> My argument isn't to make it easy to program in C from ML, but to make it > so that the syntax follows the `least surprise' principal. Also the general > idea of being able to name things, which can't be done if it is more than one > type separated by commas. I'm becoming more convinced.