Should true/false for Choice be encoded as 0x55 and 0xAA instead of 0/1?

[sonti-lurker]

Why are true/false encoded as 0u8 and 1u8 and not something with constant hamming weight like 01010101 and 10101010?

I understand that it's not required for constant time operations but it seems to be common practice in protected cryptography libraries. It provides a starting point for defense against power analysis. It would also prevent some timing issues if an extra smart compiler or processor tries to optimize a & b for the case that one of them is all-0.
It does complicate the code somewhat, I suppose.

[tarcieri]

Per #101 I think the best choices are zero and non-zero.

This is because hardware predication instructions work in terms of zero and non-zero. Using anything other than zero for false will not work with those instructions.