FOKKINGA@HGRRUG5.BITNET (03/18/88)
Our University TCP/IP network has been assigned a class B internet number. We decided to use subnets and we expected 8 bits to be insufficient for the addressing of hosts, so we assigned 6 bits to subnets and 10 bits to hosts. However our main ethernet network keeps growing and growing and all kinds of local nets are interconnected by means of bridges. In the near future we will run out of host addresses. So we will have to reduce the number of subnet bits still further. This has the disadvantage that networks that are connected via IP routers do consume a large amount of addresses for only a few hosts. Our question is therefore: If we want to keep the same number of subnet bits, do we have to add routers or gateways in the network or is it possible to use some trick so that more than one subnet address can be used on the same physical network? And finally: How do people manage to live with 8 bits for host addressing. Is there some disadvantage in using such a large ethernet network that we have overlooked so far? Thank in advance, Douwe Fokkinga Rekencentrum Rijks Universiteit Groningen BITNET: FOKKINGA@HGRRUG5
jas@MONK.PROTEON.COM ("John A. Shriver") (03/19/88)
There is no restriction with having one Ethernet be two, three or even four subnets. The cost will be that machines on different subnets of the Ethernet will have to use a router to communicate. One trcik can be played to work around this: assign the ethernet subnets so that they only differ in the lowest bit(s) of the subnet portion of the address. You tell the gateways the true number of subnet bits, but lie to the hosts and tell them there are one (or more) less subnet bits. They then can communicate between the subnets without using routers. All the routers MUST agree on the number of subnet bits used on a given network.