The screenshot doesn't show Stunt Mania, as the more observant amongst you will have noticed. It instead shows my little 'car physics test bed' program. which I am using to get a decent and well implemented mathematical model of a car, before I go assaulting the main code. The trick is obviously getting the right compromise between flexibility and complexity, especially if we want to include models as diverse as the various suggestions currently adorning the guest book!

For reference, the test app 'correctly' (i.e. as far as designed) models the following behaviour at present : understeer, oversteer, weight transfer, slip. I need to do a little bit of fixing on such minor matters as steering (the requirement for the front wheels to point in the same direction is currently a soft constraint, whereas a hard one is requierd), brakes (there are none) and gears (at present exactly one gear with a very unrealistic torque output is simulated). I may upload this test once it is complete.

In other news, an updated file format for objects which includes a whole wealth of useful collision information and a 3ds importer to write to this new format are both approaching maturity. Also, only 18 of the 56 Sports 4d Driving objects remain unstarted, although arguably as many as 40 remain unfinished.

Other changes are purely that I have now written a (non-perfect) Stunt Mania to 3ds exporter, and an importer is on the way. Which is good because other members of the team are starting to come up with models. Some 3ds files have been added to the Downloads page, but I'm fairly sure translucency information is not properly exported in it. I seem to be messing up 3ds percentage chunks. If anyone can help . . . .

On the topic of file formats, it is looking quite a bit as if Sports Car GT type cars will either be the native format, or will be supported for import. See Speedsims.com for a massive collection of these + some editors.

Unfortunately development has slowed due to a problem with my machine. As the screenshot implies, I have installed an NVidia Riva TNT2, which was meant to speed up debugging massively since it supports rendering in a window, where my 3dfx does not. But in actual fact it has slowed the whole ordeal, because it simply refuses to work with Windows 95 - the OS just resets itself on startup. Luckily I have Windows 2000 installed on this machine also, and it works perfectly in that - I'm just greatly hindered by the fact that my machine is hardly Windows 2000 material to say the least. Even immediately on startup without running a program, the OS has allocated ~64mb of memory, and I only have 48mb of RAM! In fact, my machine only supports a maximum of 64mb of RAM (which I have ordered), so things will never be great.

Don't worry though, the machine is not unusably slow, but I hope I can come up with a solution so that I am not stuck with Windows 2000 until the year 2005 or whenever I can next afford a new machine!

Anyway, enough of the dull personal stuff. I hope to have a new demo available within a month, which will be much the same as the last one, but with some more realistic physics (you'll feel as though you are manouvering a rectangle, correctly rotating and feeling gravity, rather than the always upright sphere without proper gravity of the last release) and possibly some more track shapes if I get the 3ds to Stunt Mania importer working in time.

After the surprisingly high quantity of publicity Stunt Mania has attracted - for the record I wasn't expecting anything like this response, especially not for it to end up on opengl.org - I'm starting to put together something of a team! Obviously it wouldn't be fair to go mentioning names while most involvements are still speculative, but people have expressed an interest in potentially helping with the following two key areas :

• artwork
• artificial intelligence

These areas being key not only because they can make or break a game, but also because I am not any good in them. Also, Filippo DN has pointed me toward the GPL library, ODE, which means that working on the physics has shifted completely from me coming up with code for maths I more or less understand but which there is a lot of to me learning an entirely new library and interfacing it with what I want to do. I don't think the time frames will differ that much (we're still talking a month or two at least before I get anything I'm happy with), but the result should be a much faster and more stable implementation since the author of it is a professional in the field and I am not.

Also, as a result of both the arrival or artists and also an unrelated offer, support for importing your favourite object formats (3ds, blender, lightwave, those sorts of things) into Stunt Mania should eventually be forthcoming.

Finally, I have discovered there is such a game as Hard Drivin' 2, distinct from Race Drivin', which features a track editor, and I have loosely started trying to reverse engineer the file format it uses. The editor is substantially more complicated (and hence flexible) than that offered by Sports 4d Driving, so reverse engineering may prove too much for me, but should I succeed the benefits will be very large - Stunt Mania will be able to offer the classic Hard Drivin' track!

• The compiled vertex array extension does not work on many older cards, including most 3dfx's, but works on all recent ATIs and NVidias with up to date drivers. However, it decreases the frame rate on those cards.
• On some GeForce cards, the ground texture appears to jump a little when rotating.
• On many cards, the floor can appear to fight with the road or water for prominence

None of which should be too difficult to deal with. It was also pointed out that things would look nicer if I added some lighting and included some higher resolution textures for people with cards that can take them. I'll look into both of those. Particularly, I don't actually have any higher resolution textures!

New code written since the 06/07/2002 release is minimal, but I have finally bothered to package my matrix/vector operations into classes, and have written some quaternion code as part of the preliminaries for rigid body dynamics. The screenshot shows a quick test of the quaternion based rotation.

Also there are some new game objects & some fixes to some of the existing ones, mostly just related to the bridges & ramps.

I am today back at my development machine. I have made numerous small changes, including getting together some much nicer texture maps (a couple provided by 3DCAFE.COM), and a small programmatic change makes it very obvious when your graphics card can dispose of the texture's alpha channel and dedicate more storage to colour information - providing a better display for many people with slightly older hardware. Although the texture data file has swolen in size by a factor of about 10.

I have also renovated slightly the method used to convert Sports 4d Driving legacy track files into 3d data, reducing the total polygon cost. Which, combined with a smarter, hierarchical approach to object level clipping, means the prospect of even us 3dfx voodoo 1 owners being able to play without horizon clipping is very much back.

Collision detection isn't finished yet, but I am tantalisingly close! Expect the next demo very soon.

Additionally the engine now uses compiled vertex arrays where possible. In total the whole game gets a large frame rate boost, although I don't suspect anyone with much above very low end hardware (like my 3dfx voodoo 1) will notice the difference.

I've created a few further game objects, mostly to do with the raised track sections, but I'm getting a bit bored of graphics related stuff, so have started on the collision detection and what will eventually be the Sports 4d Driving style car physics model. My structure of choice is a quadtree, should you wonder. I am hoping to have scenery interaction nearly cracked in the next demo, coming soon-ish. It'll have to wait a short while though, since I am now away from my development machine for a short period.

The more technical observer might be interested to learn that I have discovered glPolygonOffset (which, for some reason, is not documented within the Microsoft VC 5 help files), which greatly eases file format & display considerations. It means I can draw objects which are coplanar with the grass without them either appearing incorrectly or requiring their own grass polygons.

Mip maps are currently generated by means of box sampling down the original texture in RGBA space as a pre-processed effect - which can throw up some odd effects. Particularly notice how the stripe in the middle of the road appears to get fatter as it gets further from the camera.

The screenshot is taken with all texture filtering turned on, so is near exactly how your 3d accelerator would present things.

As can be seen, the corner is not properly surrounded by grass. This is partly because I am becoming a little irritated with the limitiations of procedurally establishing track pieces in the main program. All of that will now be suspended whilst I establish a set of file formats for various things, and then start using those.

Also visible is that I have used a texture with American style road markings. This is an attempt to match the American stylings of the Sports 4d object set, which I feel I need to mention since I have never been inside American, and could do with some help regarding the markings expected at different road junctions.

Although texture filtering is switched off in this, the debug version, it is on for release builds. Also the current textures suffer a little due to having come from a paletted source, but this is not a necessary restriction.

Also in the engine are the preliminaries of orientation (a special orthogonal matrix, since you ask), location & communications between parts of the collision detection code.