55 < fenn> the noun objects classes (pin-Z, face-A) are what define the operators
12:56 < kanzure> hm, I'm reminded of apache configuration expressions and .htaccess for allow/disallow
12:56 < fenn> and please stop calling things ports unless they are actually a means of input/output to/from a system
12:56 < kanzure> pipes?
12:57 < fenn> interface!
12:57 < kanzure> oh right.
13:00 < kanzure> where are the operators defined?
13:00 < kanzure> and the nouns get to be physical quantities?
13:00 < fenn> i was going to explain that but i thought you might get OOP-fever and run off
13:00 < kanzure> go ahead.. I might just back away slowly
13:01 < fenn> in the pin-Z example, it's an instance of class "pin" which inherits from "conductor" and from that you get all the electrical evaluators
13:01 < fenn> so they are just class methods
13:01 < fenn> you know about overloading right?
13:02 < kanzure> no
13:02 < fenn> so i could overload the || operator to Geom.parallel_to()
13:02 < kanzure> yes
13:02 < kanzure> sorry, you mean the programming term
13:02 < kanzure> yeah.
13:03 < fenn> then when i type A || B if A and B are of te proper types (or have the right attributes) it gets evaluated
13:03 < fenn> i know nothing about building a constraint solver
13:04 < kanzure> I do not think making 'pin-Z' inherit from 'conductor' is a good idea in general. I was thinking that in the case where you have (partHere operator physical-quantity) the operator will know to check for constraints on the part that are of the same type as the physical quantity
13:04 < fenn> obviously i want to avoid random thrashing; so there has to be some way to pass the constraints along, but i dont know what it's going to be doing with that info
13:05 < fenn> yeah of course; i wouldnt want to type all that out either
13:05 < fenn> but it's saying the same thing essentially
13:05 < fenn> pin-Z connected to 5V means pin-Z conected to (hypothetical conductor at voltage of 5V)
13:06 < kanzure> what do you mean by passing constraints along?
13:06 < kanzure> pass from what to what, in what example scenario?
13:06 < fenn> well i imagine the constraint solver wants to modify a bunch of variables
13:06 < kanzure> ?
13:06 < kanzure> example?
13:07 < fenn> and if it had access to what the constraints were it would find some smart way to modify a particular variable first so as to minimize the search time
13:07 < kanzure> the way I've been treating constraints has been as static things
13:07 < kanzure> not as a multi-variable optimization problem
13:07 < fenn> say your constraint is A || B
13:08 < fenn> one way is to randomly rotate B around until the evaluator returns true
13:08 < kanzure> and if it's not satisfied, you complain.
13:08 < kanzure> you wouldn't rotate it, you would just set it or check it
13:08 < fenn> another way is to return the angular difference and use iterative root-finding like newton-raphson or eulers method
13:09 < kanzure> setting it to be true (by changing B's coords etc.) might create an interference, which is just something that needs to be checked
13:09 < fenn> yes and for trivial constraints it's possible to just set it to the same angle
13:09 < fenn> but these SMOP things like "just set it to the same angle" always turn into huge messes
13:10 < kanzure> suppose you have two interfaces with surfaces (suppose they are interfacing pegs and a magical superglue medium that is unspecified/irrelevant)
13:10 < kanzure> eh. well. it's hard to explain but I don't think parallelism is a problem here
13:10 < kanzure> because this is about part mating, no? so it's kind of made to set it to be parallel
13:10 < fenn> ok so something like two 1x2 lego bricks right?
13:10 < kanzure> sure
13:10 < kanzure> I guess it's possible that there is another parallel constraint somewhere
13:11 < kanzure> ah, suppose you have three lego bricks
13:11 < kanzure> two of the lego plancks have to be parallel, but connected by a lego brick in between (bridged)
13:11 < fenn> isnt two complicated enough already? :s
13:11 < kanzure> well I was thinking that the only time you would need parallelism is in mating two interfaces
13:11 < kanzure> but that's not entirely true
13:12 < kanzure> in the case of those three bricks, you start with part1, part2, part3, you mate part1 and part2, and then (part1+part2) with part3
13:12 < kanzure> so in the end it really is just a combination of parallel mates
13:12 < kanzure> so maybe I was right
13:12 < fenn> it's not just parallel
13:13 < fenn> the lego studs have to fit into the lego holes
13:13 < fenn> this is geometry + elastic deformation + some amount of force
13:13 < fenn> and the geometry has some tolerance info too
13:14 < fenn> because we're talking about a class of objects, not a specific lego brick that we've measured down to the nanometer
13:14 -!- cis-action [n=cis-acti@31-33-205.wireless.csail.mit.edu] has joined #hplusroadmap
13:15 < fenn> anyway that's part mating, which is just a subset of constraints
13:15  * fenn curses the US education system for sucking so poorly
13:16 < fenn> bastards never taught multivariate calculus
13:17 < fenn> the solver will have to handle discrete and continuous variables
13:17 < fenn> so for example a chemical or a bolt size is a discrete variable
13:22 < kanzure> right
13:22 < kanzure> I still think that a multivariable constraints problem should be offloaded into another system
13:23 < kanzure> for checking compatibility it should be a different issue.
13:23 < kanzure> for instance, whether or not two interfaces should interface with each other based off of basic domain evaluations of what permissible forces
13:23 < kanzure> has anyone ever played tetris up to a ridiculously high level and now everything is accelerated even after playing it? I'm sort of in that state right now
13:23 < kanzure> and somehow symbols in words are now tetris objects being combined into weird shapes. hrm. 
13:25 < fenn> dude, the universe is a tetris game
13:25 < fenn> duuuude
13:28 < kanzure> totally
13:32 < kanzure> in multivariable calculus you sometimes define permissible domains or ranges with other functions
13:32 < kanzure> so that's kind of what I'm thinking here, especially re: the triples
13:32 < kanzure> when it comes to geometric constraints, let's let the CAD packages that already do that, do that
13:32 < kanzure> especially for large geometrical problems with lots of constraints
13:32 < kanzure> but for non-geometrical constraints, I think we should provide that
13:32 < kanzure> no?
13:38 < fenn> right
13:39 < fenn> i do want to retain some low-level control over what gets evaluated though
13:43 < kanzure> what do you mean?
13:43 < kanzure> I was thinking it would be nice to be able to specify ignore such and such parameters, or threshold ranges so that you can fuzzy-match stuff, etc., and other things like that
14:54 < kanzure> thanks fenn for http://www.pythonocc.org/wiki/index.php/Install_Linux#Using_SCons
14:56 < fenn> i think he might have fixed that already
15:00 < kanzure> for non-geometric constraints, are we good to go with tuples?
15:03 < fenn> for geometric constraints too
15:04 < kanzure> no, because that involves multivariable constraints
15:04 < fenn> so?
15:04 < kanzure> I mean, a multivariable solver
15:04 < fenn> so does everything else
15:05 < fenn> you can't solve any constraint completely separately
15:05 < fenn> it would be nice, but the universe hates you
15:05 < fenn> design is all about tradeoffs
15:07 < kanzure> so I guess there would have to be a way to "pin" a surface if it solves a constraint, and then go on to the next constraint, and if wants to pin another surface in such a way that it conflicts with the first operation, then it fails? but then that doesn't mean that no solution exists, so.
15:07 < kanzure> blah.
15:07 < kanzure> please don't make me write a multivariable constraints solution engine
15:08 < kanzure> I guess I was proposing graphsynth as a backend for that to the brlcad people
15:08 < kanzure> but it just finds "some" solutions, not necessarily optimal solutions
15:08 < kanzure> (although it can find all solutions)
15:08 < fenn> i'm just looking for "good enough" solutions
15:11 < kanzure> something in the literature about constraint graphs
15:16 < kanzure> "A constructive approach to solving 3-D geometric constraint systems using dependence analysis"
15:45 < kanzure> fenn: hm, constraint solvers don't seem to be something solved.
15:49 < fenn> heh
15:49 < fenn> meta constraint solver
15:51 < kanzure> throw in the phrase "hierarchical bayesian analysis" and you just might be handed a phd
15:57 < fenn> i would have to do an irrelevant review of the literature first
16:00 < kanzure> don't bother, I've already done that :/
16:01 < kanzure> does MINION do doubles/floats?
16:04 < kanzure> since when are constraints present in assemblies? 
16:04 < kanzure> constraints are mostly for a geometrical definition of an object..
16:04 < kanzure> of a single object
16:04  * kanzure is trying to avoid having to write a constraint solver
16:20 < fenn> eh?
16:20 < fenn> constraints are what make the parts go together
16:20 < fenn> face A must be in contact with face B etc
16:40 -!- ybit [n=heath@unaffiliated/ybit] has quit [Read error: 104 (Connection reset by peer)]
16:42 -!- ybit [n=heath@unaffiliated/ybit] has joined #hplusroadmap
16:57 < fenn> should constraints have an order (sequence) or weights?
16:58 < fenn> or if/or/else structure perhaps
17:08 < kanzure> "face A must be in contact with face B" okay?
17:08 < kanzure> sequences are fine
17:08 < kanzure> if you add a collection of parts together and they interfere, then you have an error
17:08 < kanzure> now, that might just be the first options for placement of each part, so try options 1, 1, 2, and then try options 1, 1, 3
17:08 < kanzure> etc. but it does not require a full-blown constraints engine
17:09 < fenn> i'm thinking weights are the way to go
17:09 < fenn> preferred, important, critical
17:09 < fenn> safety-critical
17:09 < kanzure> sure, different options having different weights, or something
17:10 < kanzure> (you mean weighted-scoring, right?)
17:10 < fenn> we could make the order they're listed affect the order in which things are evaluated perhaps?
17:11 < kanzure> wait, where are these things being listed?
17:11 < fenn> in yaml you have to explicitly put something in a sequence for the order to matter
17:11 < fenn> in package metadata i guess
17:11 < fenn> it's not really metadata (since it's data)
17:11 < kanzure> but if you're making a package there shouldn't be "options" really, you should just say "here is what will work" (and if you're wrong, your packaging skills suck)
17:12 < kanzure> erm
17:12 < fenn> i'm not so sure
17:12 < kanzure> what are we talking about?
17:12 < fenn> i'd like to leave some flexibility
17:12 < kanzure> say that you want to mate partA with partB
17:13 < kanzure> there may be several different options for that mating to occur
17:13 < kanzure> partA might be a peg, and partB has two holes.
17:13 < fenn> like, if your tolerances cause the part mating to fail in a statistically small portion
17:13 < fenn> or, if you dont have a 1k resistor or something
17:13 < kanzure> if you make an skdb package that consists of partA+partB, then you should say "this is the option that I want to use"
17:14 < kanzure> option #1: @ the first hole, option #2: @ the second hole
17:14 < fenn> if it absolutely must go in that hole then mark it as critical
17:15 < kanzure> okay, so here's a scenario where it might be critical
17:15 < kanzure> suppose you do that peg+hole assembly, and now you have a third part
17:15 < kanzure> and this third part has a thingy hanging off that intersects the hole-plate if it's on one side of the hole-plate as opposed to being ok on the other side of the hole-plate
17:15 < kanzure> so that would be a case where options are useful.
17:15 < fenn> i'd argue that there are a lot of options like that
17:15 < kanzure> (we'd detect an interference/collision in the case where the thingy is hanging off and intersecting the hole-plate)
17:16 < fenn> usually arbitrarily selected just to get shit done, but if we left it vague it would allow many more choices to be made down the road
17:16 < fenn> also you can optimize an entire system if you leave it vague
17:16 < kanzure> I can see that as going either way in terms of being good/bad
17:17 < kanzure> lots of options => remember yer computational complexity analyses
17:17 < fenn> then when you solve for all the example cases, there's your standard which you can put a gun to everyone's head and make them follow
17:17 < fenn> sure, within computational feasibility
17:17 < fenn> that's what order of operations is for, i guess
17:18 < kanzure> in terms of collision detection there's no way this is going to tax a GPU unit or something
17:18 < kanzure> games do tons of wacky shit with collision detection and overtaxing the system
17:18 < kanzure> hm.
17:18 < fenn> if option A is failing on an "important" constraint, skip it and move to the next one (but don't discard it entirely)
17:18 < kanzure> right
17:19 < fenn> GPU collision detection is low-rez
17:19 < fenn> i wouldn't trust it
17:19 < kanzure> oh? ok ok, cpu-computed collision detection.
17:19 < fenn> now, GP-GPU stuff might be applicable
17:19 < fenn> that's just massively parallel computation
17:20 < fenn> was reading about larrabie yesterday
17:20 < fenn> seems like it's going to be the future for the next decade or so
17:20 < kanzure> who?
17:20  * kanzure just finds lots of stuff on linkedin
17:20 < fenn> intel's GP-GPU
17:20 < kanzure> http://gpgpu.org/
17:20 < fenn> sorry, larrabee http://en.wikipedia.org/wiki/Larrabee_(GPU)
17:20 < kanzure> I thought we already had something like GP-GPU working
17:21 < fenn> it's very limited
17:22 < kanzure> how can it be low-resolution if I'm spawning 100k+ triangle objects
17:22 < kanzure> blah
17:22 < fenn> because it uses 16-bit numbers
17:22 < fenn> i guess
17:24 < fenn> would like to read a nice simple explanation of vectorized computation
17:29 < fenn> i wonder if you could use shedskin to run your python on the GPU with CUDA
17:29 < kanzure> um, so a practical question
17:29 < kanzure> if two surfaces are mating, like a peg in a hole (but the hole has a face at the bottom)
17:30 < kanzure> how do you set that up? would you tell it to make the peg and hole bottom share a face? 
17:30 < kanzure> or just set it at some very low resolution pixel-amount away from each other?
17:30 < fenn> tell what?
17:31 < kanzure> what if you were exporting the assembly to a CFD prog?
17:31 < fenn> i have no idea
17:31 < fenn> every time you talk about that my eyes glaze over
17:31 < kanzure> talk about CFD?
17:31 < fenn> yeah
17:32 < kanzure> ok, well, say you were exporting the model to something that needs to know whether there is a closed surface or if there is a gap between the two faces
17:32 < fenn> like as an iges file?
17:32 < kanzure> ok, sure
17:32 < kanzure> I guess it would make more sense to keep things separate
17:32 < kanzure> but to what extent? and how is that determined
17:32 < fenn> usually there is some value called epsilon which defines the numerical resolution
17:32 < kanzure> because truthfully they *are* separate parts
17:32 < kanzure> oh?
17:33 < kanzure> is this a header thingy in iges files?
17:33 < fenn> if it's an interference fit, you can do two things: model the flexing of the part geometry, or make the solids intersect
17:33 < fenn> nothing is ever going to be 'exact'
17:33 < kanzure> in iges, intersecting solids are still separate entities?
17:34 < fenn> iges isn't solids i think.. it's just boundary representation (?)
17:34 < kanzure> ok, so I guess it depends on the export format
17:34 < kanzure> which is something OCC could handle
17:34 < fenn> bleh.. this is low level detail stuff
17:34 < fenn> very format dependent
18:01 -!- anthonyl is now known as anthonyl|out
18:13 -!- cis-action [n=cis-acti@31-33-205.wireless.csail.mit.edu] has quit []
18:26 -!- wrldpc [n=worldpea@pool-173-48-214-204.bstnma.fios.verizon.net] has quit []
18:33 -!- samrose [n=samrose@c-24-11-214-181.hsd1.mi.comcast.net] has joined #hplusroadmap
18:58 -!- cis-action [n=cis-acti@146-115-127-170.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com] has joined #hplusroadmap
19:07 -!- samrose [n=samrose@c-24-11-214-181.hsd1.mi.comcast.net] has quit [Remote closed the connection]
19:24 < gene> so what's the status on the fablab fenn?
19:24 < gene> any progress ont he mechmate?
19:33 -!- samrose [n=samrose@c-24-11-214-181.hsd1.mi.comcast.net] has joined #hplusroadmap
19:38 -!- samrose [n=samrose@c-24-11-214-181.hsd1.mi.comcast.net] has quit [Remote closed the connection]
19:41 -!- gene__ [n=chatzill@wireless-128-62-96-128.public.utexas.edu] has joined #hplusroadmap
19:52 -!- fenn_ [n=fenn@cpe-72-177-52-180.austin.res.rr.com] has joined #hplusroadmap
19:53 -!- gene [n=chatzill@wireless-128-62-96-128.public.utexas.edu] has quit [Read error: 110 (Connection timed out)]
19:57 -!- fenn [n=fenn@cpe-72-177-52-180.austin.res.rr.com] has quit [Nick collision from services.]
19:57 -!- fenn_ is now known as fenn
19:58 < gene__> fenn fablab status?
19:58 < gene__> how is the mech mate going?
19:58 < fenn> um, no new progress
19:58 < fenn> i haven't been there for a week; synchronization issues
19:59 < fenn> we are going to start building rooms soon i think
19:59 < fenn> today i found out whatsisname in san antonio finished the cnc router i was helping with
19:59 < fenn> steve langford
20:00 < fenn> but that doesn't really matter i guess
20:08 < kanzure> did you ever go down and check it out?
20:08 < fenn> no i just found out today
20:09 < fenn> i probably wont be driving 2 hours each way just to watch it go
20:09 < kanzure> no, I mean, did you ever go down to check 'it' out? the project
20:09 < fenn> what?
20:09 < kanzure> did anything ever happen with steve
20:09 < fenn> yeah i put his router together and he paid me a bunch of money :)
20:10 < kanzure> so you only now found out that it was put together?
20:10 < fenn> it was 90% done and i thought i fried the main chip somehow (or maybe it was a dud)
20:10 < kanzure> okay
20:10 < fenn> the deskCNC thing
20:11 < fenn> so we were waiting for that to arrive, i got distracted helping les with some metalworking for a few days, then i got on mars time and a couple weeks passed
20:11 < kanzure> gasp my IP address
20:11 < fenn> now i'm on daylight time again so i emailed him saying i could go work on it, he replied that he had finished it
20:13 < fenn> say.. are you in the new apartment?
20:15 < kanzure> “Rate-and depth-dependent nanomechanical behavior of individual living Chinese hamster ovary cells probed by atomic force microscopy.”
20:15 < kanzure> no
20:15 < kanzure> I move in May 10ish.
20:16 < fenn> "poke it with a stick and see what happens"?
20:16 < kanzure> also, we will have to hold monthly meetings to plot to take over the world or something, especially now that I have conquered these 100 square feet and am moving up to 700.
20:16  * kanzure lives in a hole :(
20:16 < fenn> i'll have my people talk to your people
20:16 < kanzure> you have people?
20:17 < fenn> well, seven's pretty smart
20:17 < kanzure> huh?
20:17 < fenn> (the bird)
20:18 < fenn> yes, the bird is my secretary
20:18 < fenn> just don't tell all the other CEO's
20:19 < kanzure> bibliography of random stuff I've been reading: http://sata.serveftp.org/~bryan/papers/2009-04-06.bib
20:21 < fenn> i hate how authors only shows their first initial
20:22 < kanzure> kind of makes the name "Y. Li" useless
20:22 < fenn> so that's generated by zotero?
20:25 < kanzure> yes
20:27 < kanzure> http://sata.serveftp.org/~bryan/papers/2009-04-06_2.bib
20:27 < kanzure> now with less hamster bullshit
20:29 < fenn> what's with the hamster cell papers?
20:30 < kanzure> it was for a post to diybio and didn't know it would clutter up my other library folder thingy
20:30 < kanzure> the second link clears it up.
20:31 < kanzure> so it looks like I'm starting to spill over on to other servers
20:32 < kanzure> well this is odd
20:32 < kanzure> Regli replied to me by email
20:33 < kanzure> but in the email he also included campbell
20:33 < kanzure> who I did not originally mention
20:33 < kanzure> was emailing Regli asking for the reeb graph software out of line?
20:35 < fenn> hell no
20:36 < fenn> if an academic publishes a paper about their software they need to share it with other researchers
20:36 < fenn> it's scientific rigor
20:36 < kanzure> but it's kind of like tattle-telling on a student
20:36 < kanzure> "haha! your student emailed me!"
20:36 < kanzure> (he doesn't actually say that)
20:36 < kanzure> (he just says provisional patent, blah blah blah)
20:36 < fenn> provisional patent?
20:36 < kanzure> on the shape similarity stuff
20:37 < kanzure> yes
20:37 < kanzure> and that's why he can't share the code
20:37 < fenn> well, that's FUCKING BULLSHIT
20:37 < kanzure> yep.
20:38 < fenn> if you want to keep secrets, don't publish papers about them
20:39 < gene__> so the router works
20:39 < gene__> does it?
20:40 < fenn> so i hear
20:40 < fenn> he's going to send some video
20:40 < gene__> is it in the lab?
20:40 < fenn> no, it's a sign making shop in san antonio
20:40 < fenn> not a community project
20:41 < gene__> so it's not going to go to the fablab?
20:41 < fenn> no
20:41 < gene__> so the fablab doesn't have a CNC machine
20:41 < fenn> yep
20:41 < fenn> and you don't have a reprap
20:42 < gene__> indeed I do not
20:42 < fenn> you know, i bet my reprap will be finished before yours
20:42 < gene__> the stratasys is still in operable
20:42 < gene__> not if I print it on the SLS machines
20:42 < fenn> for a while there i was panicked because i thought you had a working CNC machine
20:42 < fenn> but it was that other gene, the submarine propeller guy
20:43 < gene__> Submarine propellor guy?
20:43 < fenn> some comment on slashdot
20:43 < kanzure> gene__: there's a guy that comes in here when you're gone
20:43 < kanzure> since you steal his username
20:43 < fenn> heh
20:43 < gene__> oh shoot
20:43 < kanzure> but we can never really tell when it's you or him
20:43 -!- gene__ is now known as genehacker
20:43 < genehacker> I have a doppelganger?
20:43 < fenn> looks like you're still logged in from utexas
20:44 < kanzure> there's a nickserv command to kill a doppleganger
20:44 < genehacker> ok
20:44 < fenn> the other guy is registered with nickserv i think
20:44 < genehacker> well he seems like a pretty nice guy
20:44 < fenn> yeah he sure talks a lot..
20:45 < fenn> that whowas thing is annoying
20:45 < fenn> if i wanted whowas i'd type /whowas
20:45 < genehacker> so what is his relation to submarine propellors?
20:46  * fenn shrugs
20:46 < fenn> i think he was a sonar operator during the cold war
20:47 < genehacker> and what can we do to get us a CNC machine?
20:47 < kanzure> "any missiles yet?"
20:47 < kanzure> "nope, still cold"
20:47 < kanzure> *yawn*
20:47 < fenn> and was whining on slashdot about how the (koreans? japanese?) "accidentally" exported six axis CNC milling machines to soviet russia
20:47 < kanzure> oh, that post
20:47 < kanzure> yes, I remember it
20:47 < fenn> and suddenly all the subs got quieter
20:47 < genehacker> hey I got some papers on fluidic missile guidance system if you want them
20:47 < genehacker> oh yeah that guy
20:48 < fenn> so, that's the other gene
20:49 < genehacker> so we need some sort of automatic fabricator to truly be a fablab
20:49 < genehacker> how can we get one?
20:50 < genehacker> I could see if I could print of 2 sets of reprap parts...
20:50 < kanzure> genehacker: hey, you know you'd need 225 m^2 of the spiral filter design in order to process 1 L/sec
20:50 < kanzure> why did we make the spiral?
20:50 < genehacker> to see if it works
20:51 < kanzure> but you know it doesn't
20:51 < genehacker> that's why
20:51 < genehacker> I don't know if does
20:51 < kanzure> there's nothing in the literature that indicates scaling it up past laminar flows would be a good idea
20:51 < genehacker> huh?
20:51 < genehacker> what do you mean
20:51 < kanzure> I mean that everything that has used it is something like 4 microliters/sec
20:51 < fenn> you know what a laminar flow is, right?
20:51 < genehacker> yes
20:52 < fenn> and water in a 1cm wide channel won't be laminar flow
20:52 < genehacker> what about the parc paper?
20:53 < genehacker> they used something similar to this
20:53 < kanzure> 92 mm/sec
20:53 < genehacker> damn I want to take fluids in the summer
20:53 < genehacker> really?
20:53 < kanzure> no, wrong paper, sorry
20:54 < genehacker> parc paper as in the original paper I showed you guys
20:54 < kanzure> yeah, so the parc paper is the one where they were using a 10 mm by 10 mm design
20:54 < genehacker> why won't they let me take fluids in the summer?
20:54 < kanzure> 300 micrometer wide channels.
20:54 < genehacker> no they used a  bigger one
20:55 < genehacker> anyway I need to design another filter design to test
20:55 < genehacker> and also, look up the meaning of the word experiment
20:55 < genehacker> bye
20:55 < kanzure> look up the meaning of the word 'hypothesis'
21:01 < fenn> you could reproduce these methods and adapt them for the application you have in mind, but then our university's attack dogs will sue you silly
21:08 < kanzure> release the hounds!
21:08 < fenn> hm another day and i still havent put in any catalog orders..
21:08 < kanzure> to mouser?
21:08 < kanzure> erm, catalog-people
21:08 < fenn> allelectronics, dealextreme
21:09 < fenn> will be ordering from arrow and questcomp as soon as i figure out wtf i am doing
21:09 < fenn> oh and enco
21:10 < fenn> important that, since les has no raw materials and lowes wants to rape me
21:10 < fenn> $15 for two feet of grid beam
21:10 < fenn> i'd say everything in that store is about 2x the enco price
21:10 < fenn> no, worse than that
22:04 < kanzure> "Do you know of an adjective for a creature which is an emergent property of an extremely sophisticated computer network?"
22:05 < kanzure> digimon? emergicon? ghost from a shell?
22:39 -!- wrldpc [n=worldpea@pool-173-48-214-204.bstnma.fios.verizon.net] has joined #hplusroadmap
22:40 < wrldpc> http://en.wikipedia.org/wiki/E-Prime
22:50 < ybit> yup
22:50 < ybit> i saw that sometime last year
22:50 < ybit> has some good ideas, but you take it too far
22:52 < ybit> for instance at one point i was coupling it with p-prime and some other rules
22:54 < ybit> russian is similar to e-prime
22:54 < ybit> or was it p-prime. i dunno. time for bed
22:54 < ybit> gn
22:55 < kanzure> haha, here it is. http://www.youtube.com/watch?v=u0R9QtAEiQ4
22:55 < kanzure> tetris brought to life.
22:56 < kanzure> the last two seconds are hillarious