jME-Physics Website

Okay, now that I'm a developer can I say that the jME-Physics site needs SERIOUS help without offending anyone?  Please let me know if that would offend someone so I don't actually say it. :-p



If someone isn't currently maintaining it I am willing to put a couple hours into it to at least make it useful.  As it stands it's just pretty much a pile of uselessness.  I understand it was kinda thrown together, but jME-Physics is getting pretty powerful and the site doesn't do well to convey that jME-Physics even actually functions. :o  I mean, a WebStart demo would be a good place to start!



darkfrog

jME Physics is very powerful. Perhaps we should join together? I know that originally, we decided to seperate the two to keep focus, but now that it's in the state it is, it might be time to merge physics into the main baseline. Just throwing out ideas here, I know DP's originally wanted it in the core.

vote for core

Hmmm, I agree to an extent.



What do you think about making it part of CVS but have it as a separate module jar so people don't have to add it to their projects if they don't want to and don't have to worry about have ODE DLLs.



Also, when we get around to doing that I think we really need to rework the whole API to make it engine neutral.  This way we can eventually utilize something other than ODE if we want and just provide a default ODE implementation.



darkfrog

Firstly jME physics needs to work with the current release, before worrying about redesigning it:-)



[edit: Thought I better add I'd do it myself but I'm a bit stretched working on some other tech for my prototype(s) (physics is used only in some of them) and uprooting for a move to Tokyo]



That said, it might be worth looking at reworking it around the Open Dynamics Framework ( http://www.physicstools.org/) , like Novedex did for their AGEIA PhysX SDK



http://www.ageia.com/novodex_downloads.html





DP was looking at Novodex integration, not sure how far he got.



Physics API is a real force mulitplier for making simple games quite dynamic. I think it's critical that jME keep it in synch.


Well, if it was to go into cvs, it would certainly be com.jmex.physics.* (the same structure it maintains in the sourceforge repository) and as such be built as a seperate jme-physics.jar. Just like all the other modules in jME.



I haven't really discussed this with Renanse or Per, so don't put any faith this happening.



Removing ODE dependencies from jME-Physics should probably happen whether or not it moves repositories, wouldn't you say?

+1 for bringing it in.

First I adapted jmephysics to work with jme 0.9 again.



As for the jmephysics redesign: yes, it's definately necessary, if it's moving into the core or not! Especially the implementation is quite messed up… the api is ok but not good.



I think it's ok, that jmephysics is in another repository than jme. This ensures that jme is not dependent on jmephysics in any way. So if jmephysics moves to the jme repository it should reside in an extra source directory and should always be configured as a separate module/project in the IDEs, to avoid introducing unwanted dependencies.



But we should consider to release it with jme!

This would require that someone that is changing the jme-api has jmephysics in his IDE, too, and changes and tests it with jme. For me this would be ok, obviously, as I'm maintaining jmephysics either (as much as rl lets me  ).

Irrisor, was it broken again?  I fixed it (at least thought I had) a few days ago and checked it into CVS…



Didn't realize I was striking up such a discussion with this thread, but I guess it was needed. :slight_smile:



I think we need to discuss a standard abstract API completely separate from any physics framework and after we've got a solid solution we can go about creating an implementation with OdeJava for it.  I think that is the next step.  In the next few weeks if I have time I will attempt to do so.  In fact, it may be a good idea to provide this abstract API as part of the base jME project and just have a separation only for the implementations of it.



Thoughts?



darkfrog

darkfrog said:

Irrisor, was it broken again?  I fixed it (at least thought I had) a few days ago and checked it into CVS...

Yes, it was broken due to the changed boundings - boundings are no Geometries any more. They are even no Spatials which made handling of them in jmephysics quite different. Additionally compounded physics objects were broken. Your fix is only a week ago, but there were more than 60 checkins to jme since that  :)

Great! Thanks for the update.

What are you guys doing breakin' jME-Physics for? :-p



So you've fixed it and checked in the changes?  Just trying to make sure there aren't any known issues at this point.



darkfrog

Okey, entering this discussion a little late…



I vote for putting jme-physics in the standard jme distribution, just like any other of the com.jmex.* packages. The core shouldn't be dependent on any of them, so I don't see that as a "risk".



And yes, we need to work heavily on the design, making physics engines work as plugins. I think I'm gonna have the time to do it (together with darkfrog and irrisor) sometime soon!

three tumbs up on that !

this should not only take rid of the annoying double checkout of libs through cvs but also help keep physics in sync with the rest.

ok, ok it wasn't that bad actually :slight_smile:

Pretty please add it to the core. I cannot access CVS on sourceforge for the life of me.

Perhaps we should try to get a nightly build put up for jME-Physics at least in the short-run so people are taken care of.  The last "release" of jME-Physics seems to have been March of this year.



Thoughts guys?



darkfrog

Yes, both, a nightly build and a release that works with jme.9 would be great. Although Per and me did not find the time to do it…

Is there an automated process that pushes up the nightly build of jME?  If so we could just add another ant file to checkout and build for jME-Physics as well.  If not I'm happy to put something together. :-p



darkfrog

Already exists.  See the very first link in the side menu of www.jmonkeyengine.com

I know it exists…what I'm asking is if it's an automated process or a manual process?



darkfrog