Hi Bill,
I'm sorry for the delay in responding. Annette did ask me to answer last week, but I'm afraid I got distracted. Okay, we're going to go into some history here. The spatial reference contains several items:
coordinate system
xy/z/m domains
xy/z/m resolutions (earlier called xyunits which was 1/xy_resolution)
xy/z/m tolerances
The latter three were originally defined because geodatabase feature classes (originally in ArcSDE) stored coordinate values as positive integers. In fact, they were originally had strict limits and were updated to "high precision" several years later. There was a maximum number of allowed integer values. A combination of the minimum x, minimum y, and EITHER the max x, max y OR the xyunits or xy resolution value controls the third thing. so if you set the min x/y and the xy resolution, the max integer values force how large the max x/y values can be. Or if you set the min/max x/y values, then the xy resolution (xyunits) value is then calculated. When we went to "high precision", that increased the maximum integer values by...a lot.
The default minimum x/y values are set based on what's allowed mathematically for a projected coordinate system rather than the area of use for that particular pcs. At the time, for a custom PCS, we would have no idea what the area of use would be so better to use the allowed mathematical limits of the map projection.
Resolution (1/xyunits) values determine the precision of stored coordinates. Our defaults for x/y are 1/10 mm or 0.0001 m. If you set xy domain to match the area of use, that usually forces the xy resolution to be really, really small which is unrealistic and can make handling and processing the data difficult. Instead, we usually set the minimum xy values and the resolution which instead makes the maximum xy values very, very large, much more than the area of use of the mathematical limits of the map projection.
Min values for z and m are arbitrarily set to -100000, similar resolution values. For GCS (degrees usually), we use corresponding values to the defaults of a pcs, but the minimum xy are set to -400 and resolution is around 10e-9, I believe.
Tolerance values have to be at least 2x the corresponding resolution value but our defaults set them to 10x, so 1 mm or its equivalent.
In case I didn't explicitly define these earlier, an area of use is the where a PCS should used, like California State Plane zone VI covers San Diego, Imperial, Riverside, and Orange counties, while the xy extent is the mathematically valid area of the map projection. California VI uses Lambert conformal conic so it's most of the world.
There is at least one white paper that discusses some of this. Here it is: Understanding Coordinate Management in the Geodatabase.
I hope this helps.
Melita
