Well, looking at some tutorials (like this one : http://www.adamdawes.com/programming/tokamak/02_Cubes.html) it seems that it doesn’t do the time-based thing alone, but you can specify the interval values between two iterations (code is at the near-end of the page listed above). Is this sufficient ?
Chman
it does solve the time based thing:
fElapsed = GetElapsedTime();
.
.
.
gSim->Advance(fElapsed);
:D
Then tokama seems like a really good idea
Will it be possible to make it run under other platforms than Windows?
It’s Windows only.
I was browsing the forum, and from that it looks like support from the developers is not very active. The current version took 7 months to get released, and bug reports on their forum are not being answered (e.g. http://www.tokamakphysics.com/phpbb/viewtopic.php?t=513 )
The best alternative would be an opensource physics engine in Java of course, but that’d require some serious effort. 
Right, the SWIG things need many changes… All types constructors are protected and should be public, etc… I’ve changed those things, but I think nothing can equals a good binding written by hands 
Chman
To be honest. I still prefer it if we created a binding to Novodex rather than Tokamak because Tokamak isn’t as heavily updated as it should be. However, what they do have in there is still better than ODE anyways (feature wise). But due to the lack of a linux and a macosx build, i am hesistant (Tokamak that is).
Just one example of Tokamak being ALOT better than ODEjava is that it uses 970Kb of ram with 30 Ragdolls whilst ODE uses 10Mb initialising! And Tokamak is faster too and is frame rate independant.
We have a working ODE-jME binding. But I dont want to go down a route and then realise that we should have gone down a different one.
DP
Have you tried the Newton physic engine ?
Seems pretty good for games… It has cool features and is frame rate independant. Some demos here…
Don’t know about the license though…
Chman
I can’t imagine it’d be free - then it would be prompted at the front page 
Well, I’ve downloaded the sdk, they said “a free physic engine for developpers”, but then you just find a copyright note… Nothing on license for commercial use…
I’m gonna mail them to get an answer
Chman
Keep in mind that sum of us hope to make a game good enough to sell one day.
Badmi is right, it’s why I don’t think using the novodex engine is good, it will only be usefull to make free physic demos…
About the newton dynamics engine, here’s what I got by mail :
The SDK is free for commercial use. You just need to give us proper credit.
So I think we could take a look at this engine and see if it's good for jme... If not, doesn't matter, we'll find another engine ;)
Chman
but do we all agree that ODE is not the solution we are looking for?
Yes, I agree. ODE is good to make some demos, but that’s all. It’s buggy and unstable, and miss a lot of cool features, so…
I’m voting for tokamak (but it’s windows only) and Newton dynamics (a linux and mac ports are on the way)… Maybe someone knows others physic engines ?
I just know about novodex, havok (pretty cool but so expensive), tokamak, newton and Ode…
The last solution is to make our own physic engine in pure java… But this could take monthes…
Chman
From looking at the poll, it looks like the majority wants Newton.
I suggest we do something like this: Me and DP consolidates the odejava binding now (it’s like max 1 hour of work) - commenting everything and doing some cleanup. Then release it as 0.2.
Then we can all concentrate on the Newton binding.
I just thought about something…
Jme display is abstract so that we can use JoGL or LWJGL or anything we want to draw to screen. Why not making the same thing with Jme ? Making the physic part abstract, the user could have the choice (the default binding developped by the team would be the one you choose).
Chman
Yeah, I think so too (and proposed it earlier in this thread).
Sure, il put that framework in. However, what i wont do is maintain both ports. Il be glad to maintain one port and have the framework to do another (like the JOGL).
So which one is it? Newton or ODE. From the poll, it looks like Newton wins and to be honest, its much more work and better results and less bugs.
We’l lay down the framework. But its up to the ODE enthusiast to create the port.
Thats how i feel about it, DP
From the poll, it looks like Newton wins and to be honest, its much more work and better results and less bugs.
Careful... while Newton might look great from the outside, no one has started using it extensively to make a real decision yet. You might run into just as many deal breaking problems as ODE is giving you. At least with ODE you have real tangible results, and Per has a damn fine game shaping up. By all means play with Newton and the rest, but I'd reserve committing to it until you are sure it does what you want... if it were me.
Me and DP have decided to continue with the ODE binding. Expect v0.2 by this evening!