My system for parity/CP/EP is as michael described. I use 2 sets of corner permutation algs: the "standard" ones, as on lars' site, and the "parity" ones, which solve corner permutation while affecting parity as well. This makes for a little bit more recognition during CP (you check if the top and bottom layers are valid PLL's (on 3x3), if both are or both aren't, do a regular CP. If one is but the other isn't, do a parity CP), but it's not too much longer, and I think my worst parity CP algs are still better than most parity EP algs. That's the other thing: this reduces you to only half of the edge permutation algorithms. The good half. Also, for those who are still learning, the half that are solvable completely with simple 2-2 swaps in combination, like dan suggested.
The other tip that michael gave that I'll clarify is doing EP with CP algs. Basically, you can use the "standard" CP algs to do any combination of U's and H's during EP. As you (should) know, all the standard CP algorithms are basically ortega PBL's. They use the square-1 as a turn-limited 2x2.
Take the J/J algorithm. /(-3,0)/(3,3)/(0,-3)/ Very quick, very easy to remember. You should also know that you can execute the algorithm from a different angle (say, after a (1,0) or a (1,-1)) to move different edges with the same corners. Well, try to do J/J from the normal angle, then again from a different angle. It will do some sort of U-perms. It's fairly easy to figure out how the angles you use affect where your U's are, which direction they cycle, and how to use other CP algorithms (ones that do N's, to be exact) to do H perms in combination as well.
And yes, being able to mirror algs is a great skill. When you think that nearly half of all algorithms on the puzzle are simply a vertical reflection of another, you will realize how important it is. Unless you like learning algs you already know over again.
