Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!ut-sally!im4u!rutgers!mit-eddie!genrad!decvax!ucbvax!ALEXANDER.BBN.COM!jr
From: jr@ALEXANDER.BBN.COM (John Robinson)
Newsgroups: comp.emacs
Subject: Re: Pseudo ttys for Jove
Message-ID: <8703092334.AA19899@ucbvax.Berkeley.EDU>
Date: Mon, 9-Mar-87 18:34:56 EST
Article-I.D.: ucbvax.8703092334.AA19899
Posted: Mon Mar  9 18:34:56 1987
Date-Received: Tue, 10-Mar-87 18:55:52 EST
References: <8703091928.AA05097@icst-cmr.arpa.ARPA>
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 23

>> OK folks, another radical porposal. Who needs software pty's if you've
>> got real hardware ttys. Just get yourself an extra multiplexor and connect
>> null modem cables between every pair. Run a getty (or even a shell!) on
>> the odd ones and nothing on the even ones. Rename the even ones /dev/pty[0-3]
>> and the odd ones /dev/ttyp[0-3] (I am assuming an 8 channel mux).
>> 
>> OK, OK, so I'd really have `real' pseudo-ttys than `pseudo' real-ttys,
>> but if that's what you're stuck with it just might work.

Not too well.  Unix tty output generally blasts along at a great clip,
while Unix tty input is real sluggish.  The output direction from the
real process would quickly outrun the input buffers to emacs or
whoever owns the "pty".  So you'd have to turn on both flow control
and TANDEM, thus losing ^Q and ^S each direction.  "Real pseudo" tty's
can and do provide much better flow control than this because the
kernel can synchronize the user processes with much greater ease in
ways that don't mess up the user data stream.

/jr

>> 	Why did Paul Simon name his album after Elvis Presley's house?

Why did Ritchie & Thompson name input flow control after a non-stop computer?