Daniel Russel wrote: > - The first one, which is just semantic, is what is the "IMP Kernel"?
At the C++ level, it is whatever is linked into libimp.so. I am not fond of configure-time optional extras going into there, since it makes testing and support that much harder.
> - The second is, what should be built into the IMP package (i.e. > installed by a single install command). My answer to that is that it > should include everything that is likely to be supported as long as > IMP is supported. So something that is specific to a particular > project and will probably not be picked up by anyone else when the > implementer leaves should not be.
Indeed. All of our supported IMP modules would go into the IMP package. But we would have other IMP modules that are not supported, still in development, or are for our own use, of course, and they would be distributed separately or not at all.
Ben