morgan@jessica.stanford.edu (RL "Bob" Morgan) (04/18/89)
I'm puzzled by various references I've seen recently, in this forum and elsewhere, to the operation of the proposed 10BaseT Ethernet. Is the draft spec available from anywhere, either on paper or (better) on-line? Is there some nice, descriptive article somewhere that talks about 10BaseT operation and terminology? When the battle to define the standard was going on last year, I thought there was a clear split between the "3Com approach" and the "Lattisnet approach." The 3Com method involves simply modifying the standard Ethernet transmission to work acceptably over twisted pair, more or less like a Balun does for other coax-to-TP purposes. In this case the TP is really just an Ethernet cable, the transceiver functions are performed at the station, and the machinery in the wiring closet is just a multiport repeater with one Balun per port. The Lattisnet method, on the other hand, uses the TP as an extension of the transceiver cable, so that most (some? all?) transceiver functions are performed by the wiring-closet machinery, which is more-or-less a multiport transceiver. The device at the station is required to multiplex the signals that are normally carried on the multiple-conductor transceiver cable so they'll fit on the TP, but isn't really a transceiver in the normal sense. There is no repeater in the normal sense (auto-partitioning, retiming, etc) at all. Of course, the Lattisnet approach "won" the standards battle. Now, I've seen people referring to the 10BaseT wiring closet device as a repeater. Is this just loose terminology or is my understanding of 10BaseT/Lattisnet wrong? Thanks, - RL "Bob" Morgan Networking Systems Stanford
kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England)) (04/19/89)
In article <1627@Portia.Stanford.EDU> morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: > >I'm puzzled by various references I've seen recently, in this forum >and elsewhere, to the operation of the proposed 10BaseT Ethernet. Is >the draft spec available from anywhere, either on paper or (better) >on-line? Is there some nice, descriptive article somewhere that talks >about 10BaseT operation and terminology? > I have some of the slide presentations from some of the meetings from last year when the spec was still very much up in the air. Nothing of late, but they are arguing about "link state" and type of equalization at this point. Actually, I understand that there has been another meeting and the 10BaseT has a proposal for 802 to consider, so all those issues are resolved by now. They probably ended up by leveling the playing field (ie, break all existing implementations :-) But, hey, I don't want to step on any mines here. >When the battle to define the standard was going on last year, I >thought there was a clear split between the "3Com approach" and the >"Lattisnet approach." Well, it was more like the DEC/3Com camp versus the "concentrator" camp, of which LattisNet was but one approach. Two main camps with subcamps if you will. :-) >The 3Com method involves simply modifying the >standard Ethernet transmission to work acceptably over twisted pair, >more or less like a Balun does for other coax-to-TP purposes. In this >case the TP is really just an Ethernet cable, the transceiver >functions are performed at the station, and the machinery in the >wiring closet is just a multiport repeater with one Balun per port. That's right, but with attendant loss of capability. Ethernet was not designed for unshielded media like TP. The committee decided not to go with an inferior technology, nor to cop out and adopt both standards. Good for them. (My opinion, no flames.) >The Lattisnet method, on the other hand, uses the TP as an extension >of the transceiver cable, so that most (some? all?) transceiver >functions are performed by the wiring-closet machinery, which is >more-or-less a multiport transceiver. The device at the station is >required to multiplex the signals that are normally carried on the >multiple-conductor transceiver cable so they'll fit on the TP, but >isn't really a transceiver in the normal sense. There is no repeater >in the normal sense (auto-partitioning, retiming, etc) at all. > LattisNet is kinda like a transceiver, but the "other" approach is very much a repeater approach. The twisted pair are basically transmit and receive pairs, with the concentrator and the transceiver doing the collision detect and CD signalling. The repeater approach makes 10BaseT more consistent with other IEEE media and gives you some of the fault isolation of IEEE-spec bit-level repeaters. Again, the right choice in my opinion. >Of course, the Lattisnet approach "won" the standards battle. No, I think they lost. The "repeater camp" won. >Now, >I've seen people referring to the 10BaseT wiring closet device as a >repeater. Is this just loose terminology or is my understanding of >10BaseT/Lattisnet wrong? > Just a little dated, I think. Of course, LattisNet still works and Synoptics may go ahead and build a 10BaseT option. The real test of the 10BaseT is when we have some vendor interoperability. A little more time required. Disclaimer: I am an interested party, but my information may be incorrect or dated in details. Please don't take my opinion comments as fact. Vendors welcome to speak for themselves. --Kent England, Boston U
kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England)) (04/20/89)
In article <17142@cos.com> smith@cos.UUCP (Steve Smith) writes: >In article <1627@Portia.Stanford.EDU> morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: >> >>I've seen people referring to the 10BaseT wiring closet device as a >>repeater. Is this just loose terminology or is my understanding of >>10BaseT/Lattisnet wrong? > >I think the tranceiver vs repeater problem is more a matter of >terminology than anything else. The TP is the "transmission medium" of >the network, even though it can have only one device on each end. In >other words, each "network segment" can have only two devices on it. >(If you're wondering, yes, you can hook two DTEs together directly.) >Since the wire closet doodad connects multiple segments, it is a >"repeater" in 802.3 terminology. > It's a little more than terminology. LattisNet is more like a multiport transceiver (maybe it *is* exactly a MPT) while 10BaseT is a multiport repeater. It does have some ramifications. For example, 10BaseT can isolate faults to the port. LattisNet isolates to the board, or multi-port. SynOptics says that they will be compatible on the board level, ie mix LattisNet and SynOptics 10BaseT in the same chassis. SynOptics graciously offered LattisNet to the committee, but they opted for a format more compatible with other IEEE specs, for example the fiber optic spec, FOIRL. And that is why two TP xcvrs should plug together and work like a sex-reversed repeater. Not sure they will, but they might.
mkd@mtunh.ATT.COM (Mark Darby) (04/25/89)
In article <1627@Portia.Stanford.EDU>, morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: > > I'm puzzled by various references I've seen recently, in this forum > and elsewhere, to the operation of the proposed 10BaseT Ethernet. Is > the draft spec available from anywhere, either on paper or (better) > on-line? Is there some nice, descriptive article somewhere that talks > about 10BaseT operation and terminology? > > When the battle to define the standard was going on last year, I > thought there was a clear split between the "3Com approach" and the > "Lattisnet approach." The 3Com method involves simply modifying the > standard Ethernet transmission to work acceptably over twisted pair, > more or less like a Balun does for other coax-to-TP purposes. In this > case the TP is really just an Ethernet cable, the transceiver > functions are performed at the station, and the machinery in the > wiring closet is just a multiport repeater with one Balun per port. > The Lattisnet method, on the other hand, uses the TP as an extension > of the transceiver cable, so that most (some? all?) transceiver > functions are performed by the wiring-closet machinery, which is > more-or-less a multiport transceiver. The device at the station is > required to multiplex the signals that are normally carried on the > multiple-conductor transceiver cable so they'll fit on the TP, but > isn't really a transceiver in the normal sense. There is no repeater > in the normal sense (auto-partitioning, retiming, etc) at all. > > Of course, the Lattisnet approach "won" the standards battle. Now, > I've seen people referring to the 10BaseT wiring closet device as a > repeater. Is this just loose terminology or is my understanding of > 10BaseT/Lattisnet wrong? > > Thanks, > > - RL "Bob" Morgan > Networking Systems > Stanford I've been involved with the 10BaseT task force since I began work with AT&T in August '87. I'll try to answer some of your questions here the best that I can. First and foremost, the "Lattisnet" proposal did not win acceptance in the 10BASET task force. It wasn't even one of the three remaining contenders in the proposal race before the final proposal won out. The three final proposals were from Hewlett Packard, 3Com and David Systems. A note that the final offering from 3Com was a joint effort with DEC. The 3Com proposal is as you have mentioned. The proposal was withdrawn because technical feasibility couldn't be shown. Because the signal involved in their proposal (using a BALUN) is DC coupled and with a data content that is much lower in level it is a great source of interference to other services and likewise is much more susceptible to interference from other services. The David Systems proposal is one I admittedly know little of. I do know that it was based on a two-pair system, each link connecting a MAU to a modified multiport repeater. The MAU in question is asymmetric; that is the transmit link was based on Manchester signaling, the receive link was based on modified Manchester signaling. David Systems likewise withdrew their proposal, but I am unaware of the reason (anyone who can expand on this, please do so). The Hewlett Packard was the first proposal to be presented to the 10BASET task force. It was also the proposal which was finally accepted by the 10BASET task force after 3Com and David Systems withdrew their proposals. The HP proposal is based on specifying the MAU only. Each link is 2-pair from MAU to MAU, each MAU having an AUI interface. The central point of a 10BASET network for this proposal is assumed to be a multiport repeater (MPR), an entity which is already standardized in IEEE (see Chapter 9, IEEE 802.3 Supplement Book). Signaling is Manchester in both directions. There are fundamental differences between the HP and SynOptics proposals. The SynOptics proposal has a network architecture based on a MAU to active concentrator link. The active concentrator is not an IEEE repeater in any sense in that there is no retiming, no fragment extension, no auto-partition capabilities at the central connection point which would maintain network reliability. SynOptics LOST on their proposal because of this. While the MPR specification has the Auto-partition algorithm as an optional feature, 10BASET has mandated that the algorithm be mandatory on a per port basis. Just a note that, while the SynOptics proposal didn't win in the standards body, SynOptics has an MPR in their product line available now. The extent of its functionality is not known, however. There are also some operational differences, especially in the areas of transmit voltage levels and equalization techniques. The differences are substantial enough to warrant increased scrutiny when attempting to bundle equipment which is not even close to the proposed 10BASET draft with other services, including equipment which IS close to the 10BASET proposed specifications. The draft is now in letter ballot form. The document is P802.3I/D6, it is currently being voted upon by the 802.3 working group, and the cut off date is May 19, 1989. It's expected that 10BASET will meet in June to resolve any NO votes from the balloting. Try to find a voting member of 802.3 to get a copy, or you can obtain a copy from IEEE. I hope this helps! --------------------------------------------------------------------- Mark K. Darby AT&T Bell Laboratories AT&T: (201)957-2706 200 Laurel Ave. uucp:..!att!mtunh!mkd Middletown, NJ 07748 DISCLAIMER: The above is a statement of my own opinions and does not reflect those of the corporation for which I work.
pat@hprnd.HP.COM (Pat Thaler) (05/27/89)
> / hprnd:comp.dcom.lans / smith@cos.com (Steve Smith) / 3:32 pm Apr 19, 1989 / > In article <1627@Portia.Stanford.EDU> morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: > > > >I'm puzzled by various references I've seen recently, in this forum > >and elsewhere, to the operation of the proposed 10BaseT Ethernet. Is > >the draft spec available from anywhere, either on paper or (better) > >on-line? > > The 10BASET standard is currently out for ballot to the IEEE 802.3 > committee. If it sails through everything, it might be a standard by > this summer (from past experience, don't bet on it). However, the spec > itself probably won't change much if at all. > Actually, the 10BASE-T draft will need to be ballotted by the Technical Committee on Computer Communications (TCCC) after it passes IEEE 802.3 working group ballot. Then it goes to the next IEEE Standards Board meeting for approval. Therefore, the earliest date at which it could be a standard is January 1990. The 10BASE-T task force will be meeting June 19-21 and at the July 802 plenary to resolve ballot comments. > You should be able to get copies from the Chair of the 10BASET task > force: > > Pat Thaler > Hewlett Packard > 8000 Foothils Blvd. R3NF2 > Roseville, CA 95678 > (916) 785-5238 > Remember this is a draft for comment and is subject to change. If you just want the draft, contact my secretary, Suzanne Palacios. The phone number above is actually hers. My phone number is (916) 785-4538. > >I've seen people referring to the 10BaseT wiring closet device as a > >repeater. Is this just loose terminology or is my understanding of > >10BaseT/Lattisnet wrong? It isn't loose terminology. The 10BASE-T draft uses the repeater defined in Section 9 of the 802.3 supplements to connect 10BASE-T segments together. (If you have trouble finding Section 9, the order of sections in the supplement is 10, 11, 9, 12. That's because they ordered the sections by supplement number rather than by section number.) 10BASE-T is being developed by the 10BASE-T task force and involves the contributions of people from many companies. It was developed to meet a list of objectives identified at the outset. It is not 10BASE-T/(any proprietary product). > > - RL "Bob" Morgan > > Networking Systems > > Stanford > > -- Steve > (smith@cos.com) ({uunet sundc decuac hqda-ai hadron}!cos!smith) > "Truth is stranger than fiction because fiction has to make sense." > ---------- Pat Thaler Hewlett Packard