As usual, investigating something relatively mundane yields some fascinating information. Any zip code database you come across in the next few months/years will probably be out of date -- zip codes are offically "mapped" by the census, and the Census Bureau hasn't completed last year's -- and it seems like the results of the survey are accurate to a county level.

The post office, naturally has their own interface which you can automate through LWP trickery. They indicate that the longest their data will rot is 2 months

If this is for big commercial use type work, maybe you wanna buy the TIGER/ZIP+4 database the USPS offer -- it cross-references ZIP code to geographic data.

The odd thing is that every database ( including the unfortunately named MAGGOT) warns that any data you get from any source is probably inaccurate to a (varying) degree.

Anyway, with the wide variety of possible ways to manipulate a ZIP code, here's one way to fulfill your original request :

#usr/bin/perl -w

use strict;
my $zip = "12345";
my @accepted = qw (12390 12351 12345 01241);
foreach my $thiszip (@accepted){
    print "Matched ", cmpZIP ($zip, $thiszip), "\n";
}

sub cmpZIP {
    my ($zip_a,$zip_b) = @_;
    if ($zip_a eq $zip_b){return 5}
    
    my ($ichi, $ni, $san)=unpack ("a3aa",$zip_a);
    my ($ein, $zwei ,$drei)=unpack ("a3aa",$zip_b);
    
    if (($ichi eq $ein)&&($zwei eq $ni)) {return 4}
    if ($ichi eq $ein) {return 3}
    return 0;
}
[download]

Although, after reading up on the subject, you may want to change this to a2a3 or a2aaa to reflect the state/county encoding in the ZIP.