I think you want something like the following.

use strict;
use Parse::RecDescent;

my $parser = Parse::RecDescent->new(q(

        start: chunk(s)

        chunk: tag | raw

        tag  : stag raw etag
        { 
        print "tag: $item[1] $item[2] $item[3]\n";
        #$return = "$item{stag} $item{raw} $item{etag}" 
        $return = "$item[1] $item[2] $item[3]";
        }
        | stag raw tag(s) raw etag
        { 
        print "tag: $item[1] $item[2] $item[3] $item[4] $item[5]\n";
        #$return = "$item{stag} $item{raw} $item{etag}" 
        $return = "$item[1] $item[2] $item[3] $item[4] $item[5]"; 
        }
        #outputs: 
        #tag: <open> tag 1  <close>
        #tag: <open> tag 2  <close>
        #tag: <open> tag 3 b  <close>
        #tag: <open> tag 3 a  ARRAY(0x1a55ad8) tag 3 c  <close>
        #tag: <open> tag 4 b  <close>
        #tag: <open> tag 4 d  <close>
        #tag: <open> tag 4 c 
        #         ARRAY(0x1a8f664) tag 4 e 
        #     <close>
        #tag: <open> tag 4 a 
        #     ARRAY(0x1a8f670) tag 4 f
        # <close>
        raw  : m{([^<]*)}
        stag : m{<open>}
        etag : m{<close>}
));

#slurp mode
local $/='';
$parser->start(<DATA>);

__DATA__
<open> tag 1 <close> 
notatag 1
<open> tag 2 <close>
notatag 2
<open> tag 3 a <open> tag 3 b <close> tag 3 c <close>
notatag 3
<open> 
    tag 4 a 
    <open> tag 4 b <close> 
    <open> 
        tag 4 c 
        <open> tag 4 d <close>
        tag 4 e 
    <close> tag 4 f
<close>
notatag 4
[download]

I am 95% happy with this code. If only I wasn't getting those ugly hex refs in the output.

Maybe some other monk can help with that :)

thomas. UPDATE: Follow up at Parse RecDescent Nesting (Followup)