I've heard great things about your renderer and I must say I'm looking forward to it's release. Any day now, is it not?
If found your questions about z-depth quite relevant and would like to share my opinon:
Format:
I alwas keep to .iff when working with Maya and Shake, for obvious reasons; it's the native format of both programs. It's my understanding though that .tiff is quite popular when compositing in other systems, so an option of directly outputting to this format could be a nice feature.
Value range:
A logarithmic representation limited by the clipping planes is of course the most efficient use of data, but I've found that this, and even simple inverted depth values, confuse a great portion of the compositing community. I think the straight z depth would be most appreciated. This would also make 32 bits more or less obligatory.
Anti-aliasing:
I know of many people who would insist on averaging the values, svearing that it would reduce artifacts. Illogical as it might be, there is a reason for this: Depending on what you'll be using the depth map for (z-comps, dof or z-driven color manipulations) a different requirement is made on the edge values, and although anti-aliasing the depth map is not a correct way to go about it, the softer edges will actually reduce artifacts in some istances. My personal opinion though is to use the value of the object covering the largest area, but only for overlapping objects. When bordering to the background you'd want the depth of the foreground object (rather than nothing) to avoid premultiplication errors.
Hope I could be of some assistance.
Good luck to you!
Andreas Karlsson
Ego Studios
