In the xerox programs, i refers to the dimension across records, j to dimension along (within) 1 record.
  Better:   i is the slow dimension, j is the fast-varying.

Postscript puts the fast dimension along x (increasing left), slow dimension (i) along y (increasing up).

BART puts fast dimension along y, increasing down.
	Plotting directly (no transformation in matrix)
	   gives image rotated 90 degrees ccw - rotate 90 d cw gives true image

MAR puts the fast dimension along Y, increasing up.
	Plotting directly gives mirror image of image: Y increasing to left, X increasing downwards
	   (i.e. we ust switched axes: rotates image about diagonal in plane of paper so inverts.)
	If we plot directly and rotate paper 90 deg ccw then y is increasing up but x is increasing left.
 	   to correct tell Postscript to invert Y (which is realy X), starting at top and comming down.
	   (also have to translate 0 1 to get to top to start [nmax 0 0 -nmax 0 1] )
	   Rotate this 90 d ccw gives true image. 
	   (could have used matrix to correct everything, but this put fast axis svertical on paper which
	    made it print very slowly on apple laserwriter)

Raxis uses Postscript conventions- vladimir took the transformation out of matrix.
	Question about window still put xmax, xmin in i- (should be j), and vv for ymin, ymax.
	He left the rotation before drawing boxxes, circles.
	I took this out, but now i,j don't correspond to x,y
	Switch i,j at translation, now puts circles correctly(?)
	Switch i,j in window question so user can get what he wants.
	Switch i,j in ENCODE statement that prints the window choice.
	Denzo Raxis out files have y and x switched- apparently denzo considers y is fast direction.
	Switch X and Y in drawing circles.
	New Denzo .out file has direct beam position?

It is alright to use odd number of pixels in either direction.  Imstring should be of length 1/2 JN, rounding up 
	if Jn is odd. (This string holds binary data, NOT the hex encoded data).
	Be sure to write an even number of bytes for each line. If odd number of pixels,
	then write one more byte (=J1 to J3, not J2).