Finally, finally, someone in a position to know these things is backing me up. Nathan Bransford says you can't just barf out a whole first draft of a novel, then edit it later. Yes, you need to free yourself from the "inner editor" enough to get something down. Enough to write a novel as opposed to, say, flash fiction. BUT you can't build a house on a shaky foundation, as Bransford says. And that's what all these "just do it" people (e.g. NaNoWriMo) don't mention: if there's a termite of a bad idea when you start, by the time you're done, that termite will have devoured your novel. If there's a twist on the track at the outset, it can throw your train off the rails and it might never get back on. And so on. Analogize at will.
So it seems to me the process of beginning must be process of writing, stopping, thinking, rewriting, rethinking. You can't rush this process, nor should you despair if you aren't just plowing ahead, vowing you will fix everything later. Some stuff, if you get it wrong, isn't fixable. The key is to recognize what is fixable later, and what isn't. Before you've had time to make these distinctions, I'd argue that a NaNoWriMo-style sprint isn't worth it--unless you are so blocked that you won't write at all unless you do this. There is too much advice to writers that advocates powering through at all costs, and dismisses all noodling and hemming and hawing as procrastination. Not so, or not always. Again, you have to learn to tell the difference, and that takes some practice. In the early stages, the bottom line is not if your novel is getting bigger, but only if it's getting better. Once you're satisfied with the foundation, then, yes, write like crazy.
There's a Paris Review interview to back this up, too. See Peter Carey.