It may have taken me a year of chipping away at it and my process is definitely suboptimal, but at least I have the satisfaction of having worked it out myself.
Also, play with it! It's an incredibly simple but totally mind bending device.
There is a trick to finding solving patterns known as commutators. For instance, see http://w.astro.berkeley.edu/~converse/rubiks.php?id1=basics&...
Once you learn the basic idea of the commutator, you can easily make up you own system. Essentially you select a part of the cube and make some specific change to it. For instance, you come up with a move sequence that turns one top corner but doesn't affect any other top cubie. If you were to reverse that move, you would be back to normal or course. So if you rotate the top before reversing the move: 1) The only change on top will be a different corner turning 2) The rest of the cube will go back to it's original state.
Therefore, the only difference is that you rotate two corners. It takes some practice to learn how this is done, but I can solve any cube-like puzzle (ok... except Square-1)and I don't have any memorized system.
Thus [A] B [A'] B' is all you need. You just need to find a proper [A]
The way I attacked it (and suspect people naturally would) was a layer at a time. When you're doing that you try to preserve what you've already solved.
Once you have 2 layers sorted and you're trying to do that last one there's not much room to manoeuvre. I spent countless hours coming up with a combination that preserved my top 2 layers and following where each individual piece moved in the final layer. If nothing else, it's great for memory and spatial awareness.
I probably wouldn't have bothered to learn to solve the 3x3x3 and 4x4x4 if I hadn't read how to do it. There's a lot of enjoyment to be had from learning the different methods, and my understanding of how the cubes function is still changing as I learn more efficient methods.