Seems like I must be as bored as Fastolfe....
Here's my own version which doesn't require fixed lengths and is more flexible in ways.

#!perl -w
use strict;
use Data::Dumper;

my %fields_types;
my @data;
my %prev;
my $biggest;

while (<DATA>) {
    if (/^#(.*)/) {            # new field declaration
        my @fields = split ' ',$1;
        $biggest = @fields if @fields > $biggest;
        $fields_types{scalar(@fields)} = \@fields;
    }
    else {
        my @line = split /(?<!,)(?<!\s)\s+/, $_;    # split on whitesp
+ace not following a comma
        if (exists($fields_types{scalar(@line)})) {
            my %tmp;
            @tmp{ @{ $fields_types{scalar(@line)} } } = @line;

            # use data from most recent full line of data
            if (scalar(@line) != $biggest) {
                for my $field ( keys %prev ) {
                    $tmp{$field} ||= $prev{$field};
                }
            }
            else {
                %prev = %tmp;
            }

            push @data, \%tmp;
        }
        else {
            warn "No definition for " . scalar(@line) . " field(s) at 
+input line $.\n";
        }
        
    }
} 

print Dumper( \@data );

__DATA__
#Name ID PS Gender Age Month Code Cap Pool
#Month Code Cap Pool
LName, FName 99999 99 M 99.9 12/2000 Add 99.99 99.99
11/2000 New 99.99 99.99
[download]

Update: Woo, my 100th post, too bad more than half of them came out under 10 points.