Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!nrl-cmf!ames!amdcad!sun!pitstop!sundc!seismo!uunet!cos!smith
From: smith@cos.com (Steve Smith)
Newsgroups: comp.dcom.lans
Subject: Re: CATV data networking question
Message-ID: <14881@cos.com>
Date: 8 Feb 89 23:18:37 GMT
References: <1422@ucsd.EDU> <27676@bu-cs.BU.EDU>
Reply-To: smith@cos.UUCP (Steve Smith)
Distribution: all
Organization: Corporation for Open Systems, McLean, VA
Lines: 72
Organization:

In article <27676@bu-cs.BU.EDU> kwe@buit13.bu.edu (Kent England) writes:
>
>	You might want to be careful about committing to the broadband
>token bus standard [802.4].  It may not go anywhere in terms of
>installed base.
>	Ungermann-Bass has dropped support for the broadband token bus
>medium.  They say it is due to lack of market interest.  They will be
>supporting MAP on Ethernet and, I suppose, other IEEE media standards.
>	U-B pioneered MAP support through INI, a joint venture with GE
>that GE bailed out of a while back.  U-B knows the MAP market through
>INI, so I would take their "advice" seriously and I question whether
>broadband token bus will ever reach a critical mass of support.  
>	Anyway, broadband MAP is looking rather deadend-like.  But
>perhaps broadband token bus has uses outside of MAP.  I would be
>interested to know what other users think and what other vendors
>already support 802.4.  Doesn't someone make a .3-.4 bridge already? 
>
>	Kent England, Boston University



There are a number of manufacturers of 802.4 stuff out there, and I get
no impression that they are going to dry up and blow away any time soon.


I have seen a couple of 802.3 - 802.4 bridges.  They all have the
problem that 802.4 allows packets up to 8192 data bytes.  If a long
packet is routed to an 802.3 network (max length 1500 data bytes), it is
simply discarded.  They should work fine for 802.3 to 802.4, though.


There are several reasons that I see for 802.4 not going anywhere.

 1.  802.4 is a very complex protocol.  It's only been relatively
recently that a single chip controller has been available.  Before this,
controllers were either custom chip sets or microcoded bitslices.
Expensive.

 2.  While one of the attractions of the broadband medium is that the
CATV components are inexpensive and widely available, the headends and
modems are *expensive*.  The modems by themselves insure that the cost
per connection will be at least $1000.

 3.  802.4, until the last year or so, has been a moving target.
Revisions to the spec have been depressingly frequent.  I believe that
the current spec is Version K.  This might be called "How to Scare Away
Chip Manufacturers".

 4.  802.4 requires a significant ammount of network management.  Unlike
Ethernet/802.3, where all of the "operating paramaters" are fixed, 802.4
has a whole bunch of paramaters that must be set in each station, and
they must all be the same across the network.  The results of screwing
up can get gruesome.

 5.  At least at one time, there were severe interoperability problems
between INI's 802.4 stuff and everybody else's.  I don't know what the
current status of this problem is.

 6.  *RUMOR ALERT* I have heard that, several years back, INI got a
contract to wire up a *big* factory with their 802.4 stuff.  Reportedly,
it never worked.  At all.  The company that owned the factory was *not*
impressed.  This gave 802.4 a really bad reputation.
     If anybody has any more info on this rumor, I would be interested
in hearing about it.


DISCLAIMER:  These are strictly my own opinions.  Any resemblence to
any real opinions, public or private, is strictly coincidental.
-- 
                -- Steve
(smith@cos.com)    ({uunet sundc decuac hqda-ai hadron}!cos!smith)
"Truth is stranger than fiction because fiction has to make sense."