Another solution, which may applied for all those messages that appear NOT to follow the typical /^--/ signature indicator is to read lines from the back of the message up, until you get to a white space line. Then if the last read line (eg the line after the white space) starts with any repeated symbols, such as - or *, it's a good bet that it's a signature. This will err on the side of keeping anything that doesn't look like a sig, even though it might be.

my @message = (<MESSAGE>);
my @possible_sig;
my $line = pop @message;
# Read past white space that may trail message
while ( $line =~ /^\s*$/ ) { $line = pop @message; }
# Read in what may be the sig...
while ( $line !~ /^\s*$/ ) {
    unshift @possible_sig, $line;
    $line = pop @message;
}
# at this point, @message contains the original message,
# minus the last 'paragraph' minus trailing whitespace.
# @possible_sig contains this last paragraph.
# If the first element of @possible_sig does look 
# like a standard sig, then we just drop it, and @message
# is ready to go.  Otherwise, just reattach it.
if ( $possible_sig[ 0 ] !~ /^[-\*%]{2,}/ )
{
    push @message, $line;  # (put that whitespace back!)
    push @message, @possible_sig;
}
# @message can now be written to file.
[download]

update: er, code was right, comment was wrong. @possible_sig should be dumped if it looks like a sig.

---

Dr. Michael K. Neylon - mneylon-pm@masemware.com || "You've left the lens cap of your mind on again, Pinky" - The Brain