PhysX Read/Write capability in libPlasma

[Trylon]

Unfortunately true...

When did that ever stop us though? The first PyPRP worked just like that.

[Zrax]

libPlasma *tries* to be a little less hacky... But with Plasma, that's sometimes hard

Regarding PhysX, I would really rather keep it out of libPlasma. I'm trying to keep libPlasma as open and available as possible, and adding PhysX into it puts a pretty severe blocker on that. More likely, for the PhysX baked data that must somehow eventually make it into the PRPs, I'm thinking a plug-in can be provided for apps that can make use of such a creature (like PrpShop or PyPRP 2), and that plug-in can be downloaded and licensed completely separately from libPlasma and company. This was sort of the direction I've had in mind for a while now, but nobody's ever bugged me about it until now, so I just kept not dealing with it

[Trylon]

Another idea of mine was just using an external executable file with a defined input and output (probably temp files) as an optional addin for either pyprp and/or libplasma. I think that would satisfy the whole GPL license stuff as well, since no actual linking is done. (And it would save some time hacking, since one can just use the PhysX SDK like Robert has done)

[Robert The Rebuilder]

Once you all decide which way to go - be it DLL/DSO or socket connection to external process - and have an interface designed, I'll retool the PhysX library code accordingly.

For all you lawyers out there, here's a question. If you download NxOgre's latest version (NxOgre.1.5.5.BloodyMess.zip) and look at the source code, you'll see it is sprinkled with PhysX SDK calls. So, why is Robin Southern (aka 'betajaen') able to do this, whereas Cyan has to rip out said calls from their code before going open source? Is it because MOUL is a licensed application (which cannot go open source due to section 5.f of the PhysX SDK EULA) while NxOgre is just a library? Or is it because NVIDIA hasn't gotten around to sending their C&D yet?

[Tsar Hoikas]

Once you all decide which way to go - be it DLL/DSO or socket connection to external process - and have an interface designed, I'll retool the PhysX library code accordingly.

For all you lawyers out there, here's a question. If you download NxOgre's latest version (NxOgre.1.5.5.BloodyMess.zip) and look at the source code, you'll see it is sprinkled with PhysX SDK calls. So, why is Robin Southern (aka 'betajaen') able to do this, whereas Cyan has to rip out said calls from their code before going open source? Is it because MOUL is a licensed application (which cannot go open source due to section 5.f of the PhysX SDK EULA) while NxOgre is just a library? Or is it because NVIDIA hasn't gotten around to sending their C&D yet?

I don't know where all of you are getting the idea that the PhysX code is the problem for MOOSE. Section 5f in no way prohibits the open sourcing of MOUL. It just prohibits viral licensing (GPL), which is not present in the LGPL.

[ddb174]

We don't yet know what license open-source Plasma will have if it happens. And libPlasma is GPL, isn't it? (As a side note, LGPL is still viral, but I know what you mean: it does not apply to linked libraries, and so does not pose a problem in this particular case.) And you probably don't *want* libPlasma to be LGPL either ;)

[Zrax]

And you probably don't *want* libPlasma to be LGPL either

That's correct

[Tsar Hoikas]

(As a side note, LGPL is still viral, but I know what you mean: it does not apply to linked libraries, and so does not pose a problem in this particular case.)

Oops! I'm glad you can understand my garbled posts

[Branan]

A shared library that gets loaded dynamically is not enough to satisfy the GPL, unless we only distribute source and let the user build it themselves (like how proprietary driver vendors send out the glue code for the Linux kernel, but not a full binary).

At this point, I think we need to just flesh out a socket and/or intermediate file spec, so Robert can finish his code. The other option, of course, is to reverse-engineer the format and build our own exporter, but that's a lot of work for likely little gain.

We would still be able to distribute Robert's tool in the same source and binary packages as the rest of things, it'll just have to have a more liberal license than GPL (or be GPL with an exception to link against PhysX).