| --- Day changed Thu Apr 17 2008 | ||
| -!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has quit ["Leaving."] | 00:36 | |
| -!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has joined #hplusroadmap | 08:05 | |
| -!- Enki-2 [n=weechat@c-71-234-190-248.hsd1.ct.comcast.net] has quit [Read error: 104 (Connection reset by peer)] | 08:07 | |
| -!- Enki-2 [n=weechat@c-71-234-190-248.hsd1.ct.comcast.net] has joined #hplusroadmap | 08:13 | |
| -!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has quit ["Leaving."] | 08:39 | |
| -!- Enki-2 [n=weechat@c-71-234-190-248.hsd1.ct.comcast.net] has quit [Read error: 110 (Connection timed out)] | 10:45 | |
| fenn | 80MB movie of a nanomotor.. wtf | 16:44 |
|---|---|---|
| fenn | havent these guys heard of youtube | 16:45 |
| -!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has joined #hplusroadmap | 18:14 | |
| kanzure | What happened to everyone? | 18:14 |
| [Users #hplusroadmap] | 18:36 | |
| [ drazak] [ fenn] [ kanzure] [ krebs] | 18:36 | |
| -!- Irssi: #hplusroadmap: Total of 4 nicks [0 ops, 0 halfops, 0 voices, 4 normal] | 18:36 | |
| fenn | enki appears to be having connection issues, and epitron was being a jerk so i told him to piss off | 18:37 |
| fenn | aside from that, nobody else hangs out here regularly | 18:37 |
| fenn | i think 'you are what you cache' is stupid | 18:48 |
| fenn | it totally ignores generalization/abstraction | 18:48 |
| fenn | most people function on heuristics most of the time | 18:49 |
| fenn | caching is not the same as learning | 18:49 |
| kanzure | you should see my ~/cache/ folder | 19:18 |
| kanzure | btw, I was thinking that I would forward you all of my emails for a single day | 19:18 |
| kanzure | just to catch a glimpse | 19:18 |
| kanzure | today is a great day for this | 19:18 |
| kanzure | I need to go setup a blog | 19:19 |
| kanzure | Andrew is arguing with dad about why he should use blogspot instead of a custom-installed blog | 19:19 |
| kanzure | sigh | 19:19 |
| fenn | i cache things to preserve them for sharing with others or re-evaluation in the future, not because it makes them a part of me | 19:27 |
| fenn | i have some web pages which are total crap, which i preserved on the off-chance that they might be right | 19:27 |
| kanzure | fenn: do you not cache your cells, your food, your thoughts? | 19:44 |
| kanzure | Domain name : HEYBRYAN.ORG | 19:50 |
| kanzure | Created on : 2005-04-22 01:58:46 | 19:50 |
| kanzure | Expires on : 2012-04-22 01:58:46 | 19:50 |
| kanzure | Last Updated : 2008-04-17 23:39:47 | 19:50 |
| kanzure | Years Purchased: 4 | 19:50 |
| kanzure | fenn: do you remember "How to make a complete map of everything you think" ? by Lion Kimbro? | 19:56 |
| kanzure | http://www.speakeasy.org/~lion/nb/ | 19:56 |
| kanzure | there are also some other items to process through over at http://heybryan.org/mediawiki/index.php/2008-04-17 include coderepos - don't know if that site is legit or not | 19:57 |
| kanzure | fenn: http://heybryan.org/~bbishop/cgi-bin/blosxom.cgi search for 'apropos' on this page. It seems to be a tag-based method of searching for shell commands. This is something that we might want to integrate | 20:16 |
| kanzure | ... basically we need a streamlined command disambiguation interface so that users can plug simulatable modules together, right? Of course, they have to provide terms, or the autospec error has to provide a context. | 20:16 |
| kanzure | haha, wearable sleeping bag | 20:17 |
| fenn | it's not tag-based, apropos just searches the man-page title | 20:23 |
| fenn | title and description | 20:25 |
| kanzure | blah | 20:30 |
| kanzure | that's not good | 20:31 |
| kanzure | is it bad that I forgot where my blog was for a whole month? | 20:31 |
| kanzure | but it's alright now - I have it updating on rsync from my local cache to the blog directly, so it's all good | 20:31 |
| kanzure | I should probably have it dumping into a wiki | 20:31 |
| fenn | you could have a meta-wiki page that mirrors your blog | 20:33 |
| fenn | ... or not | 20:33 |
| kanzure | mirrors? I'll just dump into it | 20:36 |
| kanzure | because if I'm using rsync, then aren't I just doing what we want to do with ikiwiki anyway? | 20:36 |
| kanzure | it's basically a current versioning system where I assume that all of my files are up to date ;) (all of my files on the laptop, that is) | 20:36 |
| kanzure | which obviously isn't smart :) | 20:36 |
| kanzure | but humorous. | 20:36 |
| fenn | mmm aptera is the first decent car design in a while | 20:37 |
| kanzure | re: car designs; my (completely fun, entertaining, understanding, probably ASD-spectrum) psych teacher drives a box for a car, but she loves it since it gets ~38 mpg. Not aerodynamically feasible + it's fun to laugh at somebody for driving a box. | 20:38 |
| fenn | 38mpg from a box eh? | 20:39 |
| kanzure | a scion, I believe | 20:39 |
| kanzure | but that just means toyota | 20:39 |
| fenn | hmm aptera claims 300mpg? | 20:40 |
| fenn | wtf does that mean | 20:40 |
| kanzure | something in the news recently re: 2200+ mpg from a student team on ~400 kg. | 20:40 |
| kanzure | argh, I hate losing notes | 20:41 |
| kanzure | I accidentally overwrote my 2008-04-17 page on mediawiki today because some guys were talking with me while I was working, I had two windows open I think | 20:41 |
| kanzure | the guys were actually tards, but I decided to entertain them | 20:41 |
| kanzure | 'computer voodoo magic' and such. | 20:42 |
| kanzure | fenn: try: | 20:42 |
| kanzure | config = yaml.load(file('config.yaml', 'r')) | 20:42 |
| kanzure | except yaml.YAMLError, exc: | 20:42 |
| kanzure | print "Error in configuration file:", exc | 20:42 |
| kanzure | actually, | 20:42 |
| kanzure | try: yaml.load("!!python/object:__main__.Hero") | 20:42 |
| kanzure | try that, and check the 2nd line (or array index 1) and you'll see a nicely formatted error message | 20:43 |
| fenn | how do you overwrite a page on mediawiki? editing in the form? | 20:43 |
| kanzure | I was typing up an edit on one window, and another in another window | 20:44 |
| kanzure | and I saved on one, closed on another or something | 20:44 |
| kanzure | anyway, it was just that snippit of code there | 20:44 |
| kanzure | found from http://pyyaml.org/wiki/PyYAMLDocumentation | 20:44 |
| kanzure | I was playing around on the laptop with python earlier this morning, I remembered to cache the py-yaml documentation and diveintopython.org's book on to my system | 20:44 |
| kanzure | I think it's ~/cache/skdb/ atm. | 20:44 |
| kanzure | the #yaml guys don't know about parser implementation details | 20:45 |
| kanzure | (ironically, the guy answering in #yaml is a person I found in #math a few weeks ago) | 20:45 |
| fenn | how do i make an error in the config file? it just seems to eat whatever i feed it | 20:45 |
| kanzure | see my 'actually,' message | 20:45 |
| kanzure | try: | 20:45 |
| kanzure | yaml.load("!!python/object:__main__.Hero") | 20:45 |
| kanzure | because you don't have class Hero. | 20:46 |
| kanzure | thus there must be an error | 20:46 |
| kanzure | and this is the exact sort of error that we are interested in | 20:46 |
| kanzure | please don't forget 'import yaml' | 20:46 |
| kanzure | and apt-cache search yaml | grep python | 20:47 |
| fenn | yep | 20:47 |
| kanzure | I had some thoughts on the streamliner-disambiguation program for connecting different programs together or automatically selecting programs to solve autospec problems ... this streamliner should be able to be used automatically or by a user; *this* is what uses the file format IO specs, and it generates something | 20:48 |
| kanzure | it generates either a bash script or python script or something | 20:48 |
| kanzure | and this script would attach/pipeline all of the commands together | 20:48 |
| kanzure | and manage all of the files generated and make sure that they all end up at the same place | 20:48 |
| kanzure | and then *this* script, the one that is generated, is the one that the user ultimately uses | 20:48 |
| fenn | i did some help(exc) stuff and found the __dict__ which contains: | 20:49 |
| fenn | {'note': None, 'problem': "cannot find 'Hero' in the module '__main__'", 'context_mark': <yaml.error.Mark object at 0x84f072c>, 'context': 'while constructing a Python object', 'problem_mark': <yaml.error.Mark object at 0x84f072c>} | 20:49 |
| kanzure | not in all cases does the user have to disambiguate for the program - sometimes it says "well, this last module passed out XYZ and we need TUV, and there's program X that can convert between those two" and do automatic chaining | 20:49 |
| kanzure | hm | 20:49 |
| kanzure | it doesn't have 'Hero' in a variable name though | 20:49 |
| kanzure | and that's unfortunate | 20:49 |
| fenn | no | 20:49 |
| kanzure | but it's easily parseable. | 20:49 |
| kanzure | I guess we can go update the yaml code base if we want | 20:49 |
| kanzure | do I care that much? | 20:49 |
| kanzure | I can at least go suggest to the developers to output errors in yaml :) | 20:50 |
| fenn | i expect the surrounding code is much more general | 20:50 |
| kanzure | hm? | 20:50 |
| fenn | it's probably a string that python interpreter spits out | 20:50 |
| kanzure | ooh | 20:50 |
| kanzure | tokens | 20:50 |
| fenn | eh? | 20:50 |
| kanzure | erm, I'm thinking of something different | 20:50 |
| kanzure | but it might be possible to use the py-yaml parser tokens to find the class name | 20:50 |
| fenn | i mean, its not pyyaml's job to figure out what the problem is | 20:51 |
| kanzure | http://pyyaml.org/wiki/PyYAMLDocumentation#YAMLError | 20:51 |
| fenn | cannot find 'Hero' in the module '__main__' <- this is a python error | 20:51 |
| kanzure | sure, but it should report the problem | 20:51 |
| kanzure | oh? | 20:51 |
| kanzure | well then. | 20:51 |
| kanzure | then a general underlying debug yaml layer for python would be awesome | 20:51 |
| kanzure | but beyond our scope at the moment | 20:51 |
| fenn | god i dont know how people think pouring unrefined kerosene on ground dead cows is appetizing | 20:52 |
| fenn | but it stinks up the whole neighborhood | 20:52 |
| kanzure | ground dead cows? | 20:52 |
| fenn | hamburger | 20:52 |
| kanzure | from #python - (2008-04-17 19:57:01) bob2: kanzure: well, if you catch ImportError, you can parse the .msg | 20:53 |
| kanzure | try: blah | 20:54 |
| kanzure | except ImportError: | 20:54 |
| kanzure | don't know where the msg var is though | 20:54 |
| fenn | except ImportError, exc: | 20:54 |
| fenn | print exc.__msg__ probably | 20:54 |
| fenn | exc.message | 20:56 |
| fenn | No module named foo | 20:56 |
| kanzure | apparently it might be straight up msg | 20:58 |
| kanzure | just 'msg', no special namespace or whatever | 20:58 |
| kanzure | http://projecteuler.net/ - math problems for python, it looks like an evaluator server that checks your solutions to number theory problems | 20:59 |
| kanzure | some of the programming competitions out there (the 'official' ones for students etc.) use automatic 'safe-env' evaluator servers | 21:00 |
| kanzure | ex: > If we list all the natural numbers below 10 that are multiples of 3 or | 21:00 |
| kanzure | > 5, we get 3, 5, 6 and 9. The sum of these multiples is 23. Find the | 21:00 |
| kanzure | > sum of all the multiples of 3 or 5 below 1000. | 21:00 |
| fenn | i dont know where msg is | 21:02 |
| kanzure | re: eye following problems; it occurs to me that this is kind of like my multitasking. But on the other hand, multitasking to this level is common across computer users that know what they are doing. | 21:03 |
| kanzure | but I don't have eye tracking problems | 21:03 |
| kanzure | fenn: e.msg | 21:04 |
| fenn | what is e? | 21:04 |
| kanzure | oops | 21:04 |
| kanzure | (2008-04-17 20:07:46) bob2: kanzure: try: import whatever ; except ImportError, e: print e.msg | 21:04 |
| fenn | AttributeError: 'exceptions.ImportError' object has no attribute 'msg' | 21:04 |
| fenn | some code | 21:20 |
| fenn | module = __import__("package.module", globals(), locals(), ["module"]) | 21:20 |
| fenn | members = inspect.getmembers(module) | 21:20 |
| fenn | member = getattr(module, member_name) | 21:20 |
| kanzure | 1) securely? | 21:20 |
| fenn | yep | 21:20 |
| kanzure | 2) import (modulename) ==> means what? | 21:20 |
| fenn | er.. huh? | 21:20 |
| fenn | are you asking about the python statement? | 21:21 |
| fenn | "import module" | 21:21 |
| fenn | loads code from the module "module" into an object named "module" | 21:21 |
| kanzure | so if we were using $ to represent a variable, can we do import $module_name_variable ? | 21:21 |
| fenn | no, but you can do, mymodule = __import__($var) | 21:22 |
| kanzure | in php and perl you can do $$var, where $var might be 'my module name' and $$var refers to the variable named $my_module_name in that case | 21:23 |
| kanzure | can we do something like that with python? | 21:23 |
| fenn | yes, i think it's *var | 21:23 |
| fenn | never used that though, don't you know pointers are evil? | 21:23 |
| kanzure | it's not a pointer, IIRC | 21:23 |
| fenn | reference | 21:23 |
| fenn | whatever | 21:23 |
| kanzure | it's just a way for the parser to expand it | 21:23 |
| kanzure | it's like a search-and-replace deal ;) | 21:23 |
| kanzure | (2008-04-17 20:29:40) Crast: kanzure: you can use eval to access variable variables, but much better is to try to use a dict or reflect an object using getattr (2008-04-17 20:29:48) habnabit: kanzure: getattr. but using a dict is much better most times.kanzure, not exactly like that. if you want references to an object, use an object to store it. it's also common to use a list for this purpose. | 21:26 |
| kanzure | oh | 21:27 |
| kanzure | I see you're already on that | 21:27 |
| kanzure | I don't think we need to have a variable variable in this case though | 21:27 |
| kanzure | because the script would already know of the class name if it wants to use it | 21:27 |
| fenn | eval is really hard to make secure, i gather | 21:28 |
| fenn | but it's useful as a keyword for finding ways to not use eval :) | 21:28 |
| kanzure | hrm | 21:29 |
| kanzure | tell me how this would be secure anyway | 21:29 |
| kanzure | it's not like you can entirely trust APT either.. | 21:30 |
| fenn | code modules that you download from us would be signed, meaning they dont intentionally do bad things to your computer or privacy or whatever | 21:30 |
| kanzure | please explain | 21:31 |
| kanzure | I am not familar with security models, I just assume insecurity and work from there | 21:31 |
| fenn | but anyone could modify code on the wiki, so we'd have to periodically check incoming code to see what it does, and package it into a 'release' to be signed | 21:31 |
| fenn | otherwise we get the 'commit access syndrome' | 21:31 |
| kanzure | what does it mean for it to be signed | 21:31 |
| kanzure | a bit to be checked yes/no? | 21:31 |
| kanzure | key system? | 21:31 |
| fenn | you do a md5sum or SHA and if it matches, it matches | 21:32 |
| kanzure | you would get the md5sum from the server too | 21:32 |
| kanzure | what's the point? | 21:32 |
| kanzure | (the one to check it against) | 21:32 |
| fenn | somehow that gets confounded with your key, so in effect you're saying 'this is what i intend to distribute' | 21:32 |
| kanzure | why can't somebody fake my key? | 21:33 |
| fenn | because it's a key | 21:33 |
| fenn | the client will have your public key, which can tell if the private key (confounded with the signature) is correct | 21:34 |
| kanzure | so if they can tell whether or not it is correct, then why not just bruteforce your private key? | 21:34 |
| kanzure | erm | 21:34 |
| fenn | nice long article, if you really want to know http://en.wikipedia.org/wiki/Public-key_cryptography | 21:34 |
| kanzure | bruteforce somebody else's private key? | 21:34 |
| fenn | because it will take 10^90000 years | 21:34 |
| fenn | depending on how many bits in the key, how many computers bruteforcing, etc | 21:35 |
| fenn | in general, verifying a key takes a lot less time than bruteforcing it | 21:35 |
| fenn | sound familiar? | 21:35 |
| kanzure | yes | 21:36 |
| fenn | i guess someone was making a joke when they said *var | 21:39 |
| fenn | (this is why i hate perl, you cant search for punctuation marks) | 21:40 |
| kanzure | http://heybryan.org/music/radio/ take your pick | 21:46 |
| kanzure | fenn: I never got into the whole pgp thing, mostly because I never bothered to figure out how to | 21:47 |
| kanzure | it looks like we'll want to encrypt all outgoing modules that are picked up or something? | 21:48 |
| kanzure | or store encrypted double versions? | 21:49 |
| kanzure | or just encrypt the hash? | 21:49 |
| kanzure | blah, I don't care | 21:49 |
| kanzure | you can deal with that | 21:49 |
| fenn | i'll delegate it | 21:50 |
| fenn | it doesnt matter at this point anyway | 21:50 |
| kanzure | okay, so, random brainstorming time | 21:50 |
| kanzure | get the module name that makes for the error, then proceed to check the plugin directory for that module | 21:51 |
| kanzure | load it up if we have it | 21:51 |
| fenn | btw, APT handles all this crypto stuff for us, which is part of why i wanted to use it to download software modules | 21:51 |
| kanzure | if we don't have it, start the disambiguation routine | 21:51 |
| fenn | we can just add the plugin directory to $PYTHONPATH | 21:51 |
| fenn | then python will find it | 21:51 |
| kanzure | the disambiguation software would be a way to either (1) automatically download new modules and programs to splice things together or (2) allow the user to select what's the next step (just in time programming, call it) | 21:51 |
| kanzure | yes | 21:51 |
| fenn | if we dont find the module, then we just download it | 21:52 |
| kanzure | but do we want python to load up all modules at once? | 21:52 |
| fenn | it wont load all modules at once? | 21:52 |
| fenn | hrm | 21:52 |
| kanzure | I am wondering if we should conflate mime db == the python plugins | 21:52 |
| kanzure | because the mine db idea was to use file format io specs, i.e. use this program with this input type | 21:52 |
| fenn | we have an 'installation' which is everything the user has and can do | 21:52 |
| kanzure | and clearly here we are using yaml to specify what sort of object the data is | 21:52 |
| kanzure | right | 21:52 |
| kanzure | but the user is not necessarily using everything | 21:53 |
| kanzure | okay, think of it this way | 21:53 |
| fenn | we check the installation against the technology dependency tree to build a sort of customized tree saying what the user can do | 21:53 |
| kanzure | each separate script and program | 21:53 |
| kanzure | has their own data objects | 21:53 |
| kanzure | so these programs are responsible for their modules and objets | 21:53 |
| kanzure | *objects | 21:53 |
| kanzure | but beyond this, they shouldn't really care, and instead call another program | 21:53 |
| kanzure | which is a requirement for the software to run (this would be in the meta data) | 21:53 |
| kanzure | right? | 21:53 |
| kanzure | why do we need this serialization, again? | 21:53 |
| fenn | so people dont have to edit code | 21:54 |
| fenn | its like wiki markup vs html | 21:54 |
| kanzure | okay | 21:55 |
| kanzure | but this is just for one module at a time though | 21:55 |
| kanzure | so each package is independently responsible for this right? | 21:55 |
| fenn | what are you referring to with 'module'? | 21:55 |
| kanzure | one package | 21:55 |
| fenn | technology package or code to interpret specifications? | 21:55 |
| fenn | i dont want any code in the technology package if i can avoid it | 21:56 |
| kanzure | man, I just lost my train of thought | 21:57 |
| kanzure | WTF are we doing? | 21:57 |
| kanzure | why do we have serialization? the serialization will *only* be for loading up information that the package can interpret itself, right? | 21:57 |
| kanzure | so why *wouldn't* the immediate code specify the classes? | 21:57 |
| fenn | uh.. because people are stupid i guess.. i dont know | 21:58 |
| kanzure | then use a converter | 21:59 |
| fenn | convert what to what? | 21:59 |
| fenn | think of yaml like a sandbox that people can do whatever they want without breaking things | 21:59 |
| fenn | we'll be validating the data structure it creates before running it | 22:00 |
| kanzure | what will this data struct be used for | 22:00 |
| fenn | if people just write raw python instead, they can put whatever they want in there | 22:00 |
| fenn | the data struct will contain specifications about what the technology provides and requries | 22:00 |
| fenn | the specifications need to be compatible for the entire system to be valuable | 22:01 |
| fenn | if everyone writes their own specification class, they cant be compared, it's apples and oranges | 22:01 |
| fenn | if the default level of interaction is writing a yaml file, they have to include one of the provided code modules, and it's harder to make a new code module | 22:02 |
| fenn | so people use an existing specification standard instead of making up a new one for each package | 22:02 |
| fenn | make sense | 22:02 |
| fenn | ^? | 22:02 |
| fenn | the code module also includes some sanity checks and unit tests that get run automatically | 22:03 |
| fenn | believe me, you dont want to sort code | 22:03 |
| fenn | sorting a bunch of yaml files is easier because it's dead data | 22:04 |
| kanzure | oh, right | 22:05 |
| fenn | also, you can present it in different ways, such as with ikiwiki | 22:05 |
| kanzure | we want to use data serialization for a Common Object Library | 22:05 |
| kanzure | there is another aspect of the project, then, that deserves mentioning | 22:05 |
| fenn | this is some microsoft-related concept? i dont know what that really means | 22:05 |
| kanzure | have you ever used aptitude's disambiguation program? it's an ncurses interface that lists potential matching packages | 22:06 |
| kanzure | Common Object Library: | 22:06 |
| kanzure | there's a set of common objects that you can use | 22:06 |
| kanzure | it's a library | 22:06 |
| kanzure | like sets of header files that you can choose to include | 22:06 |
| kanzure | the python plugins are the header files | 22:06 |
| kanzure | the yaml files contain instances (objects) of the classes in the python plugins / header files | 22:07 |
| fenn | yep | 22:07 |
| kanzure | okay, so the disambiguation program -> you would use this for something else, I was conflating two things together | 22:07 |
| kanzure | my 'file format io specs' is different | 22:07 |
| kanzure | that's where you splice and dice programs together | 22:07 |
| kanzure | where the program can automatically find a path to another format if needed | 22:08 |
| fenn | i was mostly thinking of the file format db for interacting with third party apps | 22:08 |
| kanzure | or the user can sit there and try out a suggested option | 22:08 |
| kanzure | well | 22:08 |
| kanzure | sort of | 22:08 |
| kanzure | think of it this way | 22:08 |
| fenn | which makes it sound more like a traditional plugin API | 22:08 |
| kanzure | many of the scripts added to skdb will sometimes be reworking of the other skdbs | 22:08 |
| kanzure | erm | 22:08 |
| kanzure | sequences of other packages | 22:08 |
| kanzure | sorry | 22:08 |
| kanzure | anyway, these sequences will just direct flow between various modules, yes? | 22:08 |
| kanzure | *various software modules | 22:08 |
| kanzure | this is essentially what the user wants to do for a particular project | 22:09 |
| kanzure | so why not let them choose what to do with output by giving them a list of programs that use that format as input | 22:09 |
| fenn | you cut and stiched those lines so many times i dont know what you just said | 22:09 |
| kanzure | thus the file format io spec metadata (not just package .deb metadata stuff) | 22:09 |
| kanzure | hm | 22:09 |
| kanzure | okay, so imagine it this way | 22:09 |
| kanzure | you just downloaded the CNC lathe tool | 22:10 |
| kanzure | and all of the python modules, .deb packages (except .skdb equivalents), etc. | 22:10 |
| kanzure | your simulated CNC lathe tool is simulated with certain runtime parameters and spits out a file that represents a finished work | 22:10 |
| fenn | agx-get install lathe tool // means i have a cnc lathe | 22:10 |
| kanzure | let's say a chair | 22:10 |
| fenn | agx-make chair | 22:10 |
| kanzure | and so this is yaml instance data | 22:10 |
| kanzure | then you're left sitting there with a theoretical (but fully physically validated, ready to be made exactly to specs) chair | 22:11 |
| kanzure | so now what? | 22:11 |
| kanzure | well, that's where the file format io db disambiguation stuff comes in at | 22:11 |
| kanzure | you would say "what programs can use <this yaml> ?" | 22:11 |
| kanzure | and it would say "well, html2pdf takes HTML yaml, here's its description yadda yadda" and it'd be a selection interface | 22:11 |
| kanzure | now, when I mean to say I was conflating the ideas together | 22:12 |
| kanzure | as in, I was thinking that the yaml object loading stuff from autospec would be associated with the file format io specs | 22:12 |
| fenn | cnc-lathe requires g-code at technology-build-time | 22:12 |
| fenn | er, crap | 22:12 |
| kanzure | yes | 22:12 |
| fenn | chair requires g-code at build-time, cnc-lathe requires g-code at run-time | 22:12 |
| fenn | right? something like that | 22:13 |
| fenn | or maybe chair provides gcode | 22:13 |
| fenn | this is why we write code :) | 22:13 |
| kanzure | you mean: physical expression of 'chair package' requires g-code at cnc-physically-expressed building-time | 22:13 |
| fenn | the word chair can refer to the software describing the chair, the physical chair itself, (others?) | 22:14 |
| fenn | let's conflate the physical chair and the technology package for chair | 22:14 |
| kanzure | simulated chair, g-code for the chair, the physical chair itself (as in, what you are editing in realtime with the CNC lathe spinning around and whatever), the entire 'chair' package and all g-code and software and python plugins and object instances (aka yaml files), ... | 22:14 |
| fenn | so we can say 'chair provides g-code at build-time' | 22:15 |
| kanzure | build-time can mean so many things though | 22:15 |
| kanzure | build-time of the simulation system | 22:15 |
| kanzure | build-time of "chipping and chopping away at wood to make the chair" | 22:15 |
| fenn | terminology.. what's a short name that differentiates between compiling and robotics stuff | 22:15 |
| kanzure | eh | 22:15 |
| fenn | thesaurus time :) | 22:16 |
| fenn | construct, fabricate, manufacture, | 22:16 |
| kanzure | it's more like | 22:17 |
| kanzure | the problems with talking about cyber->reality borders | 22:17 |
| kanzure | cyber<->physical-materials | 22:17 |
| kanzure | bits<->atoms | 22:17 |
| kanzure | http://cba.mit.edu/ Center for Bits and Atoms | 22:17 |
| fenn | hey isnt that the whole point now :) | 22:17 |
| kanzure | http://digivin.googlepages.com/ aww, Andrew just linked me to his new website | 22:17 |
| fenn | because right now cad-files are so not automated it's sick | 22:18 |
| kanzure | in general | 22:18 |
| fenn | so there's no "source code" to the manufacturing process | 22:18 |
| kanzure | I think that our system idea is going to work | 22:18 |
| kanzure | however, the more we obsess on those details, the less we get done | 22:18 |
| kanzure | I think our system is modular enough that it can be implemented on top of our basic ideas | 22:18 |
| kanzure | yes? | 22:18 |
| kanzure | on top of the basic idea of fetching data types from the web | 22:18 |
| fenn | yep sure | 22:18 |
| kanzure | on top of the basic idea of fetching packages from the web | 22:18 |
| fenn | i wouldnt be doing it if i didnt think it would work | 22:19 |
| fenn | shall we diagram or write code? | 22:19 |
| fenn | top down or bottom up | 22:19 |
| kanzure | you want to stop talking about it and just do it? | 22:20 |
| kanzure | wait | 22:20 |
| kanzure | I should have worded that | 22:20 |
| kanzure | you wanna stop talking and just code? | 22:20 |
| fenn | sometimes i get lost though if there's no diagram | 22:20 |
| kanzure | right | 22:20 |
| kanzure | it's quite a big problem space to contain in the head all at once | 22:21 |
| * fenn gets a piece of paper | 22:21 | |
| kanzure | and in fact, I am not so sure that other programmers try to cram an entire vision into their head before starting to write code | 22:21 |
| fenn | meh | 22:22 |
| kanzure | btw, I carry stacks of white printer paper with me everywhere I go | 22:22 |
| kanzure | screw lined paper | 22:22 |
| kanzure | I know how to be straight on my own, thanks | 22:22 |
| * fenn translates hand-sketched diagram into dia format | 22:24 | |
| kanzure | what happened to the UML craze for software engineering? | 22:25 |
| fenn | it turned out you cant automatically turn crappy diagrams into code | 22:26 |
| fenn | else the map is the territory | 22:26 |
| fenn | easier to write code and draw a broad overview, than to make a detailed diagram | 22:27 |
| kanzure | how are the python plugins served? | 22:37 |
| kanzure | are they treated as packages too? | 22:37 |
| kanzure | obviously these are significantly restricted packages in comparison to something like a 'metalcastingtechniques' package | 22:37 |
| kanzure | (which may include exact methodology for making a machine to do metalcasting) | 22:37 |
| kanzure | I suppose the answer is yes | 22:38 |
| kanzure | since the plugins can be of a certain python-class (skdbPythonPluginClass for now) | 22:38 |
| kanzure | so you'd serialize that object to yaml, and then you can load that data up or something | 22:39 |
| kanzure | hrm | 22:39 |
| kanzure | that doesn't make much sense | 22:39 |
| fenn | http://fennetic.net/pub/irc/autogenix.svg (and .dia if you want it) | 22:39 |
| fenn | sry that took so long, is reason i used paper | 22:39 |
| kanzure | if the plugins are a package in skdb, then how are they served? if autospec finds it needs to get a specific plugin, then ... hm. I guess the solution is to have a package that has a type of 'python plugin' which has a common variable specifying the URL to get the plugin source code. Then just make sure your autospec program can use that 'python plugin type'. | 22:40 |
| fenn | oh ffs the svg is crap | 22:40 |
| kanzure | shit Opera doesn't understand SVG | 22:41 |
| kanzure | have a suggestion for an SVG viewer? | 22:42 |
| kanzure | dia is it? | 22:42 |
| fenn | dia just dropped about 90 geek points | 22:43 |
| fenn | for shitty svg export | 22:43 |
| fenn | http://fennetic.net/pub/irc/autogenix.png | 22:43 |
| drazak | Oh god, geek points! | 22:44 |
| drazak | dia-- | 22:44 |
| fenn | dia *= 0.1 | 22:44 |
| kanzure | fenn: Paul (OSCOMAK.paul)'s geek points are through the roof. He studied at Caltech in a building called von Neumann Hall. | 22:44 |
| fenn | where they disdained space travel and computer information networks and replicating machines, sure | 22:45 |
| fenn | fuck those assholes | 22:45 |
| fenn | fuck 'em hard! :) | 22:45 |
| fenn | they humor the geeks by promising space travel, when really its just a nuclear arms race | 22:46 |
| kanzure | hah, "so we decided to name this building after von Neumann just in spite" | 22:46 |
| kanzure | cruel | 22:46 |
| kanzure | but really, answer the question | 22:46 |
| kanzure | how are the python plugins served? are they treated as packages in skdb/APT-server-side-equivalent? | 22:47 |
| fenn | we could setup a debian repository and have the autospec download autogenix-$modulename package | 22:47 |
| fenn | or we could do some fancy git stuff? | 22:47 |
| kanzure | and if so, then the meta-data for that package would simply be something standard that all autospec/autogenix things *must* understand from version 1.0 and up | 22:47 |
| kanzure | well | 22:47 |
| kanzure | I think it should be a package | 22:47 |
| kanzure | since it's kinda like a header | 22:48 |
| fenn | dunno, how does gentoo handle this stuff? | 22:48 |
| kanzure | weird, I just had a meta-philosophical insight on the nature of bootstrapping compilers, computational complexity, and well-defined fabrication automation | 22:49 |
| kanzure | it's *fundamentally hard* to describe this sort of system since you have to have such unique vocabulary | 22:49 |
| kanzure | you have to just sit down and *fucking code it* | 22:49 |
| kanzure | you *cannot* verbally detail it, that is implicitly impossible | 22:49 |
| kanzure | but anyway | 22:49 |
| fenn | you mean empirical data is in the loop? | 22:50 |
| kanzure | yes | 22:50 |
| drazak | gentoo handles everything greatly :D | 22:50 |
| fenn | or, build, test, back to the drawing board? | 22:50 |
| drazak | What are you guys trying to do? | 22:50 |
| kanzure | drazak: Wait, haven't we explained yet? | 22:51 |
| drazak | kanzure: which part? and I haven't read a lot of backog, either | 22:51 |
| fenn | drazak: apt-get for rapid prototyping (and related activities) eventually leading to self-replicators | 22:51 |
| kanzure | fenn: Sort of, yes. The only way to 'fully realize it' and 'remember all at once' is to just make it. You can't do a complete description of it, like you said. UML stuff. So therefore. | 22:51 |
| fenn | or, 'emerge' if you prefer | 22:51 |
| kanzure | so that's why I was getting to the "*must* understand from version 1.0 and up" | 22:52 |
| kanzure | you have to have the bootstrap stuff in there | 22:52 |
| kanzure | luckily our gcc friends boostrapped the compilers for us (hehe) | 22:52 |
| kanzure | and bootstrapped operating systems for us | 22:52 |
| fenn | we have no bootstrapped rapid prototyper | 22:52 |
| kanzure | which is of course what we're doing here | 22:52 |
| kanzure | bruteforced boostrapping ;) | 22:53 |
| fenn | no, we're writing the OS for a rapid prototyper, basically | 22:53 |
| kanzure | werm | 22:53 |
| kanzure | bootstrapping | 22:53 |
| kanzure | hm | 22:53 |
| kanzure | perhaps | 22:53 |
| kanzure | the analogy doesn't completely hold of course | 22:53 |
| kanzure | because if you downloaded and installed and configured all skdb software into the prototyper, | 22:53 |
| kanzure | that would be lots of waste methinks | 22:53 |
| fenn | but some of the code can be applied to regular humans in the meanwhile, like von nuemann running his cellular automata on paper | 22:53 |
| fenn | or turing running a LCM on paper | 22:53 |
| kanzure | was he really doing 60x60 cell states on paper? by hand? | 22:54 |
| fenn | dunno | 22:54 |
| * fenn points at the mythos | 22:54 | |
| kanzure | that's ... sick. | 22:54 |
| fenn | its not hard actually | 22:54 |
| fenn | just repetetive | 22:54 |
| kanzure | :) | 22:54 |
| fenn | repetetetititive | 22:54 |
| kanzure | hehe, I find myself saying that sometimes | 22:54 |
| fenn | so, the hard part will be convincing people without a rapid prototyper that this is worthwhile, rather than just googling for 'fart catcher' or whatever they're making | 22:55 |
| kanzure | nah | 22:56 |
| kanzure | that will not be too hard | 22:56 |
| kanzure | basically what we do is formalize a few projects | 22:56 |
| kanzure | something appealing | 22:57 |
| fenn | b/c rapid prototypers are more or less useless if you have to do everything by hand, and nobody is interested until they have a rapid prototyper laying around (idle) | 22:57 |
| kanzure | and start attracting attention in that way | 22:57 |
| drazak | (rapid prototyper? :P) | 22:57 |
| fenn | see how reprap is floundering, despite massive hype | 22:57 |
| kanzure | drazak: 3D printer | 22:57 |
| kanzure | think of that | 22:57 |
| kanzure | fenn: yes | 22:57 |
| drazak | ohhhhhh | 22:57 |
| drazak | yeah, seen those | 22:57 |
| kanzure | fenn: I am amazed at reprap. it's a self-replicating viral meme, that's sure enough | 22:57 |
| fenn | the meme was around before the reprap | 22:57 |
| drazak | the college around here has one | 22:57 |
| fenn | drazak: did they make anything with it? | 22:58 |
| kanzure | http://www.last.fm/user/superkuh "I enjoy recursion, dissipating local energy gradients, lipid bilayers, and biophysics in general." | 22:58 |
| drazak | fenn: they've done a few things | 22:58 |
| drazak | fenn: one of my teachers at the middle school was looking to get one for the school | 22:58 |
| fenn | did they make parts for another reprap!!!!! | 22:58 |
| kanzure | "I enjoy recursion" <-- a good line to copy around the web | 22:58 |
| drazak | for our science olypiad team | 22:58 |
| fenn | nobody seems to be doing that, which is the stated goal, fuckers | 22:58 |
| kanzure | isn't it olympiad? | 22:58 |
| drazak | yeah, missed a letter | 22:58 |
| kanzure | but you've done so consistently | 22:59 |
| fenn | you dont 'get one' you have to make it, because nobody is making parts with their repraps (because the overall design sucks (not the implementation)) | 22:59 |
| kanzure | hrm, re-order: but you've consistently done so | 23:00 |
| fenn | spelling is arbitrary | 23:00 |
| kanzure | 'do what I mean, not what I say' | 23:01 |
| fenn | nonetheless, people who write 'hi how r u 2day' are fucking idiots | 23:01 |
| fenn | i need to go dissipate some local energy gradients, bbiab | 23:01 |
| drazak | I agree | 23:02 |
| kanzure | hm | 23:04 |
| kanzure | I have an idea | 23:04 |
| kanzure | let's call these "stoppers" in the netwrok | 23:04 |
| kanzure | *network | 23:04 |
| kanzure | basically, I was considering the case where we have a python-plugin package with metadata | 23:04 |
| kanzure | the user progs would have to be able to interpret the metadata | 23:04 |
| kanzure | but this metadata might require another package | 23:04 |
| kanzure | and then another and another | 23:04 |
| kanzure | it's obviously recursive and that's fine, sure | 23:04 |
| kanzure | but at some point you have to just *stop* and download some final code | 23:05 |
| kanzure | I propose we call this a 'stopper' | 23:05 |
| kanzure | this is a type of meta-data file that is the most ridiculously basic file ever | 23:05 |
| kanzure | this is the one that points to one URL and has no other data what's so other | 23:05 |
| kanzure | just file.txt: http://heybryan.org/code.py | 23:05 |
| kanzure | that's all it has | 23:05 |
| kanzure | then, the user-progs download that code and run it | 23:05 |
| kanzure | this is the way to terminate the endless chain of program requests and requirements upon requirements | 23:06 |
| kanzure | I also propose two layers for skdb servers (not the ikiwiki software stuff) | 23:09 |
| kanzure | skdb layer 1 -- *raw* data, this is the .py files and the .yaml files and even zip files of HTML and PDFs for human processing | 23:09 |
| kanzure | in the ideal situation, those would be assembled into zip files and stored locally on our servers | 23:10 |
| kanzure | skdb layer 2 - the metadata files. These would be cached by all users. They point to the resouces in layer 1. | 23:10 |
| kanzure | and also specify the hell out of other stuff and so on :) | 23:10 |
| kanzure | there is a core set of functionality that autogenix and other related tools *must* have in order to bootstrap the system | 23:14 |
| kanzure | the server-side stuff doesn't have to be bootstrapped, it's pretty easy really | 23:14 |
| fenn | back. one problem with apt is it requires superuser access | 23:15 |
| kanzure | don't see why that would be required here | 23:16 |
| kanzure | I also think we can pretty easily do away with apt as the frontend | 23:16 |
| kanzure | I mean, the frontend stuff sounds simple to me - as far as APT goes | 23:16 |
| fenn | it shouldnt be required, but that's how it works (because 99% of packages modify the installed system) | 23:17 |
| kanzure | as for the extra stuff - like being able to interpret the metafiles and being able to update (by reading the 'stoppers' on skdb) - that's a bit more tricky | 23:17 |
| fenn | we wouldnt use apt for _our_ frontend, but for installing software (deal with key signing and software depenency resolution) | 23:17 |
| kanzure | ooh | 23:19 |
| kanzure | here's a tricky loop | 23:19 |
| kanzure | program io specs (like what file formats the program accepts as input for what parameters) can be in yaml | 23:19 |
| kanzure | *but* py-plugins (that provide the tools to interpret yaml) have yaml metadata | 23:19 |
| kanzure | and pyplugins help to interpret prog io specs (loop back to square one) | 23:20 |
| kanzure | thus I think I have proven the need for bootstrapping | 23:20 |
| fenn | uh, this is no big deal, you just download all the software for the base system in one go | 23:21 |
| kanzure | correct | 23:24 |
| fenn | re: layer 1/2; do you need the data (layer 2) to make a dependency tree? if not, how do simulators fit into the picture? | 23:24 |
| fenn | i guess simulators could get run by the package developer, to automatically generate metadata | 23:25 |
| fenn | so the metadata files wouldnt be the exact same files as the source yaml on the wiki, but would be derived from them | 23:26 |
| kanzure | huh? hold on | 23:27 |
| kanzure | you need layer 2 to make the dependency tree | 23:27 |
| kanzure | layer 2 includes all of the yaml, I would think | 23:27 |
| kanzure | on the other hand | 23:27 |
| kanzure | on layer 1 there would include yaml for other things wouldn't it? | 23:27 |
| kanzure | that could be a different dependency tree, perhaps? | 23:28 |
| kanzure | I thought that simulators would be ran at user-end? | 23:28 |
| fenn | why would the user run simulations? | 23:28 |
| kanzure | the metadata files should be on the wiki | 23:28 |
| kanzure | fenn: ex: mission optimization? | 23:28 |
| kanzure | feasability simulations? | 23:29 |
| fenn | autogenix already says 'its feasible' | 23:29 |
| fenn | autospec | 23:29 |
| fenn | bah | 23:29 |
| kanzure | are you sure? | 23:29 |
| fenn | they could doublecheck | 23:29 |
| kanzure | autospec does that for file formatting | 23:29 |
| kanzure | and yaml and so on and making sure you have the available modules | 23:29 |
| kanzure | but then this does not tell you whether or not you can provide enough 'juice' or other various characteristics | 23:30 |
| kanzure | right? | 23:30 |
| fenn | no, autospec reads the metadata and builds a dependency tree, in order to say whether something is feasible or not | 23:30 |
| kanzure | oh | 23:30 |
| fenn | it determines if there's enough juice | 23:30 |
| fenn | what else would it do? | 23:30 |
| kanzure | check if the yaml file parses and whether or not it needs to go get more python plugins from the repository | 23:30 |
| fenn | file format is something else, interfacing with simulators and crap | 23:30 |
| fenn | yes, it needs to get updated plugins, correct | 23:31 |
| kanzure | somehow saying that no end-user does simulations sounds fishy/wrong | 23:32 |
| kanzure | not to mention the new issues as to how autospec would do a full feasability check ... that's just doing metadata dependencies; you'd have to encode "I need 20 amps" as an object ... which seems nasty. | 23:32 |
| fenn | yes, you have to encode 'i need 20 amps' | 23:33 |
| fenn | otherwise you can only say 'i need powersupply xyz20' | 23:34 |
| kanzure | so it would be through the python plugins that autospec would learn how to check specs for amperage and so on? | 23:35 |
| kanzure | 'learn' | 23:35 |
| kanzure | ouch | 23:35 |
| kanzure | I don't mean to use the word learn | 23:35 |
| kanzure | if epitron was here he'd say something about brains | 23:35 |
| fenn | heh | 23:35 |
| fenn | yes autospec needs plugins to understand what '20 amps' means | 23:35 |
| fenn | also what 'i need' means | 23:36 |
| fenn | probably expressed something like peak or continuous | 23:36 |
| kanzure | soo | 23:37 |
| kanzure | in other words | 23:37 |
| kanzure | autospec needs to have a standard library of simulation tools to work with those spec-params or whatever | 23:38 |
| kanzure | in other words | 23:38 |
| kanzure | a simulator | 23:38 |
| kanzure | but this is a global simulator, it seems, all in one place | 23:38 |
| kanzure | which is different (+better) from my original idea of encoding the simulators into the chair.py code itself | 23:38 |
| kanzure | this way, the python plugins provide ways to 'deal individually' with aspects of a part | 23:38 |
| kanzure | and it could be a graph-based traversal or something | 23:38 |
| kanzure | where it just satisfies a certain region of the graph/tree | 23:38 |
| kanzure | and then moves on to another part | 23:38 |
| kanzure | but of course, doing a very deep search through possibility space would be more intense | 23:39 |
| kanzure | and I don't know how to make a common interface for doing *all* types of walks through the possibility space of verification or whatever | 23:39 |
| fenn | i'm not sure graph traversal is really the best way to imagine it | 23:39 |
| kanzure | know what I mean? | 23:39 |
| kanzure | it has to meet requirements, right? so when you meet one requirement, you have to make sure you can meet the others at the same time | 23:39 |
| kanzure | kind of like balancing a chemical equation | 23:40 |
| fenn | sure, and also find the shortest distance | 23:40 |
| kanzure | uh | 23:41 |
| kanzure | so what are you proposing | 23:41 |
| kanzure | I mean, it sounds like you've completely made the entire skdb project sound pointless | 23:41 |
| kanzure | and made autospec do all of the effort | 23:41 |
| kanzure | while not specifying how it could *possibly* work with all sorts of varied packages and so on | 23:42 |
| kanzure | I guess the 'units' stuff can be encoded in yaml and python plugins and the packages in skdb | 23:42 |
| kanzure | and then you just try to find ways to disambiguate the units | 23:42 |
| kanzure | into common expressions of the same exact things | 23:42 |
| kanzure | kind of like finding the greatest common denominator | 23:42 |
| kanzure | and then factoring it out of the entire picture | 23:42 |
| fenn | there are base units that can't be factored out | 23:43 |
| fenn | believe it or not, there is no difference between a joule and a kg*m^2/s^2 | 23:44 |
| fenn | i dont see how it's pointless to boil down equivalent things | 23:44 |
| fenn | and then say 'feasible' if they're equivalent (or satisfied conditions) | 23:45 |
| fenn | like, power provided must be greater than power required | 23:45 |
| fenn | error provided must be less than tolerance required | 23:46 |
| kanzure | okay | 23:46 |
| kanzure | makes sense now | 23:46 |
| kanzure | http://heybryan.org/docs/cache-index.txt | 23:48 |
| kanzure | my ~/cache/ | 23:48 |
| kanzure | does not include my 30 GB of texts | 23:49 |
| fenn | just looks like bad organization? i dont get it | 23:52 |
| fenn | obviously these are more important than the other 30GB | 23:52 |
| fenn | some of them are just trash though, waiting to be cleared out | 23:53 |
| fenn | you could sort a database chronologically and by importance at the same time, with a 3d view or by changing font size | 23:56 |
Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!
