I usualy suggest using a proper parser when working with XML/HTML. Several are available, see HTML::TokeParser Tutorial from the tutorials section of this site.