I think that's the point: Browsers don't bother to validate HTML, but they have to validate XHTML to know whether to pitch a fit about errors as required by the spec.