Slashdot Mirror
Life After MS-DOS: FreeDOS Keeps On Kicking
angry tapir writes "FreeDOS — the drop-in, open source replacement for MS-DOS — was started after Microsoft announced that starting from Windows 95, DOS would play a background role at best for users. Almost two decades later, FreeDOS has survived and, as its creator explains in this interview, is still being actively developed, despite achieving its initial aim of an MS-DOS compatible OS, which quite frankly is somewhat amazing."
You can have both!
Install FreeDos in the c:\dos folder of your DosBox machine. You'll get most of FreeDos' new functionality, while keeping the useful features of DosBox.
see here: http://www.dosbox.com/wiki/TOOLS:FreeDOS
As a developer, I can say no, that "problem" was just a dumb way to do physics, and it's been fixed forever to anyone who wasn't a moron, even in the AT/XT days. Back in the day we just checked wall clock time / CMOS ticks, you know, the ones we used to modulate PC speakers to make different frequencies, that's what we used to update game state and prevent binding physics to CPU speed. Today, RAM is plentiful, so I do physics state "commits" in fixed step sizes, say 30hz or 60hz, and interpolate from the last physics state to the temporary state that's being rendered.
If enough time has passed to process the next full physics step, then it's processed and committed. Otherwise I reset state to last commit, and process the partial physics step, but do not trigger any important events like player death in the temp step. If the system is too slow to run a partial physics step without immediately requiring another full physics step then the partial steps are not processed and the game rendering updates screen positions only after the physics step can be computed. This is important for deterministic physics, for demo replay and also for synchronized network games because:
UpdatePhysics( 20ms ); UpdatePhysics( 20ms );
is not always equivalent to:
UpdatePhysics( 40ms );
Due to a number of factors, especially rounding errors.
UpdatePhysics( elapsedTime );
Is the worst on fast systems -- those very small fractions of time lead to physics explosions -- not the particle effect kind, the bounce off an object through the floor and to the other side of the universe kind.
For comparison:
Here is my rope physics with a fixed physics step size frames.
Here is the same physics running with actual elapsed time each frame.
The latter comments mention tiny jiggles propagating into a frenzy. Those tiny movements coupled with very small elapsed times create the physics explosion.