[ax25-layer2] The lights.

[pete at ae5pl.net]

Good questions...

> -----Original Message-----
> From: James Ewen
> Sent: Friday, September 23, 2005 9:42 PM
> To: Discussion of Link Layer use of AX.25
> Subject: Re: [ax25-layer2] The lights.
> 
> Am I out in left field? Where does the routing go to when it 
> is removed from the TNCs? The NSR concept doesn't appear to 
> remove layer 3 functionality from layer 2, it simply removes 
> the ability of the user to control where packets go, and 
> hands it over to someone else.

The implementation proposed is a combination of the NSR for UI packets
and v2.2 implementation for connected-mode packets (restricting chained
digipeaters to a maximum of 2).  It is also proposed that layer 3
implementations _based_ on AX.25 such as those described previously not
be implemented in the digipeater nor should they be implemented at
endpoints.

Yes, the NSR algorithm does remove the ability of endpoints to define
paths for UI (broadcast) packets.  This, however, is what facilitates
providing layer 3 protocols a transparent, universal layer 2.  For any
layer 3 protocol to work with transparently with a layer 2, the layer 3
must be able to do discovery of other layer 3 devices on the specific
layer 2 interfaces.  NSR provides this in such a way that the endpoints
do not need to be pre-configured for local RF network topologies.  This
is the critical piece because this means that path determination for
connected modes can now be done algorithmically instead of manually
changing configurations as you move from LAN to LAN.

But, as you point out, the NSR algorithm by itself does not make AX.25 a
usable layer 2 protocol.  It does, however, facilitate this by providing
a "universal" discovery capability to higher layers.  When the other
pieces such as v2.2 implementation with connected mode connectivity in
the digipeater and at the endpoints allows LAN connectivity to be done
without any prior knowledge of local RF network topology.

If we move AX.25 LANs in this direction, we would be able to have
application and layer 3 implementations that could be created which
would work across _any_ AX.25 link (and any other layer 2 link, for that
matter).  If people can write to a standard interface that they know is
implemented in most areas where their application will be used, they can
concentrate on making applications, not making RF routing algorithms.
It also means that their applications can be used anywhere a layer 2
network exists, regardless of what protocol makes up that link layer at
each interface.

> A network of TNCs with the network capabilities removed 
> leaves me looking at a bunch of hardware that can not talk to 
> anywhere.

The exact opposite.  A network of TNCs with network capabilities removed
leaves you with network interface devices that let you talk to the
world.  This is because the routing is done outside of the link layer
and off the single RF channel used by digipeater chaining.  Does this
require an Ethernet interface on the other side of each endpoint?  No.
Does it require something more than simply attempting to repeat packets
multiple times on the same channel?  Yes.  What we have today is a bunch
of hardware that can not talk anywhere unless you are using the specific
higher layer protocol that is implemented on a particular channel and
then you are limited to how far you can go by simple physics.  By
implementing a universal interface for layer 2 AX.25, any AX.25 network
can be used for any layer 3 and above applications.  That common
interface opens up the world.

73,

Pete Loveall AE5PL