fmbutt@mrbt.sw.stratus.com (Farooq Butt) (12/14/90)
Net.TCP/IPGurus:
(I'll save net.bandwidth and keep it brief...)
Is IEEE 802.3 the very same thing as the "Ethernet" standard ? In what
ways are the two different ? I know that 802.3 is usually implemented
at a hardware level, what about "Ethernet" ? Which one is older ?
Does using one seriously preclude you from connecting with machines
using the other flavor ? Finally which is "better" ;-) ?
thanks for your input,
-fmb
--
Hi-Tech Disclaimer: NOTHING in the above article has any relationship
to reality. If any reality correspondences are found,
please notify me IMMEDIATELY. Any threats or abuse
of any kind is purely unintentional. My employer is not liable. mjhammel@Kepler.dell.com (Michael J. Hammel) (12/15/90)
In article <3443@lectroid.sw.stratus.com>, fmbutt@mrbt.sw.stratus.com (Farooq Butt) writes: > Is IEEE 802.3 the very same thing as the "Ethernet" standard ? In what > ways are the two different ? I know that 802.3 is usually implemented > at a hardware level, what about "Ethernet" ? Which one is older ? Version 2.0 of Ethernet was published in November 1982. The first edition of 802.3 was published in early 1985. > Does using one seriously preclude you from connecting with machines > using the other flavor ? Finally which is "better" ;-) ? They can coexist on the same wire but unless your physical interface is sophisticated enough to recongize both types at least some of the data will not get through. Which is better is probably a religous question. ;-) They are not the same, but they are somewhat similar. The Ethernet frame format looks like this: 8 octet preamble 6 octet destination address 6 octet source address 2 octet type 46-1500 octets of data 4 octet CRC The 802.3 frame format looks like this: 7 octet preamble 1 octet start frame delimiter 2 or 6 octet destination address 2 or 6 octet source address 2 octet length 46-1500 octets data and pad (in 10BASE5 definition) 4 octet CRC The preamble and SFD of 802.3 are basically the same as the preamble of ethernet. The 10BASE5 version of 802.3 specifies use of 6 octet addresses so these are the same as ethernet. The type field of ethernet is, I've been told, always less than some number and the length field of 802.3 is always greater than that number. This *might* be a way to tell the two frame types apart, but don't count on it. Both formats require frame sizes between 64-1518 octets total (excluding preamble). In ethernet the client layer must guarantee this size. In 802.3 the 802.2 LLC can hand off invalid size frames to the 802.3 MAC layer and the MAC layer will compensate (at least it will if the LLC provides a frame thats too short; I'm not sure what will happen if its too long). I got all this info from an article by Gary Kessler in the July 1987 issue of LAN Magazine. Its a very good article if you'd like to read more about the specific differences in the two formats. Michael J. Hammel | mjhammel@{Kepler|socrates}.dell.com Dell Computer Corp. | {73377.3467|76424.3024}@compuserve.com #include <disclaim/std> | zzham@ttuvm1.bitnet | uunet!uudell!feynman!mjhammel #define CUTESAYING "Your cute quote here"
gaj@hpctdja.HP.COM (Gordon Jensen) (12/15/90)
>Is IEEE 802.3 the very same thing as the "Ethernet" standard ? In what >ways are the two different ? I know that 802.3 is usually implemented > >at a hardware level, what about "Ethernet" ? Which one is older ? >Does using one seriously preclude you from connecting with machines >using the other flavor ? Finally which is "better" ;-) ? Ethernet and 802.3 are used more or less interchangably these days. Ethernet was the original network developed back at Xerox ages ago. IEEE made some minor changes - redefined the type field as length, AC coupled the tranceivers, some other minor stuff. I have seen old Ethernet type 1 MAUs oscillate with the inductance of the isolation transformer in 802.3/Ethernet type 2 nodes. You shouldn't see any problems unless you have some really ancient equipment, since virtually everything that I know of that has been made in the last six years conforms(more or less) to the IEEE spec. The only real difference is the definition of the type field. IEEE uses it for length. Fortunately there isn't a lot of overlap so you can often figure out what kind of a frame you've got. You can run both on the same cable, but your software may have some problems.
koning@koning.enet.dec.com (Paul Koning) (12/18/90)
|> |>They can coexist on the same wire but unless your physical interface is |>sophisticated enough to recongize both types at least some of the data |>will not get through. Which is better is probably a religous question. ;-) The sophistication needed is trivial. |>They are not the same, but they are somewhat similar. |> |>The Ethernet frame format looks like this: |> 8 octet preamble |> 6 octet destination address |> 6 octet source address |> 2 octet type |> 46-1500 octets of data |> 4 octet CRC |> |>The 802.3 frame format looks like this: |> 7 octet preamble |> 1 octet start frame delimiter Same as Ethernet (Ethernet doesn't have separate names for the two parts, but the definition is IDENTICAL. |> 2 or 6 octet destination address |> 2 or 6 octet source address No, only 6 octet addresses are allowed. See section 4.4.2.1. |> 2 octet length |> 46-1500 octets data and pad (in 10BASE5 definition) Not just 10Base5; it applies to all 10 MHz versions. |> 4 octet CRC |> |>The preamble and SFD of 802.3 are basically the same as the preamble of ^^^^^^^^^ They ARE the same; only the words used to describe them differ. |>ethernet. The 10BASE5 version of 802.3 specifies use of 6 octet ^^^^^^^ Actually, then 10 MHz version in general (including 10Base2, a.k.a. "ThinWire", a.k.a. "cheapernet". |>addresses so these are the same as ethernet. The type field of ethernet |>is, I've been told, always less than some number and the length field of |>802.3 is always greater than that number. This *might* be a way to tell |>the two frame types apart, but don't count on it. Both formats require It's actually the other way around: 802.3 Length is in the range 0-1500, while Ethernet ProtocolType has the larger values. And why be so tentative about this? You certainly CAN count on it; it works; it IS the way to tell the two apart. paul
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (12/18/90)
In article <13213@uudell.dell.com>, mjhammel@Kepler.dell.com (Michael J. Hammel) writes: > ...[discussion about difference between and mixing ethernet and 802.3 > deleted]... RFC-1122 as well as RFC-1042 and 1188 have good stuff about these subjects. Vernon Schryver, vjs@sgi.com
burdick@hpspdra.HP.COM (Matt Burdick) (12/19/90)
> It's actually the other way around: 802.3 Length is in the range > 0-1500, while Ethernet ProtocolType has the larger values. And why be > so tentative about this? You certainly CAN count on it; it works; it > IS the way to tell the two apart. With one exception: the Xerox PUP ethernet type lies within the 802.3 length range (it's 0200). Fortunately, it isn't used much anymore, so you probably don't have to worry about it. -matt -- Matt Burdick | Hewlett-Packard burdick@hpspd.spd.hp.com | Intelligent Networks Operation
mjhammel@Kepler.dell.com (Michael J. Hammel) (12/20/90)
In article <12710012@hpspdra.HP.COM>, burdick@hpspdra.HP.COM (Matt Burdick) writes: > > It's actually the other way around: 802.3 Length is in the range > > 0-1500, while Ethernet ProtocolType has the larger values. And why be > > so tentative about this? You certainly CAN count on it; it works; it > > IS the way to tell the two apart. > > With one exception: the Xerox PUP ethernet type lies within the 802.3 > length range (it's 0200). Fortunately, it isn't used much anymore, so > you probably don't have to worry about it. Someone else mentioned this to me but said that Xerox had changed (or was going to change, maybe) the type value to one that didn't conflict. Do you know if this is true? Michael J. Hammel | mjhammel@{Kepler|socrates}.dell.com Dell Computer Corp. | {73377.3467|76424.3024}@compuserve.com #include <disclaim/std> | zzham@ttuvm1.bitnet | uunet!uudell!feynman!mjhammel #define CUTESAYING "Your cute quote here"
donp@na.excelan.com (don provan) (12/20/90)
In article <13213@uudell.dell.com> mjhammel@Kepler.dell.com (Michael J. Hammel) writes: >The type field of ethernet >is, I've been told, always less than some number and the length field of >802.3 is always greater than that number. This is backwards, probably just a typo. The maximum legal 802.3 length is 1500. Anything value larger than that is an ethernet type. > This *might* be a way to tell >the two frame types apart, but don't count on it. On the contrary, you can count on it because it's in the 802.3 spec. (Well, sort of: the spec has a famous footnote which says that packet lengths larger than the maximum 802.3 length have some unspecified local meaning. I've yet to hear of an 802.3 network where that "local meaning" was something other than "ethernet type".) The only "illegal" ethernet types i've ever heard of were a couple of ancient Xerox PUP values that fell below the 1500 value. Both of them were reallocated legal values long ago, so i wouldn't expect to see any software using the old values. Actually, i'm not sure i'd expect to see any software using the new values, either, since i don't believe PUP is used much any more. don provan donp@novell.com
masinter@parc.xerox.com (Larry Masinter) (12/29/90)
>> > It's actually the other way around: 802.3 Length is in the range >> > 0-1500, while Ethernet ProtocolType has the larger values. And why be >> > so tentative about this? You certainly CAN count on it; it works; it >> > IS the way to tell the two apart. >> >> With one exception: the Xerox PUP ethernet type lies within the 802.3 >> length range (it's 0200). Fortunately, it isn't used much anymore, so >> you probably don't have to worry about it. > Someone else mentioned this to me but said that Xerox had changed (or > was going to change, maybe) the type value to one that didn't conflict. > Do you know if this is true? ================================================================ >The only "illegal" ethernet types i've ever heard of were a couple of >ancient Xerox PUP values that fell below the 1500 value. Both of them >were reallocated legal values long ago, so i wouldn't expect to see >any software using the old values. Actually, i'm not sure i'd expect >to see any software using the new values, either, since i don't >believe PUP is used much any more. Well, at the home of th PARC Universal Packet, there are still a few machines around that talk PUP. Officially reallocating packet types doesn't help much when you are dealing with software that you couldn't modify if you wanted to (who knows BCPL any more?)... The only product that shipped in any numbers and had PUP support was the Lisp machine, and it was given some kind of configuration switch/patch to switch the pup type to the reallocated new value. I'd guess that happened about '85. However, there's no telling if anybody bothered to reconfigure. I'm planning to bring up a netware server here, and it looks like we'll have to isolate it from the pup traffic by an appropriate router; at least the rumors are that the netware servers will not coexist. That's a nuisance, of course... -- Larry Masinter (masinter@parc.xerox.com) Xerox Palo Alto Research Center (PARC) 3333 Coyote Hill Road; Palo Alto, CA USA 94304 Fax: (415) 494-4333