Boost hi .. I'm using the finite state filter component in the iostreams library to create a state machine that can easily help me process incoming data streams from various medical sensors. I'm currently working around the following scenarios and wanted to run them by the list to see if the method I chose was the preferred one. 1- Constant datastream with no variation in datagrams. Solution: use finite state filters as they come, works ok ! 2- Variable datastreams with constant size in datagrams. Solution: use finite state filters with a transition table that can interpret anything and choosing in the processing function the right way to go, not very clean, but works !! 3- Variable datastreams with variable sizes. Solution: trying to change the finite state filter on the go, currently doesn't work. I noticed that when bytes are inserted into the state machine these don't get processed until a certain buffer size is met internally, resulting in a behaviour where data gets analyzed long after it has been inserted. I saw that the flush() function for the iostream that contains de state machine suffices when called before inserting a new character with a line such as: io::flush<io::filtering_ostream> ( this->data ); but the problem is that the only mechanism that can determine when a whole package has been received and one can change state machines to analyze a new state lies within the state machine itself (in the processing functions), I can't issue a flush() from within a state machine since it won't execute until a specific buffer size is met. Any ideas on how to go around this matter ?? I'm currently using boost 1.53 on Linux Thanks ...