scp not tolerant of extraneous shell messages

[Bryan Henderson]

>So sub-system servers are executed via a shell - the user's shell.

OK, I see.  Not sure why I didn't before, unless I was blinded my
fundamental belief that there's some difference between a subsystem
and an ordinary command.  :-)

So what is the essential difference between a subsystem and an ordinary
command?  I see that the .sshrc doesn't get executed for a subsystem, and
that sshd looks up the subsystem program by a special name rather than
the client having to supply the filename.  Is that it?  It doesn't seem
like enough to justify the extra concept.

Did this maybe work differently in the past?  Because the naming of
the functions and source code comments such as "Performs common
processing for the child ... and executing the command or shell" seem
to obscure the fact that the fundamental function of this code is to
execute the user's shell, no matter what.

-- 
Bryan Henderson                                    Phone 408-621-2000
San Jose, California