You shouldn't be using the term "length"; for sets that are not unions of intervals (with more than one point), there is no notion of length that is easy to define. There is a notion "Lebesgue measure" that ultimately does give a notion similar to "length", but it is rather more advanced than the set-theoretic notions of being countable (like $\Bbb Q$) or not (like $\Bbb R$).
The situation of ordering of rational and irrational numbers is in fact rather counter-intuitive, as it marks a sharp deviation from what happens in the finite case. Imagine you want to make a fence across a certain opening. You can put two fence posts at the ends, and connect them with a wire; there is one more fence post then there are pieces of wire. Now you can add some fence posts in between; each time you add a fence post, you also divide one piece of wire into two, so it is easy to see (by induction) that at any time you will have one more fence post then piece of wire between them. (Some people tend to forget this small difference of one, and an error in computer programming that is due to this oversight even has been named after this.)
Anyway, no matter how long you go on subdividing, there are always nearly equally many fence posts as there are intervals between then, with the fence posts staying just one ahead of the intervals. However when in one fell swoop you put in (countably) infinitely many fence posts (like one at every rational number) then a surprising thing happens: then number of intervals has become much greater than the number of fence posts. Also intervals are in general no longer delimited by given a pair of fence posts; rather they are determined by which fence posts are to their left, and which are to their right (technically this is called a Dedekind cut). In fact, supposing the set of locations of fence posts is dense, not only has the length of the intervals become zero, each interval actually contains just a single point. And furthermore supposing the fence posts are precisely at the rational numbers, the intervals correspond precisely to the irrational numbers.