We had this discussion quite a long time ago, in the days
before this WG was formalized. (on end-to-end, I think...,
but neurons die....and forums change).
I remember being quite dogmatic that TCP/IP/Satcom !=
TCP/IP/Fiber and provided numerous satcom systems engineering
link cals, BER data, rainfall, and outage information;
including actual math foundations.
The response was kind-of: "Interesting, correct, etc. etc."
But "WE" plan to treat TCP/Satcom like 10^-9 BER, available,
and reliable (but thanks, Tim, for pedantic theory and
practical expertise, having been there, done that, got the
mouse pad).
In fact, there is merit to considering TCP/IP/Satcom as a
"long fiber problem". It gets the community thinking and
working on the problem; and allows satcom designers to
know that their constraint is "10^-9 BER".
The danger, as I see it, is that we well see a future
'attitude shift' where satcom engineers will say
"hey, we can't guarantee 10^-9 BER except for a 30 minute
period during a clear day in June in the desert"......
So, I highly suggest that this WG formalize the requirement
that the TCPSAT group expects a certain BER over a certain
time period (or a table of these requirements) e.g.
BER PERIOD (example only, for illustrative purposes only)
10^-9 1 hour
10^-6 1 month
10^-3 2 years
This is the foundation of most of the assumptions, requirements,
etc. in satcom links, IMHO.
Regards!
Tim
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:37 EST