Abstract: Methods and devices for handling a buffer in packet networks, particularly in regard of loss and delay of packets, which belong to predetermined or implicitly given loss priority classes and delay priority classes. When a packet arrives at the buffer, its loss priority and delay priority classes are determined. For each loss priority class, there is a predetermined threshold value, and a total filling level of the buffer, i.e., the total number of packets stored, is compared to the threshold value of the packet's loss priority class. If the threshold value is larger than the total filling level, the packet is buffered in order to be forwarded; otherwise, the packet is lost. In determining whether a packet is to be buffered or lost, the packet's delay priority is not taken into account. In forwarding packets from the buffer, those packets that belong to higher delay priority classes are chosen in the usual way before packets belonging to lower delay priority classes.

Abstract: The transfer of data packets are monitored in a line having various logical channels. For each logical channel there is a count value and at the arrival of a data packet, or in each time slot when the transmission on said line is time multiplexed, i.e., for each time when a packet could have arrived in a time multiplexed system, only the count value for a single channel is decremented. Then it is determined whether the received data packet, if there is one, is to be forwarded or is to be discarded by the comparison of a calculated value (F.sub.n) to a threshold value (T.sub.n) belonging to the channel to which the packet belongs. In the calculation of the value (F.sub.n), the number is used of those channels which have not been decremented after decrementing the channel to which the received packet belongs. In this way, the calculated value (F.sub.n) can be made to resemble the calculated value or level value, which is used in prior methods for handling loss of packets in a network.