If I'm strongly typing variables then I can understand how my strong typing might cause VisualAPL grief, and interfere with it's normal operations. I can understand how you probably wouldn't want automatic conversions going on that might result in loss of precision.
However if I just assign variables, and let VisualAPL choose it's internal cvar type for me, I would hope it would be able to be internally consistent so that these types of operations would just work.
My own opinion is that if code that I'll call "raw APL" (no input/output to/from devices or the screen, just computation using traditional APL functions and operators) works correctly in APL+WIN, it should also work correctly in VisualAPL.
The only reason for the subtraction above is to try and get numbers into a more reasonable range. The only reason for doing this is because operations weren't working correctly with the original large numbers.
My own opinion is that what I'm trying to do above is not particularly esoteric, but pretty basic and should work.
I will try this out in your latest release and let you know what results I get.
Thanks for the new release.