If you just want to check if the string matches, Limbic~Regions example will probably do it.

Although you didn't say it, I suppose you want to catch the substrings between 'start' and 'end' - otherwise you don't have to care about overlapping. In that case you could do something like the following:

#!usr/bin/perl
use warnings;
$_ = "start b2 end start b2 b2 end start b2 end";

@Matches = /        # list context to catch all
  (?<= \b start \b )# 'start' must precede
  ( .*? )           # catch fewest possible anythings
  (?=  \b end   \b )# 'end' must follow
/gx                 # global and expressive

and print join("\n", @Matches);

__DATA__
 b2 
 b2 b2 
 b2
[download]

update: normal matching without look-ahead and look-behind will have the same effect with the example string:

@Matches = /
  \b start \b
  ( .*? )
  \b end   \b
/gx
[download]

update: I overlooked that Limbic~Regions regex catches the substring too. Shame on me!

^~Django
"Why don't we ever challenge the spherical earth theory?"