[e2e] TCP in outer space
fred at cisco.com
Tue Apr 10 14:40:27 PDT 2001
At 02:33 PM 4/10/2001 +0200, Michael Welzl wrote:
>Why isn't TCP translated into a rate-based scheme and vice versa across
>a satellite (or other wireless) link?
I don't think there's any argument that it should be, when delays are long
and bandwidth is assured; if you look at NASA's SCPS, that's about what it
does. The problem is that sessions using TCP don't usually spend all of
their time in space - a session spends a good bit of time going through
perfectly normal terrestrial parts. This is fundamentally the argument,
though, for starting out with enough data in flight to either get the job
done or get it going at a good clip.
The place where a rate-based scheme makes sense is when an identifiable
channel (mumble KBPS out of the link between here and there) is known to be
available and nobody else will use it. There are few places in the Internet
where that assurance is a given, today, but clearly on links between the
earth and other planets it can be known with great certainty. So in such
contexts, rate-based transports are both feasible and necessary.
But applying the same reasoning to email to/from the space station is a
little difficult. Network News might make more sense than SMTP in space, as
it is about moving files around. But even for that, what bandwidth would
you want to ensure, and what bandwidth would you want to say a file
transfer needed? TCP is a more natural fit, I should think, because it
automatically adapts itself to the bandwidth available to the task.
More information about the end2end-interest