[Mep-dev] The DTN Protocols and the MEP (and ACP)

[Timothy J. Salo]

I believe that the MEP project should focus on developing
a deployable prototype that provides fully functional IP
connectivity. In my opinion, any efforts such as integrating
the DTN protocols into the MEP should be deferred until after
an initial, IP-based MEP prototype has been successfully
demonstrated.

The MEP as an IP Network
------------------------

In my view, the MEP should appear to the devices that use
it to be an IP network.  That is, to a device using the
MEP, the MEP should be indistinguishable from any other
IP network.  This strategy ensures that the MEP
interoperates with existing and future IP devices and
networks (i.e., pretty much every networked device in
the known universe).  (This includes, amazingly enough
AX.25 devices -- see RFC 1226, "Internet Protocol
Encapsulation of AX.25 Frames").  Another significant
benefit of this IP-based model is that it clearly
identifies the boundary of the MEP system.  The MEP
provides IP connectivity and higher-level protocols
(which could include the DTN protocols) are implemented
in hosts that attach to or use the MEP.

I tried to articulate this architectural approach in
"Proposed Network-Centric Architecture for the
Advanced Communications Package (ACP)",
<http://www.saloits.com/papers/AMSAT2008.pdf>.
If you like, you can simply substitute "MEP" for
every occurrence of "ACP" -- the argument is the
same.  For that matter, I claim that this "make it
look like an IP network" approach should be used
for all amateur radio communications systems.
(Well, for all of the infrastructure-based systems,
those that have repeaters/routers, either terrestrial
(MEP) or on-orbit (ACP).  Things become a bit more
complicated in the absence of repeaters.)

With that background, let me make a few comments on
the DTN protocols.

Play with the DTN Protocols Today
---------------------------------

Bob McGwier wrote:
 > There is no "porting" to be done to test and learn about it.  You can
 > download dtn as source TODAY and a simple google search reveals:
 >
 > http://sourceforge.net/project/platformdownload.php?group_id=101657

Absolutely.  I suggest that people interested in the DTN
protocols start by reading the DTN RFCs and Internet drafts.
They are available at:

https://datatracker.ietf.org/idtracker/?search_group_acronym=&search_job_owner=0&search_rfcnumber=&search_status_id=&sub_state_id=6&search_cur_state=&search_button=SEARCH&search_filename=dtnrg&search_area_acronym=

Additional documents are available at:

   <http://www.dtnrg.org/wiki/Docs>

And, as Bob noted, a reference implementation of the DTN
protocols is available.  You can play with the DTN protocols
today.  The DTN reference implementation includes convergence
layers for TCP and UDP.  So, you should be able to make DTN
work over pretty much any IP network.  And, if the MEP
really does look just like another IP network, then you can
use DTN with the MEP without any additional development work.
MEP is (or should be) just another IP network.

Integrating the DTN Protocols with MEP Devices
----------------------------------------------

Would it be useful to integrate the DTN protocols into MEP
devices (ignoring the risk that this work would distract
from the completion of core MEP functionality)?  Probably.
Would this functionality be useful enough to justify the
costs and risks involved?  I'm not sure, but I suspect not.

One could think about putting DTN functionality in the MEP
repeater (presumably using the IP connectivity provided by
the MEP system).  Conceptually, this is equivalent to
adding an application host at the repeater site.  In this
case, the application protocols would be the DTN protocols.
The easiest way to realize this is to put three, rather
than two, IP interfaces on the MEP repeater (i.e., an
input channel, an output channel, and a local Ethernet
interface).  The benefit to the MEP of this approach is that
very little incremental development effort is required to
support the DTN protocols -- simply add an Ethernet
interface to the MEP repeater.  Anyone interested in
running the DTN protocols would simply add a host
running the DTN protocols at the repeater site.

A more interesting question is whether the DTN protocols
should be implemented in an on-orbit ACP payload.  While,
I don't think that the complexity of the ACP software is
yet fully appreciated, I believe that adding the DTN
protocols to the ACP payload would substantially increase
its complexity.  This is probably not the place you want
to add complexity, rather than on the ground.  But, the
same strategy could be used: simply add an application
host to the satellite to run the DTN protocols.  This
would reduce the complexity and risk added to the ACP project.

But, given the nature of the MEP and ACP networks, I
question whether adding DTN functionality to the
MEP/ACP repeater/payload would really add much benefit.
Just run the DTN protocols in the end devices, and let
the DTN protocols be invisible to the MEP/ACP system.

Let's focus on building the MEP (and ACP) before adding
(or being distracted by) additional, optional functionality.

-tjs