Given two chains of similar length, one is clearly more difficult than the other.  For example suppose the test network was 'compatible' with the official network and as a result of a network merge a client had to pick which of the two chains to use.  The only difference is the amount of CPU time that went into generating each, but they are of the same length.

I presume you could simply sum the difficulty of each block and the chain with the highest sum is chosen over the 'longest' chain?

Bitcoin does something like this. Look for "work=" in debug.log.