14:09 -!- Irssi: Starting query in freenode with CobraCommander 14:09 howdy 14:09 hi 14:10 so ... Im considering writting my own language :P 14:10 lol 14:11 motivations? 14:11 fame 14:11 ? 14:11 you know, women and money 14:11 more seriously the truth is I don't know if I want to write a "language" 14:11 maybe I just want a software, a library, a method or even a paradigm 14:12 and the more I digg in it, the less I know 14:12 and what capabilities do you want in such a thing? 14:12 to let me think about the problem in what I consider a better way 14:12 I dont really care about efficiency/speed 14:13 so I started to look at Lua, D, Gremlin 14:14 (Cobra, obviously ;) 14:14 but my question is more, what was for you the threshold 14:14 the justified decision that made you think "ok, enough, time to write my own." 14:15 and precisely, why a language rather than method/framework/library 14:15 well first thing 14:15 regarding efficiency, it *does* come up when you have to wait for your program to plow through lots of data 14:15 and that can inhibit your productivity 14:15 yes 14:15 that's why i ultimately rejected Python (and Ruby et al) 14:15 regarding your question 14:16 i was bouncing between Python and C# at the time 14:16 Im not saying it doesn't matter but I think for me right now that's not the bottleneck 14:16 ok 14:16 i was frustrated that Python's speed was slowing me down 14:16 i was frustrated that C# wasn't as productive during coding as Python... wasn't as high level 14:16 although it's gotten closer over time 14:16 i had already been thinking about making a language for a long time 14:16 but still i resisted 14:17 right 14:17 in 2005 i looked at languages that might be close enough 14:17 these included Io, D, Qu among others 14:17 they all had various problems 14:17 (Ill have to check Qu)) 14:17 all of them had maturity issues and it was not clear how long before they cleared up 14:17 i think Qu is totally dead 14:18 finally i hit a tipping point and kicked off the project in Jan 2006 14:19 so what kind of things would you do in your language or library or whatever that would be different than say Cobra or Python or Java or Ruby 14:19 think is the community matters a lot (paradoxically, for a disciplined often thought of as associal) but when you start your own you somehow trust the community as... well it's yourself 14:20 when you start your own, you take responsibility for leading the community and sticking with the project (or passing it on to someone) 14:21 what kind of thoughts are you having about capabilities and expressiveness and so on? 14:22 (damn he won't let me be elusive ;) 14:22 well, you know dataflow and pipes quite a bit I guess 14:22 :-D 14:22 so some kind of data flow language 14:22 I think it's a great principle but tricky to use 14:22 for AGI specifically or for anything? 14:23 hmmm for anything but I guess (and hope) for AGI it should work really well 14:23 mostly for "difficult problems" 14:23 that's why I don't focus on speed right now, rather on conception/modelling 14:24 so in biology you have phylogeny 14:24 formalized as a tree of organisms sharing part of their DNA 14:24 and to me, that's dataflow 14:24 and that's how I see...well everything 14:25 i assume you have reviewed http://en.wikipedia.org/wiki/Dataflow_programming 14:25 so when I have to write code and have to solve a problem thought OOP for example it's tricky 14:25 and the various things it refers to 14:25 hmm objects surround us 14:25 i bet you are sitting on a chair right now 14:25 well to me objects are inscribed in time 14:25 but hey i don't want to discourage you from finding a different paradigm for getting computers to do our dirty work 14:26 and thus not putting object in a time situation is a rather fundamental flow 14:26 flaw ;) 14:27 so yeah if you're still formulating what the language would be like then you can skip performance concerns right now 14:27 just keep them in mind in your design for the future. ;-) 14:28 you any example programs or syntax ? 14:28 right, if I can propose a solution in 5min but it takes 50years to execut it... that's a fundamental flaw too ;) 14:29 yet at least, I would have a model to improve upon 14:29 nothing wrong if the prototype is slow if you at least know how it could be sped up 14:30 I think dataflow principles are rather adapted to that, from FPGA to map/reduce, no? 14:31 yeah but i don't know *exactly* what you have in mind so i was just commenting 14:31 I don't either, Im really on shaky ground here 14:32 but Im pondering that since it seems restarting from "scratch" with new tools is getting easier and easier 14:32 it probably is easier than before, but it's still a big pain it the ass. :-) 14:32 so if I can have a full model of the environment I work with, a conceptual closure somehow, it could be beneficial to invest time doing so 14:32 i really would have used something Cobra-like if it had been out there 14:32 but teaching you a lot too 14:33 probably have a proto-AGI by now. ;-) 14:33 :) 14:33 true, I have to keep in mind my usages 14:33 anyway i think a good exercise is to jot down some example code you might write in your language 14:33 yes 14:34 this is much cheaper than implementing the language and you can creatively play around with some possibilities quickly 14:34 keep in mind that if your language is not highly derivative to existing languags 14:34 then what you have is a basic research project 14:34 which is very time consuming and may or may not pay off 14:34 nothing wrong with that. just be aware 14:34 pretence if powerful 14:36 just playing with your tool before it actually exists, like Palm Pilot prototyping by Hawkins 14:37 ok so Ill dive again in the fundamentals of programming and phylogenies, write examples in whatever syntax I see fit then bug you again then if you don't mind 14:37 sounds good 14:38 (including http://mitpress.mit.edu/sicp/ ) 14:38 http://en.wikipedia.org/wiki/Lucid_(programming_language) 14:39 yes I checked Lucid a bit too, I guess I should start by generating a phylogeny of dataflow programming languages ;) 14:42 the images for phylogeny look like class hierarchies to me. :-D 14:42 http://www.google.com/images?hl=en&q=phylogeny 14:42 but classes even through inherance don't really evolve over time, do they? 14:42 well in those diagrams you add a new branch 14:43 in class-based OO that would be a new subclass 14:43 classes in a code base *do* evolve at the hands of developers 14:43 but not automatically 14:43 unless you are trying out some GP stuff 14:43 but... in the end Im sure you could do all that in OOP or other paradigms, the problem is more how difficult it is to use that metaphor/model (hence the paradigm over language question) 14:43 well yes Im into GP/GA/GG 14:44 no expertize their but I think it's definitly not sth to neglect (hence the interest in bio/evo) 14:44 s/their/there/ 14:44 sth? 14:44 s/sth/something/ 14:46 alright. good luck! 14:46 thanks a lot 14:47 quicknote : do you mind if I post that discussion on my website? most likely nobody will read it but Id rather ask before making sth private public 14:48 sure 14:48 cya later --- Log closed Wed Aug 04 14:48:49 2010