Klaus von der Heide wrote:

[...]
2. The symbol rate is exactly  1 bit/s.  Starting every minute,
    a constant random pattern of 60 bits is sent out.
    Appropriate matched filters must be used at both ends.

That seems like a good idea. But just to start with, what about :
a)  a transmitter keyed "on" for 1 minute, then "off" for 1 minute for
instance.
b) using a GPS based frequency reference (described in QST , and
mentioned again recently on this reflector)
c) if the "on" period is 60 seconds long, even if our timing is off by a
few nanoseconds that should not matter, since the error, if derived from
GPS,  should not be cumulative
c) we could also add those into an accumulator, etc. and wait as many
days (months ?) as required.
After some preliminary success we could then go for more intricate
schemes.

As I mentioned some weeks ago, phased array antennas
at both ends would increase the signal by many dBs.

Should we consider several receiving stations, physically separated to
insure diversity reception, all listening to the same frequency, and
polling their resources afterwards?
73
Andre'  N4ICK