I know. :) (They were, however, good examples for the alternative directory structure.) Here are the problematic ones:
is_adaptable_binary_function.hpp (32) is_adaptable_binary_predicate.hpp (33)
Do we need the "adaptable" bit? If yes, then how about: is_adaptable_binary_func.hpp is_adaptable_binary_pred.hpp Maybe not ideal, but as long as you are consistent it should be OK.
is_hashed_associative_container.hpp (35) [*] is_multiple_associative_container.hpp (37) is_pair_associative_container.hpp (33) is_simple_associative_container.hpp (35) is_sorted_associative_container.hpp (35) is_unique_associative_container.hpp (35)
[*] Ok, not standard yet, but used in the N1443 proposal.
As you can see, it's mostly the associative containers that has a problem. Perhaps shorten the filename (and trait?) "associative" -> "assoc"? That keeps it just under the limit. Then what about the two first?
"assoc" sounds good to me. BTW, don't get too hung up on names it could be a bicycle shed issue again :-) John.