It'd be good to have public versions of _dbus_check_is_valid_bus_name() etc., as part of the general principle of "if a function has preconditions, then language bindings should find it easy to check those preconditions themselves".

The API I propose matches the current dbus_signature_validate and dbus_signature_validate_single functions:

/*
 * Assumes that the string is '\0'-terminated, so you must check lengths
 * separately if dealing with untrusted data from the network.
 *
 * Passing a NULL string is considered to be a programming error; callers
 * are expected to check for NULL themselves.
 *
 * If valid (which is flagged as _DBUS_LIKELY), returns TRUE.
 *
 * If invalid, raises an error (currently DBUS_ERROR_INVALID_ARGS) and
 * returns FALSE. The error message is currently generic; as a future
 * refinement, we could include a hint appropriate to the validity
 * failure in the error message, e.g.
 * "com.example.a&b" -> "name components may only contain [A-Za-z0-9_]".
 */
dbus_bool_t dbus_validate_THING (const char *, DBusError *);

with the following validators available:

* dbus_validate_path
* dbus_validate_interface
* dbus_validate_member
* dbus_validate_error_name
* dbus_validate_bus_name (accepts both unique and well-known names)
* dbus_validate_utf8

Thread: http://lists.freedesktop.org/archives/dbus/2011-July/014528.html