Prioritization of fragmented messages in an ORB is achieved by intercepting the flow of message data within the ORB, marking the current data stream location and issuing sufficient read commands to gather the message header and control area information. Then classification for priority can be performed using the data in the control area to classify the request. The classification information is stored for the message for later use by queuing and prioritization processes and the data stream marker is reset to the original location. When the queuing and prioritization processes are ready to dispatch the message, the message can be sent on to the higher levels of processing within the ORB where the higher levels of message processing code will process the message, totally unaware of the fact that the message has been classified and/or queued or prioritized prior to delivery to the higher level.