radcan and algebraic flag



I wonder why the algebraic flag is false by default in radcan. Is this
an accident of history (runs too slow on a 0.5 MIPS, 5 Mb PDP-10, used
to be experimental, radcan written before algebraic)?  Or is there
some substantive reason?

I always (naively?) think of radcan (often followed by factor) as the
"universal solvent" for algebraic expressions.  So when I start with
something simple like

             ex:  1/(sqrt(x^2+1)+x)+sqrt(x^2+1)+x

and can't get a simpler form with radcan and factor, I'm not happy.
But it turns out that radcan(ex),algebraic => 2*sqrt(x^2+1), which
makes me much happier.

The main argument *against* having algebraic:true during radcan is
that there would then be no convenient way to have it be *false* if
necessary.

I'm tempted to suggest that algebraic *always* be on by default, but
perhaps users would be unnerved by ratsimp(1/sqrt(1-x^2)) =>
-sqrt(1-x^2)/(x^2-1).

                  -s

PS Peculiarly, ratsimp(1/sqrt(1+x^2)),algebraic does NOT change the
form: bug, I think.