--- Thomas Schmidt <thomas.schmidt@xxxxxxxxxx> wrote: > > Roland, > > It seems you actually want to setup a new parametric modeller. The > brave = > ones rule the world. :-) That really hits the nail on the head - I am glad you said the "brave ones" allthough I am sure you thought of 's/brave/crazy/' here. Anyway, such a modelling kernel would be some really interesting project, especially if it is backed up by a database as was discussed earlier. A CAD system that allows a user to specify dimensions by parameters is probably built upon some form of a constraint solver. I assume that the term "parametric" is used because what the user actually notices and sees is, that an edge can have a length that can be changed at any time and the endpoints will automagically move. Parametric is a term that refers to the application of the solver not it's implementation. The first and most succesful system that implemented this feature was certainly Pro-E, they even named the company Parametric. Roland > I assume that you use the term "parametric" not to refer to the > parametr= > ic functions of curves and surfaces, > but to a system, where objects interact by constraints. I have ever > wonde= > red why the term "parametric" > is used to describe a constaint solving system. Does anybody know? > Anyway, you might want to have a look here : > http://www.cs.washington.edu= > /research/constraints/ > The constaint solver they describe, are mostly based on local > propagation= > algorithms > and therefor not very well suited for (complex) 3D modelling, but you > get= > a good introduction into the world of > constaints and their solving algorithms. There is even some code for > a fe= > w solver. > I also found many research papers about constaint solving algorithm > on ht= > tp://citeseer.nj.nec.com/cs. > Search for constaints or solver. Mostly academic papers, but > interesting = > anyway. > __________________________________ Do you Yahoo!? The New Yahoo! Search - Faster. Easier. Bingo. http://search.yahoo.com