jpayne%bbn-vax@sri-unix.UUCP (08/19/83)
From: Jonathan Payne <jpayne@bbn-vax> Recently I was wondering why you can't detach yourself from the system, and then attach from somewhere else, like on tops-20. Then I remembered what csh says when you try to suspend your login shell: Can't suspend a login shell (yet). Does that mean somebody is planning to implement the equivilent of detach and attach? And would it be difficult?
msc@qubix.UUCP (Mark Callow) (08/21/83)
A tops[12]0 like detach and attach facility for Unix (at least bsd Unix)
would be somewhat complicated by the necessity for establishing
TERM and TERMCAP environment variables for the new;y attached terminal.
--
Mark Callow, Saratoga, CA.
...{decvax,ucbvax,ihnp4}!decwrl!
...{ittvax,amd70}!qubix!msc
decwrl!qubix!msc@Berkeley.ARPAMBM@mit-xx@sri-unix.UUCP (08/22/83)
From: Michael B McIlrath <MBM@mit-xx>
Return-path: <jek@SCRC-VIXEN>
Received: from SCRC-HOUSATONIC by SCRC-TENEX with CHAOS; Mon 22-Aug-83 09:07:03-EDT
Date: Monday, 22 August 1983, 09:07-EDT
From: James E. Kulp <jek@SCRC-VIXEN>
Subject: [Jonathan Payne <jpayne@bbn-vax>: Suspending login shell]
To: Michael B McIlrath <MBM@MIT-XX>
Cc: jek@SCRC-TENEX
In-reply-to: The message of 22 Aug 83 08:13-EDT from Michael B McIlrath <MBM at XX>
Date: 22 Aug 1983 0813-EDT
From: Michael B McIlrath <MBM@MIT-XX>
Date: Fri, 19 Aug 83 9:06:39 EDT
From: Jonathan Payne <jpayne@bbn-vax>
Recently I was wondering why you can't detach yourself from the system,
and then attach from somewhere else, like on tops-20. Then I remembered
what csh says when you try to suspend your login shell:
Can't suspend a login shell (yet).
Does that mean somebody is planning to implement the equivilent of detach
and attach? And would it be difficult?
-------
I've thought about this a lot, since its the main outstanding UNIX network
functionality lossage that I can think of. Unfortunately there is no
appropriate data structure which can hold that state of a "session" after
your carrier (virtual or otherwise) drops. The cheap way to do this is
by creating a new software tty driver which is just a detached tty that
doesn't do anything. The right way is to have a data structure called
a session, but this would involve many global changes. But my days
of making major changes to UNIX (and probably my ability to get them
adopted by Berkeley) are over. I'm almost to the point of doing this
(the former), since its so annoying.
(Sometimes one more message is enough to get me going).
-------rehmi@umcp-cs.UUCP (08/24/83)
I did something like this once with first mpx files, and then with ipc when
all the various filesystem-trashing mpx bugs crawled out of the woodwork
onto me. They pseudo-worked, but still have many bugs, and it seems easier
to do with pseudo-ttys. Or, leave your terminal logged in overnight :-) !
There was one cute crock in the mpx code where I went to read kmem to do
something like ioctl(mpxfd, FIONREAD, &i) to see if there was anything to
read on the mpx file. Well, it *was* cute at the time!
-rehmi--
By the fork, spoon, and exec of The Basfour.
Arpa: rehmi.umcp-cs@udel-relay
Uucp:...!harpo!seismo!umcp-cs!rehmi
...!{allegra,brl-bmd}!umcp-cs!rehmijames@umcp-cs.UUCP (08/24/83)
Just to make this work, you could write a kmem-writer program which would change the control-tty of your shell, along with any other important data. Maybe those new files called FIFO's (pipe with a name?) would be better than pseudo-tty's.
notes@CSvax:Pucc-H:pur-phy.UUCP (09/04/83)
#R:sri-arpa:-436000:pur-phy:12700017:000:647 pur-phy!crl Aug 27 22:14:00 1983 TOPS-20 (Twenex) loses your terminal type on a detach/attach. With UNIX, it would probably be possible to have a .attach or a .continue file (analogous to the .login/.cshrc/.logout) to handle the change of terminal. Of course, stopped jobs would lose, but oh well. On the TOPS-10 system at Worcester Polytechnic Inst. before they got a 20, a feature called "freeze" was installed. With it, one could suspend a process into a file and continue at any later date. The restart was invisible--except for the tty, all open files were reopened to the right place. Anyone start on such a beast? Charles LaBrec pur-ee!Physics:crl purdue!Physics:crl