To get good coverage you need as many LEO satellites as possible so they should each be as small as possible. Intersatellite linking could be done via automated ground stations. This eliminates the need for high-power transmtters and/or high-gain antennas on the satellites for interlinkng.
Yep, that is what we have been trying to do now for 8 years with the APRS satellites on 145.825. We just need several of them in orbit at the same time. We have demonstrated dual-hops several times whenever two or more of the APRS satellites (and ARISS) are operational at the same time. If we could get 6 to 10 of the University cubesats to simply carry the 3.4" square APRS transponder (Byonics TinyTrck-4), then we would have a constellation providing nearly continuous connectivity via these satellites from any handheld or mobile APRS radio. With 6, you might have to wait 30 minutes or so to make yoru contacts. With 10 or so, you might have to wit no more than 5 to 10 minutes for connectivity.
See www.aprs.org/cubesat-comms.html
It's better to put that gain and power consumption on earth.
The advantage of the APRS satellite concept and Packet, is that we can use a 5 Watt transmitter on the satellite to be able to hit any mobile or HT using its existing omni antenna because the packet has a low dutycycle. So running 5 watts on a cubesat is easy, because the transmitter dutycycle is only on less than say 5% of the whole-orbit time. (average power 1/4 Watt)
Whereas ECHO which is on all the time, has to be set at 1/4 watt TX power because it is on all the time.
Also, EVERY APRS satellite would be on the same frequency 145.825 with no doppler to track, and since every one of them does the same generic relay, independent of callsign, then the user on the ground just operates... He does not have to do anything to go from one satellite to another...
Bob, WB4APR