eriks@yunexus.yorku.ca (Eriks Rugelis) (12/31/89)
I'm having a friendly disagreement with a colleague about a particular
aspect of 802.3 network configuration. Now, before I go on, we are not
really arguing the 'will it work' aspect but rather 'what did the designers
of the Standard intend?' (if the answer contains a description of 'why
the designers thought so', we wouldn't turn it down...)
So, onward:
One of us (not me ;-), is of the opinion that 802.3 does not always
require a repeater at both ends of a FOIRL and that if there is but a
single device at the end of a FOIRL the repeater is superfluous. By
extension, this argument says that a degenerate network could consist
of nothing but two nodes, two MAU's and the FO cable between them,
without ANY repeaters at all. But then you can start throwing in some
curves such as: Is a node attached to a DELNI attached to the FOMAU
without a repeater STILL OK? Where does one draw the line?
In a picture, is the following legal (no repeaters)?
AUI cable FO cable AUI cable
Node-----------FOMAU================FOMAU-----------Node
If it IS legal, at what point does the Standard *require* repeaters?
Hello, Pat Thaler? You've 'been there' for much of 802.3 development process.
Can you take some time again to educate the rest of us?
Thanks,
Eriks
--
---
Voice: Eriks Rugelis E-Mail: eriks@outland.yorku.ca UUCP: yunexus!eriks
Ma Bell: 416/736-5257 x.22688 Last resort: eriks@yulibra.NETNORTH
Snail: York University, 4700 Keele St., North York, Ontario, Canada. M3J 1P3pat@hprnd.HP.COM (Pat Thaler) (01/04/90)
Eriks Rugelis writes: > I'm having a friendly disagreement with a colleague about a particular > aspect of 802.3 network configuration. Now, before I go on, we are not > really arguing the 'will it work' aspect but rather 'what did the designers > of the Standard intend?' (if the answer contains a description of 'why > the designers thought so', we wouldn't turn it down...) > > So, onward: > > One of us (not me ;-), is of the opinion that 802.3 does not always > require a repeater at both ends of a FOIRL and that if there is but a > single device at the end of a FOIRL the repeater is superfluous. By > extension, this argument says that a degenerate network could consist > of nothing but two nodes, two MAU's and the FO cable between them, > without ANY repeaters at all. But then you can start throwing in some > curves such as: Is a node attached to a DELNI attached to the FOMAU > without a repeater STILL OK? Where does one draw the line? What follows, except for quotes from 802.3, is my personal opinion. Not necessarily that of IEEE or any committee. There are a number of questions here. First does the FOIRL spec require a repeater to be connected at each end? From 802.3d-1987 9.9 Medium Attachement Unit and Baseband Medium Specification for a Vendor Independent Fiber Optic Inter Repeater Link. (the FOIRL section of 802.3) 9.9.1.1 Overview. A vendor independent FOIRL provides a standard means for connecting only two repeater units. .... The FOMAU described in this document is not intended for use in connecting DTEs. That is not quite as strong as if they had said "The FOMAU shall not be directly connected to a DTE," but it gives you the intent. Of course, 9.9.1.4 also says that SQE test shall not be implemented. (A repeater MAU does not do SQE test and a DTE MAU does SQE test.) Note also that the FOIRL is defined as a subsection of the Section 9 which defines repeaters. Other MAUs have their own section. "Require" would perhaps be a bit strong, but 802.3 clearly "intended" a repeater at both ends. Since SQE test was not included in the FOIRL, one could argue for "require." Second, what if you use a MAU mux type device such as a DELNI? 802.3 does not say anything about MAU muxes. A MAU mux is a device which allows multiple DTEs to share a single MAU. 802.3 does not specify such a device. If you use one, you are outside of the scope of the standard. Several manufacturers produce MAU muxes. The idea is to produce such a device in a way that it does not affect the quality of the signal passing through it (or at least doesn't affect it any more than passing through 10 m or so of AUI cable). The small sampling that I have looked at fall somewhat short of this goal in the area of jitter. (Total jitter budget from the AUI connector on one DTE or repeater to the next is 18 ns and the total amount of jitter allowed for 50 m of AUI cable including SNR is 1.5ns. So jitter has to be held very low for a MAU mux to be no worse than 10 m of AUI.) > > In a picture, is the following legal (no repeaters)? > > AUI cable FO cable AUI cable > Node-----------FOMAU================FOMAU-----------Node > > If it IS legal, at what point does the Standard *require* repeaters? What do you mean by legal? No one is going to arrest you for using this configuration -). (Sorry, I couldn't resist.) Seriously, I would say the above usage of a FOMAU would not be what the FOIRL section intends. If the FOMAUs above provide the ability to turn on SQE test, I know of no technical reason why the above configuration would not work. Why then were the FOMAU and FOIRL defined as specifically for interconnecting repeaters? (The following is somewhat simplistic. The path wasn't quite as clear and neat as it implies.) There was a need for a FOIRL to be defined as soon as possible. FOIRL was included in the system topology described in the original 802.3 standard and people needed it to connect between buildings in those cases where coax wouldn't do. There was a clear way to define a FOIRL which worked. There was also interest in a fiber optic 802.3 network. There were a number of candidate technologies for such a network and there was no concensus on which was best overall. One of the candidates was the technology used in FOIRL. It was decided to produce the FOIRL standard while leaving open the question of which technology should be used for a fiber optic 802.3, so as to meet the immediate need in a reasonable time. Once the FOIRL standard was nearly complete, work was started on choosing a fiber optic 802.3 technology. About a year ago, concensus was reached to develop a draft which includes both an active star, 10BASE-FA, and a passive star, 10BASE-FP. Work on the draft is proceeding. > > Hello, Pat Thaler? You've 'been there' for much of 802.3 development process. > Can you take some time again to educate the rest of us? > > Thanks, > Eriks > -- > --- > Voice: Eriks Rugelis E-Mail: eriks@outland.yorku.ca UUCP: yunexus!eriks > Ma Bell: 416/736-5257 x.22688 Last resort: eriks@yulibra.NETNORTH > Snail: York University, 4700 Keele St., North York, Ontario, Canada. M3J 1P3 > ---------- Your welcome, Pat Thaler