www.a00.de > tcpgroup > 1991 > msg00070
 

TCP-group 1991


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

MAC sublayer protocol



Fred Goldstein k1io, replies to a message from Steven L. Johnson (me).

>Again it's a question of who does what to whom where.  An access channel
>needs a MAC protocol; a backbone is normally point to point and doesn't.

I'm sorry, I appeared to have misused the term 'backbone'.  You are
equating 'backbone' with a separate set of frequencies on fdx or hdx
point to point links between dedicated sites.  The term I should have
used is perhaps major routing node, for a site that is available for
forwarding traffic a significant percentage of the time.  At this
time, I was not making assumptions as to pt-pt links or dedicated
frequency usage of these routing nodes, and considered a pt-omni
connect between and end node and a routing node a particularly inter-
esting configuration.  A group of these routing nodes within a geographic
area might be able to supplement, replace or provide redundant routing
for a dedicated backbone.

>Direct 56k user access to backbones is an oxymoron unless the user and
>the backbone each have a dedicated link radio to one another, making the
>user in effect a backbone node.  The problem we're trying to solve is
>ordinary shlubs like me, who don't have dedicated radios, and just run
>packet once in a while, not 7x24.

Again, I misspoke.  I meant 56k user access to a backbone site (or
major routing node).  I don't understand why 56k would require a
dedicated pt-pt link, and I assume that said dedicated because I said
backbone.  I would think it highly desireable if an end-node user (or
even a routing hub itself) could go up and down at will without severly
impacting the network.

When you mention 'problem/ordinary-shlubs', I'm assuming you are refering
to wide area network connectivity for low-cost low-speed users (1200b),
and perhaps 2.4k and 9.6k as well.  And certainly, this is an important
service as a large number (majority!) of end nodes will continue using
this hardware.  However, on a selfish note, what tickles me would be
intermittant access to a higher speed channel available on UHF/MW; and
I believe this may involve a different set of problems and opportunities.

>CSMA means that everybody hears the channel before transmitting, not
>while transmitting.  CSMA/CD means everyone listens _while_ transmitting
>too, and can tell if they've been clobbered.  That's VERY tough with
>radio; you'd have to compare received with transmitted bits at 2x the
>repeater path delay.

I may be missing something, but beyond the fdx hardware, I don't see
the difficulty.  Unless transmitting very short frames (less than
2x path delay), the actual delay shouldn't matter.  After one starts
transmitting, the next packet received better match up byte for byte
with that transmitted or whammo assume collision.  Different than
analog collision detection used on ethernet networks, but seems
doable.

So it would seem that if you have a full duplex repeater you might
as well try CSMA/CD (if the leaf node is fdx) for added effeciency.
If BTMA, then as you pointed out you still don't have any CD, but
simply have improved a pseudo-CSMA to CSMA since the BT should
eliminate hidden terminals and guarantee carrier sense capabilities
on the channel.  However on a simplex channel with hidden terminals,
then one is still muddling between a pure Aloha and a pseudo-CSMA
channel, and I'm stuck wondering if there isn't a better way.

>Ham radio doesn't always behave rationally.

Not much does.

-Steve





Document URL : http://www.a00.de/tcpgroup/1991/msg00070.php
Ralf D. Kloth, Ludwigsburg, DE (QRQ.software). < hostmaster at a00.de > [don't send spam]
Created 2004-12-21. Last modified 2004-12-21. Your visit 2020-10-24 23:03.18. Page created in 0.0587 sec.
 
[Go to the top of this page]   [... to the index page]