Frank W. Steiner <steinfw@hotmail.com> writes:
>But if there is no written specification then some particular
>implementation of scp must be the reference one, right? Which one is it?

There's no particular reason why everyone should have chosen the same
"reference implementation" to reverse-engineer their SCP implementations
from.

I suspect that the de facto "reference implementation" will probably
turn out to be Ylonen et al's original ssh (which I think invented SCP),
or its derivative OpenSSH, simply because the source is widely
available.