Suggested broad cmp algorithm:

gistfile1.txt

infix:<cmp>($a, $b) {
     if $a ~~ Numeric && $b ~~ Numeric { return $a <=> $b }
     if $a ~~ Stringy && $b ~~ Stringy { return $a leg $b }
     if $a ~~ Positional && $b ~~ Positional { return [||] (@$a Zcmp @$b) }
     if $a ~~ Hash && $b ~~ Hash { maybe?  dunno how to do this }
     $a.WHAT leg $b.WHAT; # except it has to handle the two different classes with the same name case
}

Take two, with TimToady and benabik input:
infix:<cmp>($a, $b) {
    if $a ~~ cmp(positionalish) || $b ~~ cmp(positionalish) { return || }
    if $a ~~ cmp(Real) && $b ~~ cmp(Real) { return $a <=> $b }
    if $a ~~ cmp(Unicode) && $b ~~ cmp(Unicode) { return $a leg $b }
    if $a ~~ cmp(hashish) && $b ~~ cmp(hashish) { maybe? dunno how to do this }
    $a.WHAT leg $b.WHAT; # except it has to handle the two different classes with the same name case
}