Rushmore "lives" in the VFP core, which is also a part of the runtime files. But it is not really "rocket science", it is only a set of algorithms which evaluate all available indexes and choose which, if any, indexes will speed up the search, and how to use these indexes optimally. I don't think it would be that difficult to duplicate this functionality? Funny you mention that - back around 1992-1993, Sequitur Software published a C/C++ library called CodeBase. It provided a number of xBase-style functions that you could incorporate into applications. They had a release that supported the CDX, and provided a 'bit-optimization' capability that was pretty close to what VFP had in Rushmore. I never did exact benchmarks between the two, but from what I remember, the speeds were pretty close.
Ah, the good old days. ;)