It is small (97 strokes, including the meaningful newline), but it is also very fast. There are certainly places where a few characters can be shaved but several of those break subtle features such as one demonstrated in the example code: sorting the sample data correctly even when each record ends with a newline.

use strict;
use warnings;

sub naturalSort {

my$i;s/(\d+)/pack"aNa*",0,length$1,$1/ge,$_.=$".$i++for
# 345678 1 2345678 2 2345678 3 2345678 4 2345678 5 2345
my@x=map{lc}@_;@_[map{(split)[-1]}sort@x]
#8 6 2345678 7 2345678 8 2345678 9 234567

}

my @data= <DATA>;
chomp @data   if  $\;
print for naturalSort(@data);
__END__
amsTerdam40
amstelveen
1
AmstErdam5
Amsterdam
amsterDamed
A
[download]

Run it normally or with "perl -l ..." to see that it sorts the same with or without the newlines in the strings being sorted.