Here is an explicit solution w/o using a module that uses a recursive procedure to calculate the next combination:

use strict;
use warnings;

sub first {
  ( (1)x$_[1], (0)x($_[0]-$_[1]) );
}

sub nextstep {
  my @p = @_;
  my $i = $#p;
  $i-- while $i>=0 and !$p[$i]; # find leading 1
  return () if $i<0; # only 0s => nothing to do
  my @sub;
  if( $i==0 or not( @sub = nextstep( @p[0..($i-1)] ) ) ) {
    return () if $i==$#p;
    $p[$i] = 0;
    $p[$i+1] = 1;
    @sub = first( $i, scalar grep { $_ == 1 } @p[0..($i-1)] );
  } 
  return ( @sub, @p[$i..$#p] );
}

my @data = 'a'..'e';

my @logical = first( scalar(@data), 3 );
while( @logical ) {
  print "@logical: ", @data[ grep { $logical[$_] } 0..$#logical ], "\n
+";
  @logical = nextstep( @logical )
}
[download]

While I think it works, give the time it took me to get it right and the somewhat tricky logic, I have come to the conclusion, that using a module is the better alternative.