To preface this: in case this is mistaken for a personal attack of some sort, I'm not trying to do anything of the sort. All I want to do is promote a sensible, concise, and correct way of doing something, though this is not to say that I'm not open to new ideas.

There are a few things that set my hair on fire, not unlike the way that people rolling their own "CGI" routines does to others, and IP address and e-mail address parsing using oversimplified and/or overcomplicated regexes is one of them. It is often because the programmer is applying a superficial understanding of some concept, which only creates problems for users later.

An example is someone who writes a routine to verify postal codes, the sort of thing you see on Web sites all the time. For some reason, it is assumed that no sensible postal code would have letters in it:

if ($zip !~ /^\d+(\-\d{4})?$/)
{
     $error{zip} = "Invalid postal code.";
}
[download]

So I pound in something like 'M6K3J8' and I'm chastised, even though I can choose both country and province from a drop-down which lists them explicitly.

To bring this back to the task at hand, inet_aton is tough to match in terms of both accuracy and ease-of-use. Stuff goes in, and if it passes the mustard, a packed number comes out. No messy stuff, and no magic.

Further, if inet_aton can't figure out what to do with the address, it is virtually guaranteed that no other program will be able to either since the vast majority of Internet software uses that very function, or a functionally similar equivalent. On the other hand, even if something passes through a user constructed regex, there is still opportunity for that to fail later on when the inet_aton has a crack at it, which it most surely will if the address is ever used for a connection.

That being said, if, for whatever reason, you absolutely had to use a regex, then I would submit that an improved version would be:

sub valid_ip
{
     my $bit = '[01]?\d\d?|2[0-4]\d|25[0-5]';
     return $_[0] =~ /^($bit)(\.($bit)){3}/;
}
[download]

However, it still suffers from the same problems as the original.

Another method, though this is really not an improvement on the regex method in terms of being easier to understand, goes about it quite literally:

sub valid_ip
{
        my $bits = 0;
        foreach (split (/\./, $_[0]))
        {
                return unless /^\d+$/ && $_ >= 0 && $_ < 256;
                $bits++;
        }

        return $bits == 4;
}
[download]

I'm not sure what your remark about IPv6 compatibility implies. Obviously the standard inet_aton function will not handle IPv6 addresses. Not that the regex proposed could handle them either.

The IPv6 Address Specification is even more "open to interpretation" than IPv4, such that writing a regex for that would be considerably more difficult, and consequently, more likely to fall short of the mark.

It would, however, be easy to upgrade the inet_aton approach to IPv6. When Perl adds native support for the IPv6 equivalent method for inet_aton, or now by using a simple module written by Tony Monroe available on CPAN:

use Net::IPv6Addr;

sub valid_ipv6 
{
     return Net::IPv6Addr::ipv6_parse($_[0]);
}
[download]