[hfsig] time syncronous radio protocol query

[chris at yipyap.com]

Hi,

I recently attended the DCC in Des Moines.  I learned
about a lot of things.  As usual, I was confirmed in my
ignorance about a whole lotta stuff.

One thing that got my mind turning was the HF digital protocol
presentations about Winlink and SCAMP.

I don't know hardly anything about information theory,
practical data bandwith at HF, etc, which will probably
be obvious in my questions!

So, my mind was thinking,  how would I get information faster
through an limited, unreliable channel?  

I thought of a couple of ways to 'cheat'.

Like those old signal-flag raisers in the days of sailing ships,
I might use a coding system where a small amount of data
would be a pointer to a larger meaning which is already
located at the message destination.  That might have some
merit. If everyone is going to be using a computer with some
software on it then there is already an assumption of shared
information resources.  The software on each end could have
a dictionary of some sort.  But the efficiency of the system
relies on the frequent use of a small range of information 
bits.  Otherwise you end up with a large dictionary, and larger
pointers to get to everything.  I think I would call this a
separate passive channel.

Another way would be to have a separate active channel, one
where control information could pass outside of the data channel.
I think some computer networks work this way.  But, how do you
get a separate channel when the receiving station is far away
and HF is what we are working with?  

Ok, so here is where my ignorant questions are starting to
come to the front.  Why not use the continous flow of time
as a separate active channel?  I can see a couple of ways
this might be of use.  One would be in synchronizing data flow.
You could have a synchronous-like protocol in an async environment.
Another would be to use the current timestamp as an encoding key
of some sort to get into that shared dictionary.

I found a couple of web pages where people have talked about this
a little.  One is on the BBN website and it appears to be related
to some military research.

Support for Time-Synchronized Radios
http://www.ir.bbn.com/projects/udaan/tdma_add.pdf

Another is a study on the ACM site about improved efficiency
with time synchronization.  This one I haven't been able
to read because I'm not a member of the ACM.

Real-time support in multihop wireless networks
http://portal.acm.org/citation.cfm?id=299571

I can see some problems in this of course.  You would have delays in
signal propagation and slight shifts of synchronization between the
sender and the receiver.

So  then I get to my questions:

1) surely someone has already done this?  or shown that it is
  a bad idea?

2) how much synchronization is possible in a practical sense?  Using
   ntp or gps methods, how synchronized can two remote systems be?

3) assuming we can get the two ends of our communication system 
   synchronized to a high degree, would that
   actually improve communications in a substantial way?  (I 
   think that is what the ACM paper is about, but I can't read it.)

Thanks for the opportunity to air my thoughts.

Chris
kc0atc