Latin-1 characters and XML

kilinrax has asked for the wisdom of the Perl Monks concerning the following question:

I'm trying to write a script do the following:

Read in an .unl file of translations from one language to another
Read in a consistent-format XML file, and parse it
Apply the translations to some attributes and text elements in the file
Write the file back out, with all latin-1 characters escaped in numeric form (e.g 'é' -> 'é'

Currently, I'm trying to do this with XML::Twig, but only the attributes are coming out escaped properly, not the text.

Read in translations and construct twig:

open( TRANSLATIONS, $translations_file ) or die "Could not open '$tran
+slations_file': $!";
while( <TRANSLATIONS> ) {
  if( my ($word, $translation) = split /\|/ ) {
    push( @translations, {
                           match => qr/\b$word\b/i,
                           replacement => $translation,
                         } );
  }
}
close( TRANSLATIONS );

my $twig= new XML::Twig(
                         twig_handlers => { 'FOO' => \&foo },
                         pretty_print => 'indented',
                         output_filter => 'safe',
                       );
$twig->parsefile( $xml_file );
$twig->print;
[download]

Apply translations to parts of element 'foo':

sub foo {
  my ($t, $elt)= @_;
  # apply translations to the name, and text of all descendant nodes
  if( my $name = $elt->att('NAME') ) {
    $elt->set_att( 'NAME', apply_translations($name) );
  }
  my @descendants = $elt->descendants( 'BAR' );
  foreach my $descendant (@descendants) {
    if( my $text = $descendant->text ) {
      $descendant->set_text( apply_translations( $text ) );
    }
  }
}

sub apply_translations {
  my $text = shift;
  foreach (@translations) {
    $text =~ s/$_->{match}/$_->{replacement}/g;
  }
  return $text;
}
[download]

Input
`<FOO="Duration"> <BAR>Duration</BAR> </FOO> <FOO NAME="Airport"> <BAR>Airport</BAR> </FOO>` [download]

--->

Output
`<FOO="Durée"> <BAR>Durée</BAR> </FOO> <FOO NAME="Aéroport"> <BAR>A鲯port</BAR> </FOO>` [download]

I've also tried using XML::SAX to do the same thing, but it barfs on the xml version/encoding line, complaining 'Only ASCII encoding allowed without perl 5.7.2 or higher. You tried: iso-8859-1'.
Unfortunately another application running on this machine (not mine, so don't ask me why) requires perl 5.6.0 or 5.6.1 , so upgrading perl would be problematic at the very least.

I'm somewhat loathe to start hacking on a new version using a different module without being sure it's support for latin-1 is better than those I've tried so far.

Does anyone have any experience of doing anything like this, and have recommendations/advice they'd care to share?

Back to Seekers of Perl Wisdom