possible problem with memcache's new session feature

Hi there, I just updated to the newest memcache extension and saw a file called memcache_session.c. Curious as I usually am, I took a look at the code and saw an C implementation looking a lot like my old memcache session implementation using the session_set_save_handler in php userland code. While this will work for most people, we've had major issues with this kind of implementation in combination with an Ajax frontend. Since I saw no lock on session information it is very well possible for two requests from the same session to intervene with the session information in a negative way. Example: - request one retrieves session data - request two retrieves session data - request two changes session vars - request two finishes and writes session data - request one finishes after request two and writes it's copy of the session data (modified or original) to memcache Effectively the session data update from request two is lost! Is locking session key-value pairs something that is going to be implemented? If not (or not soon) shouldn't we warn users for this possible problem when using memcache as session storage handler? ps. I know memcache has no native locking mechanisms but it can be emulated (with some performance penalties) by using memcache->add. At session_read, try an x amount of time to get a succesful mc->add(lockkey => lockhash), failing on time out. At session_write, read lockhash = mc->get(lockkey). If not existent or retrieved lockhash differs from sessions lockhash, fail. Regards, Bas van Beek