In scalar context, it's usually called the flip-flop operator.

What it's doing is printing all the lines that contains *end* and all the lines containing *start*, and every line in between. If *start* or *end* occurs in the middle of a line, so be it. If multiple *start* or *end* tags exist in one line, so be it.

The whole concept of line is kinda odd when dealing with binary data anyway. Usually one reads fixed-width blocks of data when dealing with binary data because you'll never know how long a line will be. (0 bytes? 1 bytes? the entire file?)

Assuming fixed-width delimiters, your code should look like this:

#!/usr/bin/perl -w
use strict;

use constant BLK_SIZE => 64*1024;

sub process {
   my ($data) = @_;
   print($data);   # Or whatever
}

sub read_more {
   my $fh = shift;
   my $rv = read($fh, $_[0], BLK_SIZE, length($_[0]));
   die $! if !defined($rv);
   return $rv;
}  

my $start = '*start*of*junk*data*';
my $end   = '*end*of*junk*data*';


my $qfn = "file";
open( my $fh, '<:raw:perlio', $qfn )
   or die("open $qfn: $!\n");

my $buf = '';
my $in_junk = 0;
while (read_more($fh, $buf)) {
   if ($in_junk) {
      my $pos = index($buf, $start);
      if ($pos >= 0) {
         process(substr($buf, 0, $pos, ''));
         substr($buf, 0, length($start), '');
         $in_junk = 1;
         redo;
      } else {
         process(substr($buf, 0, -length($start)+1, ''));
      }
   } else {
      my $pos = index($buf, $end);
      if ($pos >= 0) {
         substr($buf, 0, $pos+length($end), '');
         $in_junk = 0;
         redo;
      } else {
         substr($buf, 0, -length($end)+1, '');
      }
   }
}

process($buf) if !$in_junk;
[download]