                Using the Wisconsin Network - Part 39 
                      by Any Nemec, KB9ALN

     In this part of our series, we will take a look at a
different way to network in Packet Radio - Flexnet. Flexnet is
a digipeater-based method of networking nodes, somewhat
similar to "Net/Rom", which is used extensively in Wisconsin.
Although I am not aware of any Flexnet digipeaters operating
in Wisconsin, it is interesting to look at other networking
options. Flexnet comes to us from Germany, and was developed
nearly 10 years ago. It enjoys wide popularity in Europe. but
has not caught on much here in the U.S.

     Some may be wondering why Amateur Packet radio should use
digipeaters - after all, they are inefficient and one missed
packet can translate into several retries on a marginal path.
With normal AX.25 digipeaters, this is indeed a probability.
Flexnet, however, operates differently than digipeaters that
you may be accustomed to. Packets are acknowledged over every 
step of a multiple-hop link. This makes for a reliable
connection over several digipeating "nodes". The mechanism
used to acknowledge packets is different than Net/Rom, and far
more reliable (according to it's proponents). The end-to-end
acknowledgment of basic, unenhanced digipeating is gone,
along with it's problems.

     Flexnet is AX.25 based, and has the ability to digipeat
TCP/IP frames that are encapsulated in AX.25. Net/Rom does not
have the same ability to carry TCP/IP - when forced to, it
will, but very poorly.

     A thought may have occurred when comparing Flexnet
with Net/Rom. Does a packet operator have to know the
call-signs of every digipeater in the connection path? No,
Flexnet is autorouting. In other words, it knows if it can get
to a distant digipeating node, and which route to take in order
to get there. All you need to know is the call-sign of the
station you wish to connect to, the call-sign of the local 
digipeater (node), and the call-sign of the distant digipeater.
From there, routing is automatic - the network knows the way.
The down side is that operators will not have the easy node
aliases to remember - Flexnet seems to operate with call-signs
only. It may be possible to use an alias as a call-sign, but
the authors do not mention this in their introductory
documentation.

Features:

     One very attractive feature of Flexnet is compression -
all packets sent between Flexnet Digipeaters are compressed to
save bandwidth. This helps with speed - Flexnet spends a lot
of time acknowledging packets between it's digipeaters. That
acknowledgment makes for it's good reliability, and the
compression is needed to make this operation speedier than it
would normally be.

     As was mentioned before, it acknowledgment method is
different than Net/Rom. This means that "hung" Net/Rom
connections would be a thing of the past. Connections across
state lines with multiple digipeating "nodes" is theoretically
possible (if enough nodes exist to make a path). It's authors
say it is possible to make a connection across Germany and
into other countries solely with RF based nodes. It should be
noted though, that Germany and most of Europe enjoy an
abundance of network nodes, so this more than anything makes
this possible. However, it does effectively show that we can 
overcome the Net/Rom restriction on the number of links. And 
it also shows us that Flexnet is a workable networking system
in a crowded packet radio environment.

     Flexnet is not limited in it's architecture in most ways.
It is capable of operation at both 1200 bps and 9600 bps. It
is capable of multi-port operation, so it is possible to
incorporate it into a LAN/backbone "node stack" setup. It
should be noted that there are some serious hardware
considerations that we will explore later on.

     And one appealing feature certainly is cost - there is
none. It is available through the Internet on the author's
home page.

Drawbacks:

     There are a certain number of drawbacks to Flexnet, and
they would certainly have a bearing on it's implementation.
While the software is free, a node operator may well have to
make it all up with hardware. The most common implementation
involves a number of PC Packet cards, installed in a personal
computer devoted to the task. It is possible to use a KISS TNC
with Flexnet, but all of the Flexnet capabilities are
generated in software. As of this writing (and this may be
subject to change soon), there is no plug-in TNC-2 firmware
chip that has Flexnet capabilities. So, a computer has to be
used to perform the actual Flexnet operations.

     And based on experience, an 8088 may not be fast enough
to do the trick in a multi-port "node-stack" type of setup.
Now that 386 computers are getting fairly inexpensive on the
used market, this may be less of problem than in years past.
However, most node sites are remote and "unattended". One may
go through more than a few 386 computers in a good lighting
season!

     TCP/IP operators may have an interesting time with
Flexnet. The main difference between it, and say, an X-1J node
is that it reverts to a "Virtual Connection" mode rather than
a "Datagram" mode. This means you will have a particularly
difficult time with timing parameters like IRTT, for one.

     The most significant drawback to Flexnet may be the very
thing that may make it desirable - it is different. This
means that any part of the network using it would not easily
interface to a different network. Even if a large commitment
to "change over" network software was agreed upon, there would
be gaps in coverage while the changeover was taking place.
Some may consider it a nuisance, others (like Emergency
Coordinators) may look at changeover coverage gaps as a real
handicap, temporary or not. In short, a changeover of this
nature (Flexnet or some other system) would not be entirely
painless. 

     Is Flexnet a workable networking option in Wisconsin?
Only real working experience will truly tell if it is. It may
appear not, simply because our neighbors would certainly have
a tough time interfacing with us. Add to that the fact that
there is no support for it in MSYS (the BBS software of choice
in Wisconsin) as well as other popular Host TNC programs, and
the deck is somewhat stacked against it. With the continuing
incorporation of the Internet into packet radio, it would seem
most likely to use TCP/IP as a networking protocol. However,
there are some elements of Flexnet that certainly can be
incorporated in some kind of radio adaptation of TCP/IP. So it
is certainly worth testing, for those of you who are
interested in digital networking.

So if you are hardy digital experimenter, research it via the
World-Wide-Web and check it out yourself. Find out what others
have to say about it and try it.

That's all for this time out, 73 from Andy.
