If you want to do everything perfectly formally you're better off studying a computer based theorem prover like Coq. Not only if everything proven perfectly formally, the proof is also checked by a computer. This also allows conciser proofs because the theorem prover can often generate parts of your proof.
http://mathoverflow.net/questions/14356/bourbakis-epsilon-ca...
Usually mathematical arguments are relatively informal; every time you use set comprehension notation, for example, you're skipping a few steps in your proof (and note, of course, that the impredicative definitions one can use with it are open to e.g. Russell's paradox, which is why we have axiomatic set theory in the first place).