accumulating dsps vs transforming

[Eric Dean Tribble <tribble>]

This is about the decision to accumulate dsp as we walk down the ent
vs. transforming crums.  For example, the combine operation walks down
one orgl (A) till it gets to the height of the other orgl (B).  As it
moves downward, it can either accumulate a dsp from the top of orgl A
or transform orgl B at each level of orgl A. I defaulted to the latter
method.  I finally found a practical reason to choose between the
two.  When I transform a crum, the new crum doesn't have all the
backpointers:  canopies, htrees, trace coordinates, etc.  This doesn't
break anything yet, but I will be saved a debugging hazard ( and thus
lots of time and pain) by accumulating dsp.  Since the question has
come up on occasion, I thought I'd let you all know the resolution.

dean