NET-4.TXT
                             NETWORK PROTOCOLS
                             -----------------

The  justification  for restricting  direct  user access to the backbone trunk,
and  advantages of  using  a packet repeater on busy user LAN access nodes were
discussed.   Thus  far we have been addressing improvements that can be made in
a  simplex  type network environment.    Continuing then, let's talk about node
firmware.

Two concepts  are to be  found in  the amateur network.   The first  is the one
most  are  familiar  with and is based on dynamic routing capability (netrom or
thenet).    The  advantage of dynamic routing is the network will automatically
update itself  when existing nodes go away,  or when new nodes are added to the
system.  A disadvantage of dynamic routing is that this process adds "overhead"
to the network,  or "node barf",  as one critic  calls it.      Node barf is in
reality the node-to-node broadcasts and update exchanges necessary  to make the
dynamic process work.

Since  all networks  have a finite capacity  (simplex networks have less than a
finite capacity),  this node overhead  reduces  the amount of user traffic that
can occupy the channel at any given time.  Until recently, little attention has
been given to ways of minimizing node barf.   However, node overhead can now be
significantly  reduced  through nodeware  selection  and intelligent setting of
node parameters.

Another  disadvantage  of dynamic routing is it has been inconvenient to affect
network control,  since earlier versions  of network firmware were deficient in
this regard.   With the release of TheNet Plus version 2.08,  the problems with
inadequate network control and excessive node barf have been addressed.

The  second  concept is based on a static routing technique and is the one used
with the  ROSE network  which is  also a  TNC-2 type of node.     Here, network
changes  have  to be  manually entered to a  configuration file  by either  the
NodeOp  or  someone  designated  as  the network  manager.    Since destination
routing is "fixed" into each node, there is no node overhead on the system.  As
a result, user throughput is improved.

Either  concept  has  advantages  and  disadvantages.   Both work quite well on
lightly  loaded simplex networks with the theoretical advantage weighted toward
ROSE  on  heavier loading.   However  the bottom line lays not so much with the
type  of  node firmware concept as it does with RF path design.    Swapping out
one type of  chip with another  on a properly designed network is not likely to
result in a noticeable performance difference.