Well, there's no reason to read both files into mem.. you can just process/replace the items in file 2 as you read it..
for example

open(FILE, "<$file1") || die "oh the misery, $!";
while(<FILE>) {
  chomp;
  # making the assumption that each field 
  #  is delimited by comma space, safe?
  ($id, $desc) = split(/, /);
  $update_hash{$id} = $desc;
}
close(FILE);

#create backup
rename $file2, "$file2.bak" || die "bah, skum $!";

open(INFILE, "<$file2.bak") || die "dagnabit! $!";
open(OUTFILE, ">$file2") || die "ookiemouth! $!";
while(<INFILE>) {
  chomp;
  # once again, assuming data is delimited 
  #  by comma space
  @vals = split(/, /);

  # update if a new description exists
  $vals[4] = $update_hash{$vals[2]} if $update_hash{$vals[2]};

  # write file delimited by comma space (even last item)
  foreach $data (@vals) {
    print OUTFILE "$data, ";
  }
  print OUTFILE "\n";
}
close(INFILE);
close(OUTFILE);
[download]

Not a whole lot changed, just some extra stuff, including more validation..
-Syn0