[comp.dcom.lans] Report on Localnet 86

phil@amdcad.UUCP (Phil Ngai) (11/19/86)

Here's a trip report which I hope will be of interest. If not, let
me know and I won't post such things in the future.

***************************************************************************

Copyright 1986 Advanced Micro Devices			11/19/86
							Phil Ngai

This report is the opinion of the author only, not his company.

LOCALNET 86 Trip Report

I went to Localnet 86 in San Francisco on Nov 18, in the afternoon and
talked to the vendors there to see what new products they had or were
coming out with. My interest was strictly in IP/TCP 802.3 type
products. There was a lot of broadband, etc which I skipped past.  In
general the vendors seem very unknowledgable as to the offerings of
their competition. I consider this a major failing. Many vendors
further show little thought as to how a customer might use their
product or evidence of talking to real customers. Sad, very sad.

Here are some brief notes.

ANIXTER

Raychem was demonstrating their EZF 75 ohm connector. Very simple and
quick to install. Now, they ought to broaden their line to support
other types, particularly thin-wire Ethernet cable.

BICC

Claim their multiport does SQE correctly. When a station transmits, a
flip flop is set to enable the return of SQE. After the end of
transmission, a timer starts up to cover the interpacket gap.  When
the timer expires, the flip flop is cleared, disabling the passage of
SQE to that port on the multiport.

Finally, someone that does it right! Also, they are very careful to
make their AUI connector comply with all the 802.3 specs.

BRIDGE

I didn't notice anything new at the show.

CABLETRON

Has come out with a multiport transceiver. As far as I can tell (the
salesmen are not well trained) it does SQE the DEC DELNI way and
stations which did not transmit get the SQE back whenever anyone on
that multiport transmits if the transceiver is set up to generate SQE.

I think this is wrong.

CISCO

Was notable only by their absence.

CODEX

Poor Codex has a terminal server which does not talk IP/TCP.  They
seem to be quite clueless as to the importance of standards. I was
so disgusted I didn't even look at the 4320 LAN Hub, which is
apparently another multiport transceiver.

CMC

They have a bridge product but don't really know how it compares to
DEC's LAN Bridge 100. From what I gathered, it has the same
functionality but is slower. You would expect something in exchange
for the loss of performance over a LAN Bridge 100 but you don't get
anything like programmable packet filters. It is a little cheaper. The
salesman was not very well trained.

They have a terminal server. I asked "How does it know host names?"
"You program the box with a terminal." "Each box?!" "Sure." "What if I
have 50 boxes and 600 terminals, isn't that a lot of work?"  "Well, we
hadn't thought about that. But don't people call machines by different
names? How could you ever get people to agree on what the name of a
computer was?"

EXCELAN

Says they have real networking for IBM PC-RTs running AIX, $1295 list.
Telnet, ftp, etc. If it is like their other products (I forgot to ask)
they don't do Berkeley routed.

HP

The 4971S protocol analyzer still isn't portable. Nor does it use a
standard host. I saw a lot of Compaq portable PCs at the show but not
in HP's booth.

MICOM-Interlan

Finally have a IP/TCP terminal server. Network management does not
seem to be well thought out yet but they seemed to listen, if in a
distracted way. They asked "why do you want routed or EGP, you can
just program the gateways by hand". I also asked for end to end
testing, they allowed as that might be useful. They do have a nice
display of TCP status per port.

PROTEON 

was there but all I saw was rings so I didn't talk to them.

UNGERMANN-BASS

Much more knowledgable about large networks than MICOM. They actually
have EGP in their routers. It is really stupid that they call their
routers "remote bridges". The guy I talked to was interested in
whether there was a need for routed as well as EGP. At least they've
thought about it and are asking customers. They also have a network
management station with a download server for their router and all
those things that Bridge Communications doesn't do.

A problem with their router is unsophisticated serial line management.
Actually, they only let you use one serial line between routers right
now. Based on the long lines performane we've seen, this could not
provide acceptable network availability.

VITALINK

Still doesn't do serial lines right. We'd like to rent two 56,000 bps
long lines and get 112,000 bps of throughput moving a file from one
station to another. They can't do this.  They only allow any one
connection to go over one long line. By connection, I assume they mean
from one Ethernet physical address to another. If that long line goes
down, the people on it get dropped. New connections will be over lines
that are still up.

Their rational is that out of sequence packets are unacceptable. For
our environment, it is acceptable, but they have a larger application
base to serve, like LAT. (does that stand for Local Access Terminals?)

They are experimenting with an Amiga as a network management station.
They have the right idea with multi-tasking, I think. But why not
a Compaq 386 with Xenix? Or VAX Unix binaries?
-- 
 The distance from the North end of Vietnam to the South end is about
 the same as the distance from New York to Florida.

 Phil Ngai +1 408 749 5720
 UUCP: {ucbvax,decwrl,hplabs,allegra}!amdcad!phil
 ARPA: amdcad!phil@decwrl.dec.com