Performance when using tuples

[Mathew Kuthur James]

Hi,

I understand that TLA+ supports non-flat tuples, so {a} and {<<a>>} and {<<<<a>>>>} are considered distinct. My question is - if all sets of the form {a,b,c...} are replaced by sets of the form {<<a>>,<<b>>,<<c>>...}, are there performance implications for model-checking? 

Also, is there any way to dynamically check whether a set consists of strings or tuples, and have a different behaviour in each case?
Something like:

F(S) == G(S) if S is a set of strings, H(S) if S is a set of tuples