Comments on Gepner models and type I vacua in string theory