> Oh and (R2U2)^3 doesn't count!

Of course it does! That's one of the 20 algorithms with parity.

Here are some longer ones (I try to use RULDM only, because it's fast

for me):

R2 D' L2 U' L2 D' L2 U' L2 D' R2 D' swaps UF and UR edges.

U R2 U' R2 y' M2 D L2 D' L2 D' R2 D R2 U' R2 x2 swaps UBL and UFR corners.

y' R2 U R2 U' R2 D R2 U' R2 U' R2 U R2 U' R2 U2 R2 swaps UFR and ULF

corners.

> Now that I think more about it, You can't use the standard 3x3 PLLs

> either due to the turning restrictions. Hem..., at the same time

> there would be added flexibility of not caring about the

> 4 "invisible-edges"/"mystery-edges" or being off by an E (assuming

> holding it a certain way). I bet there are tons of cool PLL algs for

> this puzzle.

You could say 'tons'... I usually got about 50-100 optimal algorithms

when I searched, so I guess you have a lot of stuff to choose from. On

the other hand, the optimal algorithms can be pretty long, and all the

180-degree turns are slow.

> On second thought, I think I'd go with a CF method. What do you guys

> think?

That's a pretty good idea, actually, except that you wouldn't be able

to do the edges in one algorithm. On the other hand, if you sort the

edges, you can do edge permutation on each layer separately, and the

algorithms for that are 12 moves or less.

> Okay I really want a domino now!

Me too. I usually emulate on a 3x3x3 (ignore middle layer) or 4x4x4.

If I ever decide to seriously speedcube it, I'll have to get or make a

real one.

I haven't decided whether it's faster to have a numbered or colored

cube. Is it a valid record, I wonder, if you use a 3x3x2 cuboid?