<input type="text" value=">">

Why not just use HTML::Parser ? That would be the correct way of doing it. And it is fast too, both to write and execution. Just subclass HTML::Parser, and use the text method, like this:

package MyParser;
use base 'HTML::Parser';

sub text { 
    my($self, $origtext, $is_cdata) = @_;
    print $origtext;
}
[download]

Autark.