Whoever voted down the above node completely missed the point. MrNobo1024 is completely correct in saying that if Perl worked as documented as far back as, say, Camel 2 then it should not have had enough information to calculate $&.