Subject: Re: libedit (was Re: bin/3011: ftp could be smarter with host:/path and URL's )
To: Michael Richardson <>
From: Grey Wolf <>
List: current-users
Date: 12/17/1996 09:19:47
Michael Richardson sez:
 *     Jason> ...uhh, definitely not in the kernel :-)
 *   I partially agree, and partially disagree. 
 *   I've been to put my $0.02 in on mouse's serial line
 * suggestions. We've really just come back to this discussion.
 *   I think the TTY driver is both too big and too small.
 *   It is too big given that most *real* ttys are being used for PPP
 * or other non-VDT uses. All that termios cruft, and the silos,
 * etc.. are just wrong for this.

How do you figure?  You at least need the termios stuff to switch
into raw mode to use PPP and other non-VDT uses.

Perhaps, though, we should have a mechanism for detaching a given
driver from a port and just attaching a low-level PPP driver.
Or whatever.

 *   Meanwhile, the pty's are being used more and more, we all use
 * xterm or telnetd, or sshd, to drive them. I think that the line
 * editing should go in the "xterm", not the application.

I disagree with this approach because it does not provide proper access
to, e.g., history recall/editing via the shell, variable/filename expansion
via ESC, and the like.  If you're going to yank the editing from the
application, you might as well yank it all (i.e. having the line editing
in the xterm and the rest of the stuff in the shell (I use tcsh, FWIW;
bash does roughly the same stuff) doesn't make sense).

 *   Also, I loath what pty's do: make us go through the kernel again.
 *   Solaris put telnetd in the kernel. They should have provided a
 * telnet protocol STREAMs module, and pushed it onto a socket, and then
 * "named" the socket to become a /dev/pts/X. This really doesn't change
 * the model much though.

As you seem to be pointing out, Solaris goofed.  telnetd does not belong
in the kernel.

You also bring up SEWER^H^H^H^HTREAMS again.  Does _any_one out there
have any prototypes of the stuff Mike Karels was working on?  I mean,
it's been declared dead for a while; you'd think he'd at least let
someone else continue on with it.  Maybe we just need to start a
similar implementation from scratch.

[BTW -- has he fallen off the face of the planet or what?]

 *   I think it is time to consider if there is a better model for
 * dealing with tty-like things. We provide a pty like facility for
 * applications that expect a real tty.

I think it's time we came to some agreement of what 'tty' is supposed
to mean.  I know what _I_ think it means, which is what it's always
been.  Certainly there are those of us who would disagree with this.