Just an idea, I'm not sure whether you can use it.

If you make a copy of a large string you've read, the two copies will have separate pos-itions, thus, you can traverse both with regexps without the need for substr'ing it.

For example:

$string1= "hello, world\n"; $string2= $string1; pos($string1)= 4; whil
+e ($string1=~/(.)/gs) {$win_right= $1; $string2=~/(.)/gs; $win_left= 
+$1; print " "x$n++, "$win_left...$win_right\n"}
[download]

prints:

h...o
 e...,
  l...
   l...w
    o...o
     ,...r
       ...l
       w...d
        o...
[download]

This will of course have difficulties on what to do at the boundary of two blocks you've read.

Update: spelling. Whether, whether, whether, whether, I'll have to learn this eventually.