Data type implementation of gluLookat()

I am wondering if gluLookat() implemented using float values for the eye point and view point. According to the VC++ docs it’s parameter list consists of double values.
I am using very large numbers from the Earth Centered Earth Fixed frame (on the order of +/- 6,000,000) for my eye point and view point. I see a very disturbing jitter between frames with these numbers. If I translate the entire system to some arbitrary system centered around (0,0,0) prior to any OpenGL calls the jitter goes away.
It the Win32 version of gluLookat() is implemented in terms of float values instead of doubles then this might explain the problem.
Any ideas??