A system and method for minimizing adverse affects of network transfer latency on packet data transmission that reduces degradation of a perceived quality of the data transmission. The system includes at least one interface device linking a first computer network to a second computer network. The interface device receives at least one group of packets transmitted across the first network in a buffer, performs packet-based media processing on the packets based on a type of information content of the packets, and transmits the processed packets to the second network. While receiving the group of packets, the interface device determines whether any packets are "late". The interface device waits longer for late packets likely to contain active speech (or active music) than late packets likely to contain silence. Further, the interface device waits a fixed time for late packets containing fax data. In the event a maximum waiting time is exceeded, the interface device declares the corresponding late packet "missing" and drops the missing packet.