From: Brian Thompson <briandthompson@ho...>  20001030 23:35:14

> > Hi all, > > I started yesterday to code the Map converter, when I realized that it > isn't going to be easy to run a 3D client using the actual structure. > > I am downloading Quake source to take a look at how it store the data > and how the map is done. > > I really feel that the actual terrain definition takes to much storage ( > and network ) space and it is not the best for 2D neither 3D. > > Any idea for a good 3D game to mirror the data structure? > Sorry I'm not a coder, although I have done some programming speed was not a concern the way it is for us. I do have several ideas about doing the 2d/3d problem but I confess they are a simplification and when developed may not be as fast as we need. First what about generating the landscape using triangles the rows alternate and are shifted be 1/2 each edge between points would be 1 meter like this 11 12 13 14 15 16 17 18 19 21 22 23 24 25 26 27 28 so that from 11 to 12 to 21 back to 11 is triangle each of these points has a height associated with them so these establish the normal by performing a cross multiplication of two edge vectors you find the normal vector to the surface. from above this would be a very simple system and the top of every surface is visible so the only things we'd need to worry about are caves and buildings and the seperate system for placing them and turn the roof invisible when someone walked into a building. I will ask what about crystal space they are open source and we may be able to convert part of they're code to do the 2d part of our system, in fct them may be willing to help with this. Another thing I've been thinking about is the landscape definition. I think that vistapro uses fractals to calculate its landscapes and the slopes and all, but you can change certain variables to make the landscpe look like you want. Why couldn't we use this idea define our landscape, the only problem with this is the calculations involved is would probably slow down the process but iof youre doing the calcualtions for a 1024x768 screen where the height mapping is done for 12x16 matrix of points the only concern would be smoothing, and storing the points for the whole world once its created. What size of files would this require? could we use the equation on a running basis or would it be just way to slow? If we shifted the landscape and only had to recreate the portion of the landscape that is newly revealed this might be a possibility how complex and taxing would the smoothing on a 1024x768 screen be if we used a 12x16 point matrix as a basis, and would a fractal be faster? Some thoughts, I'm not sure I got them all across but I hope someone can understand enough not to just think I'm stupid. Brian Thompson 