OK, I know everybody was waiting for me to use my hammer ;--) ... here is a solution using XML::Twig. One big caveat though is that XML::Twig's version of XPath is way, way, _WAY_ less powerful than what XML::XPath offers. No functions except string, complex sub expressions not supported, you name it. It does /foo/bar/text/* though ;--)

#!/usr/bin/perl -w
use strict;

use XML::Twig;

my $twig = XML::Twig->new( pretty_print => 'indented');
$twig->parse( \*DATA);

# the * means that the nodes returned will be either #PCDATA 
# or #CDATA, this would not work if the content of text was...
# not text but included sub elements
foreach my $node ($twig->find_nodes('/foo/bar/text/*')) { 
    my $data = your_munge_function($node->text);
    $node->set_text( $data);
}


$twig->print;

sub your_munge_function
  { return "munged $_[0]"; }

__DATA__
<foo>
<bar>
<text id="text1>"><![CDATA[La dee da de da.<br>Foo bar baz]]></text>
<text id="text2>">a normal text></text>
</bar>
</foo>
[download]

Wahoo! I just tested it and that solution will work perfectly. XML::Twig supports what I need to do and the code changes will be pretty minor. Thanks for your help!

-Mike