DSP Robotics Support

Just figured I would ask to see if my thinking on this was sound.
Since stereo signals can have differences in peaks between the left and right, is it customary for compressors to take the maximum value between both signals as the input level? Another option would be to process each signal separately, which I think some compressors out there may do just to be fancy, but it's probably not very useful. And you could take the average of the two signals, but I don't think that would go well.

Just curious what the norms are I guess. I've used plenty of compressors but never made one before.

compressors may do it all... some use sum of the channels, some use maximum and some compress separately.
Sum is most typical for bus compressors to glue stuff together - it prevents the compressor from messing the stereo image, which may happen during high GR in dual mono mode (L-R separate processing). Example of such compressor is the Density mk3 from variety of sound if I recall the manual correctly.
Maximum of LR peak values is mainly used by brickwall limiters - it prevents the limiter form messing up stereo image, yet passes no peaks through. I also hear compressors to use average between LR envelopes.

Ah I see, thanks!
I think summing the channels and taking the average are the same thing (average is just sum scaled).
Still not sure whether to go with the sum/average or max value. This is for color/dynamics shaping, especially for drums. I'm kind of leaning towards max.