Subject: re: Linux pseudo pty
To: Greywolf <email@example.com>
From: matthew green <firstname.lastname@example.org>
Date: 02/15/2001 18:16:41
couple of points:
the "push telnet/rlogin module in to the kernel" happened because they had
performance issues. it was pounding them *far* too much to context switch
the daemon in & out for every key press... and you have hundreds of users.
it appeared as a patch for 2.3, and was standard in 2.4.
saying "sshd uses BSD-style ptys" isn't (necessarily) correct. the above
rlogin/telnet thing is a performance hack that was required for solaris
because of the insanely expensive cost of context switching in solaris.
but that doesn't mean that *only* telnetd/rlogind use ptmx! personally,
i'd rather that all software was converted to use ptmx-style ptys on
solaris... and all the stuff i run, AFAICT, don't use the bsd ptys.
as far as pushing an ssh module in to the kernel.... well, see, the
major difference between rlogin & telnet processing, and ssh processing,
is that they're static and never change. they do a job and always do it.
ssh requires significant configuration to perform it's job (crypto keys
being the major part of this). i can't see this being a workable solution
that anyone could digest.