On 05 Mar 2010, at 23:04 , Andrey Fedorov wrote: > But a good map-maker includes the "important stuff" and discard the > "details". When "complexity" or more specifically "the possible states of > each component" are part of the "important stuff", CRT is not a good > representation of a system.
Science goes through phases. Initially you are collecting observations. There are blue boxes, red circles and all kinds of shapes in all kinds of colors. After a time a pattern starts to emerge. Whenever you see a green triangle you also see a blue square and a black icosahedron. So then you haul out the first version of your CRT and you make a public announcement of your theory: IF there is a green triangle AND a blue square THEN we can expect to see a black icosahedron. But then Andrey Fedorov comes and pisses all over your bagels and complains that yesterday he was looking in his ORDBMS and he constructed a query that showed not one, not two but fifteen _hundred_ instances where the state of the system has a green triangle and a blue square but no black icosahedron! At first you despair but then Alejandro comes along and he is thinking systemically so he's looking for invariants so what he does is he re-runs Andrey's query w/ a timestamp and - lo and behold ! He makes the huge realization that whenever T_month modulo 9 is equal to 0 the green triangles are having wild monkey fornication with the blue squares and black icosahedrons are the offspring! Major paradigm shift! Society is never the same again! Alejandro gets publicly crucified! Great honor! So yeah, no one is suggesting that you use CRT for capturing the state of your systems at discrete intervals. The purpose of CRT is to describe your mental model of what you think is behind all the observations you are making. A lot of the fighting programmers do with each other is around the argument whether 'tis nobler to suffer the arrows [1] of general interfaces to computation or to take arms against seas of unmanageable state and through normalization unto the third normal form end them. The reality is that what you're going to do depends largely on what your program is modeling. Sometimes you're in a domain (Large Hadron Collider anyone?) where everyone's so utterly and completely clueless that you're going to have to spend your entire career going painstakingly through the state of your system for every possible starting energy of e at every time interval t. Other times you're engaged in a program like STEPS and you're thinking it's largely overdue for someone to do the difficult, dangerous, underpaid and totally under-appreciated work of coming up with an (executable) general theory of computing. - antoine [1] http://www.haskell.org/arrows/ _______________________________________________ fonc mailing list [email protected] http://vpri.org/mailman/listinfo/fonc
