ng@neutron.mpr.ca (Steve Ng) (04/26/91)
I know that both Ethernet and IEEE 802.3 specify CSMA/CD as the MAC layer, but a guy in here said that Ethernet will not interwork with an implementation of 802.3. In other word, Ethernet is not conforming to the 802.3 standards. I suspicious that he is wrong, but can the Ethernet and 802.3 experts out there educate me on this subject? Thanks. Steve. -------------------------------------------------------------------------- Steve Ng INTERNET: ng@mprgate.mpr.ca or ng%mprgate@cs.ubc.ca MPR Teltech Ltd., UUCP: ...uunet!ubc-cs!mprgate!ng 8999 Nelson Way, CSNET: ng%mprgate.mpr.ca@RELAY.CS.NET Burnaby, B.C., PHONENET: (604) 293-5463 Canada. (V5A 4B5) FAX: (604) 293-5787 --------------------------------------------------------------------------
henry@zoo.toronto.edu (Henry Spencer) (04/27/91)
In article <NG.91Apr25224522@neutron.mpr.ca> ng@neutron.mpr.ca (Steve Ng) writes: >I know that both Ethernet and IEEE 802.3 specify CSMA/CD as the MAC >layer, but a guy in here said that Ethernet will not interwork with an >implementation of 802.3. In other word, Ethernet is not conforming to >the 802.3 standards. I suspicious that he is wrong, but can the Ethernet >and 802.3 experts out there educate me on this subject? Thanks. He's wrong. Ethernet rev 1, Ethernet rev 2, and 802.3 are all completely compatible on the coax, with the possible exception of cases where you are pushing the specs to the limit on things like cable length and repeater count. They are *not* compatible at the AUI level, though, so you need to be sure that transceiver, transceiver cable, and host interface agree on which flavor is in use between transceiver and host. He may be confusing the software issue -- whether the first two bytes of the data in a packet are a protocol code or a length count -- with the hardware. Ethernet revs 1+2 say it's a protocol code, 802.3 says it's a length, and almost everybody does it right and makes it a protocol code. This was simply a botch by 802.3, which most people have sensibly ignored. The two can coexist, but machines that want to communicate with each other naturally have to agree on protocol issues like this. -- And the bean-counter replied, | Henry Spencer @ U of Toronto Zoology "beans are more important". | henry@zoo.toronto.edu utzoo!henry
koning@koning.enet.dec.com (Paul Koning) (05/18/91)
|> |>I know that both Ethernet and IEEE 802.3 specify CSMA/CD as the MAC |>layer, but a guy in here said that Ethernet will not interwork with an |>implementation of 802.3. In other word, Ethernet is not conforming to |>the 802.3 standards. I suspicious that he is wrong, but can the Ethernet |>and 802.3 experts out there educate me on this subject? Thanks. |> |>Steve. |> |>-------------------------------------------------------------------------- |>Steve Ng INTERNET: ng@mprgate.mpr.ca or ng%mprgate@cs.ubc.ca |>MPR Teltech Ltd., UUCP: ...uunet!ubc-cs!mprgate!ng |>8999 Nelson Way, CSNET: ng%mprgate.mpr.ca@RELAY.CS.NET |>Burnaby, B.C., PHONENET: (604) 293-5463 |>Canada. (V5A 4B5) FAX: (604) 293-5787 |>-------------------------------------------------------------------------- |> (I'm getting the impression that there's a routing loop in the Usenet news distribution...) The claim you quoted has some grains of truth in it, but as phrased it's rather misleading and sounds far more ominous than is justified. Ethernet and 802.3 are completely interoperable at the physical level. Stations conforming to either standard will attach to the same cable and will coexist perfectly happily. The frame formats for the two standards are somewhat different. The most important parts ARE identical: CRC (error check) and addressing are the same for both standards. The way protocols are distinguished is different between the two: Ethernet uses 2-byte "protocol types", while 802 uses either one byte "SAP addresses" or 5 byte "protocol identifiers". They all serve the same purpose and provide essentially the same capabilities, but the precise means are different. In practice this is a complete non-issue. When two nodes communicate over protocol X, they have to agree that they are both using X... obviously. Part of the definition of "X" includes whether X is identified by a 2-byte Ethernet protocol type, a one-byte SAP address, or a 5-byte protocol identifier. If the two nodes don't agree on that, no communication is possible. Then again, the same applies when one node sends protocol type 10-10 and the other one is looking for 10-04! One key property is that you can tell from the header whether a frame is encoded Ethernet style (with protocol type) or 802.3 style (with SAP address or Protocol ID). This means that a single node can send and receive in BOTH formats, and tell them all apart. Perhaps the best way to look at it is that a CSMA/CD LAN allows you to communicate in 2^16+2^7+2^39 different protocols (Ethernet protocol type, SAP address, Protocol ID). FWIW, other LANs allow you just 2^7+2^39 different protocols... paul