FAITH... WITH REASON

One of the fundamental principles of logic is that the propositions must be assumed and can't be proven within the framework of the formulated argument. One of the fundamental problems of philosophy is how to determine what premises to assume to begin with. However, I'd argue that we do not have to take principles on faith. Rather, I think there's a recursive relationship between rationality and empiricism. In particular, we can make assumptions, test their consequences, and if they match the empirical results then that's evidence that the assumptions are true. If you take math as an example, the peano axioms are assumed but not proven; we continue to assume them because they seem to accurately model reality (ie, they match our empirical experiences). I would say that Occam is much the same way, and we assume it on the same principles by which we assume mathematical axioms. 

I do realize there are some realist problems with the notion of simplicity. Even within the realm of computational complexity much depends on what language we choose. Using binary is one way to try to get around this, since binary is universal and the simplest language possible (you can't create a language simpler than one that relies on true/false). That may ultimately prove impractical, but I'm just speaking in principle here. For some things, however, we needn't bother with these gray areas of simplicity/complexity: I keep using it as an example, but the conjunction fallacy applies no matter what language we choose or how we define complexity/simplicity, and the variant/variant distinction is a pretty straight-forward example of it. 

RE your second paragraph: In philosophy, metaphysics doesn't mean "beyond the physical," but just refers to whatever is--all the notions surrounding being. If there is something beyond the physical, this would fall under metaphysics; but the existence of atoms fall under the category of metaphysics as well. Most rationalists don't bother with metaphysics since we recognize that ultimately what we're concerned with is making it so our models--perceptions of how reality functions--matches our phenomenal experience of reality. If they do, we don't see much point in questioning it beyond that point since it just turns into intellectual masturbation with no experiential consequences. So when I say that Occam is an "ontological principle" I'm saying that in the sense that I (and other) conflate our empirical experience of reality with reality itself, so that it's a principle of our map-making abilities. The same way you might say when looking at a tree that "the tree exists" without going through the rigmarole of saying "I'm experiencing my visual cortex's interpretation and reconstruction of photons that are bouncing off an object that may or may not exist beyond by perception in the form that I call a tree." 

Here's a good read on the subjective of recursive justification: lesswrong.com/lw/s0/where_recursive_justification_hits_bottom/