I've been mucking with the beginning of an insecure-ish perl module. Basically, if you run taint, you have to regexp and untaint your input. This module presents a shell() function that works like the backtick operator, but will shell-escape anything passed to it. 
<p>
While working on it, I was befuddled by the following:
<ol>
<li>Quoted regexps for character classes can't be interpolated
<li>Quoted regexps for character classes can't use the qr//x for extended regexps.
<li>You have to <tt>map { s/$evil/\\$1/g; $_ } (my @a = @_);</tt><br> instead of
<tt>map { s/$evil/\\$1/g; $_ } @_;</tt><br> to get around the <b>Modification of a read-only value attempted</b>. I don't understand why map is trying to alter @_
with the s///. 
</ol>
</p>

<div class="source">package InSecurity;
use base 'Exporter';
our @EXPORT = ('shell', 'shell_escape');

# kermit may like SUVs, but I keep my environment clean. 
delete @ENV{ grep { !/^(HOME|LOGNAME|USER)$/ } keys %ENV };
$ENV{'PATH'} = '/bin:/usr/bin';

my $evil_characters = qr![\;\|\!\$\`\&\*\(\)\>\<\'\"\\]!;
# no q//x or interpolation with character classes = the suck.

sub shell {
    local($_);
    my $cmd = join ' ', map { shell_escape($_) } @_;
    return `$cmd `;
}

sub shell_escape {
    local($_); $_ = $_[0];

    s/($evil_characters)/\\$1/g;
    /(.*)/ and return $1;
}   

! caller() and do {
    print "Running insecurity tests\n";
    for (keys %ENV) { print "$_ => $ENV{$_}\n" }
    printf("\`echo hi | grep hi\` =>  %s", shell("echo hi | grep hi"));
};


caller() and return 1;
</div><hr /><tt>$_ (the bit s/// operates on) is an alias to each item on @_ inside of
map{}.  It isn't a copy of the item on the list.

Thus, when you modify $_ through s///, you are actually modifying the
elements of @_.

I would guess that you are calling shell() here as

    shell("testing");

rather than

    shell($test);

http://lists.debian.org/debian-user/2003/01/msg00940.html
http://dev.perl.org/perl6/rfc/344.html
</tt><hr /><tt>Thanks! You're spot on; one has to take care regarding where the elements of @_ came from, making the @_ copy to @a almost obligatory. -- Patrick. </tt>