[comp.sys.handhelds] symbolic Row Reduced Echelon format program

dwrfielder@dahlia.waterloo.edu (Dave Fielder) (06/25/91)

I originally mailed this letter to Marc E Blair, but have since had no 
response in over 2 weeks.  I am not sure if he is the original programmer
so I thought I'd post my questions to the net.

I have a question as to why infinities occur in row reducing matrices
when they are linearly dependant.
 
 eg. 1: [[ 1 1 a]
        [ 1 1 b]]

	rr

     1: [[ 1 1 a]
         [ 0 0 b-a]]

  is what should occur but...

     1: { { 0 1 'a*oo-b*00+2*a-b'  }
          { 0 0 '-(a*oo^2)+b*oo^2' }}

  is what I get.

I'm unsure as to why the infinities are necessary or how come they occur.
Is there a workaround?  If so, what is it?

I love the fractions flag and think it's great.  To give you 
some feedback, some of the fractions are a little off due to
rounding, ie. 100000000001/300000000002 instead of 1/3. Could one 
of the various toQ programs help eliminate alot of the fraction errors?

Thanks,
--Dave Hubert.