Without boost::filesystem, it is quite common for programs to take a path as user input and pass it to the OS without caring much if it's valid native path or not. If it isn't, the OS function call will fail. I don't think that this common use case would change much if it's ported to use boost::filesystem. This leads to my questions: if I can store any string into a path object, how are the semantics of boost::path different from the semantics of a string? What is the path type invariant? Does it give me any guarantees about the stored string? Emil Dotchevski Reverge Studios, Inc. http://www.revergestudios.com/reblog/index.php?n=ReCode