I think that a utf8 regex pattern with the /d modifier should have unicode semantics. But I'm open to contrary opinions. I had always assumed that it should, but it doesn't currently.