Many Thanks!

In D4471#102360, @Brecht Van Lommel (brecht) wrote:

I don't think this approach is correct. It tries to preserve total volume of the mesh, but that doesn't make a lot of sense for more complex shapes or meshes that have multiple disconnected components.

Well, it is not always guaranteed that the procedure will work in every case. One has to try whether one likes the result or not. If not, one should switch it off. It's the same with many things in Blender.

For example if you have a character mesh, if one arm collides with something it would cause the entire mesh to expand?

The example with the arm shows the problem of softbodies in general: If the change of the arm volume is big, the simulation is bad in every case, even without "Preserve Volume".

A solution could be to make the arm to an separate object and to connect it to the body on the shoulder. The same could be done with other parts of the body if one want to simulate a human body.

I could imagine to extend the feature to vertex groups in the future. What do you think?

There is a valid point to say this is not the right way to do it for softbodies. Cloth simulation already has air pressure, why add air pressure to softbodies as well? Softbodies are solid chunks of elastic or deformable material, not air tanks as it might seems currently. Blender really has very bad softbody code so using air to fix the volume problem doesn't really cut it. In my opinion blenders softbodies are made redundant by cloth simulations currently and would need a proper re-implementation with a finite element solver. The only thing softbodies have, that cloth doesn't have currently is plasticity, but that may very well come to cloths simulations in the future (for aluminum foil simulations for example)