A short search with Google brings me to this page, which tells me that most Netscape strings start with something like Mozilla/x.xxx(, and after that first opening parens, there is not the word compatible. This could be checked for with the following crummy code :

sub header_NS {
  # Assume that nobody uses Netscape anymore
  my $result = undef;

  # Get user agent passed as the only parameter
  my $user_agent = shift;

  if ($user_agent =~ m#^Mozilla/\d+#i) {
    # We have a Mozilla/NetScape compatible browser
    if ($user_agent !~ /compatible/i) {
      # And it's not even just "compatible", it's the real,
      # CSS burning, X crashing hellspawn
      $result = 1;
    };
  };

  return $result;
};
[download]

The above code checks for any browser that claims to beo Mozilla, but also claims to be the real thing and not something compatible. But I think your idea of using User-Agent headers for this is flawed, as there are many instances where proxies change the user agent and there are also other browsers which have real problems with CSS support. I recommend you put a "This site without CSS" link into your script, which calls your script again with the parameter ?nocss=1 appended to the URL. This allows users with broken browsers to easily use your website even if your magic detection does not work because you didn't anticipate their specific browser.