Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!mips!apple!voder!berlioz.nsc.com!falcon.nsc.com!my
From: my@falcon.nsc.com (Michael Yip)
Newsgroups: comp.dcom.lans
Subject: Re: Info on FDDI SMT
Keywords: FDDI, SMT, X3T9.5
Message-ID: <1991Apr14.204102.24156@berlioz.nsc.com>
Date: 14 Apr 91 20:41:02 GMT
References: <97250@sgi.sgi.com> <97028@sgi.sgi.com> <1991Apr11.221054.25188@leland.Stanford.EDU> <1991Apr12.233143.722@berlioz.nsc.com>
Sender: news@berlioz.nsc.com
Reply-To: my@falcon.nsc.com (Michael Yip)
Organization: National Semiconductor Corp
Lines: 50


> PMF's etc would have been a keen idea 5 years ago, before SNMP, before the
> market decide INTERnetwork management was possible and desirable.  At this
> late date, it is strange to delay SMT to perfect a network management
> scheme that only works on the local ring not even crossing bridges, to
> believe that TCP is hard in these days of free 4.3BSD and KA9Q source and
> dirt cheap MB RAMs and Mbit PROMS, to mandate a link layer management that
> needs a transport with most of the machinery of UDP/IP (the worst case size
> of some of the frames is >4500bytes).  It's as if some in X3T9.5 have
> commercial death wishes.
Well some people believe in FDDI-2 and hope that FDDI-1 takes off later
so that FDDI-2 will be in better shape.  Some people also sell SMT software,
the harder and more complicate SMT gets to be ... the more copies that
they can sell.   I agree that UDP/IP is easy to get and we should just 
leave those functions to SNMP by expanding SNMP MIB.


> Instead of delaying to elaborate and fix PMF's, X3T9.5 should have finished
> SMT 3 years ago, and started a SMT2 with all of the bells and whistles.
> In the interum we could have seen how NIF's etc. really work, to find what
> fails in real life.
Well, just remember that it took a LONG time to define the NIF also ...
it was defined once and then state machines were added so that people
can understand how to implement NIF ... it's sad.  In general, I agree with
you and we should have a "SMT1" which is a stripped down version of SMT.


> The good news is that many customers and vendors are going ahead, treating
> the current standard like the Ethernet Blue book of old, and building,
> shipping, buying, and installing networks.
I sure hope that people go ahead with FDDI-1 and SMT (draft) ... otherwise
no one will buy FDDI chips from my company and my job will be gone.  ;)


> 
> How are people planning to use FDDI2?  (That's an honest question; I wasn't
> paying attention when it started.)
> FDDI is theoretically 10x ethernet.  Any pair of cheap 1990 workstations
> worth buying can staturate one ethernet (i.e. get around 1MByte/sec with
> ttcp).  How are you going to have any useful bandwidth left in FDDI2 if you
> reserve enough to do video?  Doesn't one video channel need >= 5MByte/sec?
Don't know what good FDDI-2 can do with video/multimedia ...  it is just
to slow for things like that.  And if FDDI-2/SMT-2 take another 2 years to
finalize then we might as well forget about FDDI-2 because applications 
demand a LOT more than 100Mbit/s then.


> Vernon Schryver,   vjs@sgi.com
-- Mike Yip
   my@berlioz.nsc.com