UNIX System Administration Handbook - Evi Nemeth [162]
For example, a UDP packet being transmitted over Ethernet contains three different wrappers or envelopes. On the Ethernet wire, it is “framed” with a simple header that lists the source and next-hop destination hardware addresses, the length of the frame, and the frame’s checksum (CRC). The Ethernet frame’s payload is an IP packet, the IP packet’s payload is a UDP packet, and the UDP packet’s payload is the actual data being transmitted. Exhibit B shows the components of such a frame.
Exhibit B A typical network packet
We use the term “byte” to refer to an 8-bit data unit. In days of yore, “byte” was a more general term, so you will often see the term “octet” used in RFCs instead.
The link layer
In this section, we cover several topics that bridge the gap between the lowest layers of the networking software and the network hardware itself.
Ethernet framing standards
One of the main chores of the link layer is to add headers to packets and to put separators between them. The headers contain the packets’ link-layer addressing information and checksums, and the separators ensure that receivers can tell where one packet stops and the next one begins. The process of adding these extra bits is known generically as framing.
Two different standards for 10 Mb/s Ethernet framing are in common use: DIX Ethernet II and the IEEE 802.2 LLC SNAP.3
UNIX hosts generally use Ethernet II, as do Cisco routers. Novell and IPX networks normally use 802.2. Ethernet and 802.2 differ in some fields of the frame header but do not conflict, so receivers can determine unambiguously which format is being used by each individual packet and decode the header appropriately.
The framing that a machine uses is determined both by its interface card and by the interface card’s driver. On PCs running Windows you can choose which style of framing you want, but on UNIX you usually cannot. Both types of framing interoperate just fine from UNIX’s perspective. On the other hand, Windows machines that use different framing on the same network cannot talk to each other. As a sysadmin, you usually don’t need to worry about framing mismatches unless you are performing low-level debugging of a mixed network.
Ethernet cabling and signalling standards
The cabling options for 10 Mb/s Ethernet are fairly straightforward, but things get somewhat more complicated in the 100 Mb/s arena. Years ago, there were three different standards for twisted-pair wire (TX, which uses two pairs of Category 5 wire, and T4 and VG, which each require four pairs of Category 3 wire) and one for fiber (FX, which uses multimode fiber optic cable). Hewlett-Packard championed the VG standard (which wasn’t really Ethernet) and was the first to bring a product to market. Other manufacturers ignored the VG system and scurried to implement TX. TX is now used everywhere.
Refer to Chapter 15, Network Hardware, for more information about the various Ethernet standards. Another useful reference to the ins and outs of Ethernet is the web site www.host.ots.utexas.edu/ethernet, which is maintained by Charles Spurgeon.
Wireless networking
The IEEE 802.11 standard attempts to define framing and signalling standards for wireless links. Unfortunately, it was originally rather vague and included several parameters and options that were not fully specified. One interoperability issue you may need to pay attention to is that of “translation” vs. “encapsulation.”
Translation converts a packet from one format to another; encapsulation wraps the packet with the desired format. Windows systems tend to default to encapsulation and UNIX systems to translation; the wireless base stations must be explicitly configured. If you are deploying a wireless network, you must make sure that your base stations and the workstations they talk to are all operating in the same mode.
Laptop computer users encounter another