Not everyone likes @Hixie's spec writing style, but it makes implementing the tokenizer require almost no thought. This is a *good* thing.