Standard for the transmission of 802.2 packets over IPX networks
also
- std-49
Extracted elements (15)
Because the IPX programming interface insulates applications from knowledge of the physical layer, 802.2-over-IPX works over any physical medium without modification. This physical-layer independence is cited as a primary implementation benefit.
Encapsulating 802.2 over IPX rather than defining a custom framing allows any network protocol that supports the IEEE 802.2 Data Link Layer specification to be carried over IPX, including IP, ARP, and CLNP, and enables reuse of the RFC 1042 IP-over-IEEE-802 framework.
No unicast packet is ever sent from host1 to host2 without a prior broadcast from host2 to host1. This invariant guarantees that the immediate address of any intervening IPX bridge can be learned by inspecting the physical header of that prior broadcast, eliminating the need for an explicit bridge-discovery mechanism.
This specification does not use the IPX checksum field, following the convention of most IPX applications. No normative reason is given; it is treated as a standard omission.
Existing ARP implementations must be modified to allow cache entries larger than the standard hardware address length, to accommodate the 10-byte 802.IPX address combined with the possible immediate address of an intervening IPX bridge associated with a given IP address.
This specification obsoletes earlier documents for transmitting Internet packets over IPX networks. It differs from predecessors by supporting multiple network protocols simultaneously over IPX and by enabling packet transmission through IPX bridges.
Mapping of Internet Protocol addresses to 802.IPX addresses is performed using ARP, in the same fashion as with other IEEE 802.X physical addresses, with the distinction that the hardware address length is 10 bytes rather than 2 or 6.
An 802.IPX physical address is 10 bytes: 4 bytes of the IPX network address followed by 6 bytes of the IPX node address. This 10-byte address is used as the hardware address in ARP mappings for IP-over-IPX hosts.
IPX socket number 0x8060 has been reserved by Novell for the implementation of this protocol (802.2 encapsulation over IPX). Implementations must bind to this socket to receive encapsulated packets.
Security issues are not addressed in this memo.
A typical IP-over-IPX packet is layered as: physical header (N bytes), IPX header (30 bytes), 802.2 LLC+SNAP header (8 bytes), IP header (typically 20 bytes), TCP header (typically 20 bytes), and TCP data (up to 498 bytes).
Broadcast packets are formed by setting the IPX header Packet Type field to 0x14, Destination Network to the local network number, Destination Node to 0xffffff, and the Immediate Address field of the IPX Event Control Block to 0xffffff.
The byte transmission order for 802.2-over-IPX is big-endian.
The maximum data size of an IPX datagram is 546 bytes. With the combined 802.2 LLC and SNAP headers consuming 8 bytes, the effective Maximum Transmission Unit (MTU) for 802.2-over-IPX is 538 bytes.
Unicast packets are formed by setting the IPX header Packet Type to 0x04, with Destination Network and Destination Node taken from ARP resolution, and the Immediate Address of the IPX Event Control Block set to the physical address of the destination node or the appropriate IPX bridge.