From SRA at XX.LCS.MIT.EDU  Fri Aug  7 03:52:00 1987
From: SRA at XX.LCS.MIT.EDU (Rob Austein)
Date: Aug 6 1987  21:52 EDT
Subject: Unix catching up to ITS
In-Reply-To: Msg of 6 Aug 1987  14:00-EDT from barmar@Think.COM
Message-ID: <SRA.12324458461.BABYL@XX.LCS.MIT.EDU>

No, that's the USR: device.  But they've invented the JOB: device too,
something called "portals" I think.

As Noel Chiappa said when talking about ITS and Multics, it's really
kind of scary to think that we are just now catching up to what we
were able to do twenty years ago.


From barmar at Think.COM  Thu Aug  6 20:00:39 1987
From: barmar at Think.COM (barmar at Think.COM)
Date: Aug 6 87 14:00:39 EDT
Subject: Unix catching up to ITS
Message-ID: <8708061800.AA08725@godot.think.com>

I read the following in the Usenet newsgroup comp.unix.wizards (== to
the Arpanet mailing list UNIX-WIZARDS).  AT&T has finally "invented"
the JOB: device!

Article 3409 of comp.unix.wizards:
Path: think!husc6!seismo!mnetor!utzoo!henry
From: henry at utzoo.UUCP (Henry Spencer)
Newsgroups: comp.unix.wizards
Subject: /proc, /n/face
Message-ID: <8285 at utzoo.UUCP>
Date: Jul 10 87 19:06:20 GMT
References: <7879 at brl-adm.ARPA> <2211 at bunker.UUCP>, <6043 at brl-smoke.ARPA>, <8244 at utzoo.UUCP>
Organization: U of Toronto Zoology
Lines: 42

> V8 /proc preserves the semantics of a normal Unix directory setup *exactly*,
> unless I missed something subtle when I read the code.  My impression is that
> /n/face does likewise.  In both cases the directory hierarchy is actually a
> figment of the kernel's imagination, but it is a consistent figment with the
> same semantics as normal directories.

Several people have asked what this is about, so I suppose I should elaborate
a bit.  This stuff has been presented in papers at Usenix conferences, but
not everybody's familiar with those.

/proc is V8's replacement for the ptrace() system call and related things.
It's a slightly-odd type of file system.  Once mounted, it looks like a
single directory containing a bunch of files with numeric names.  If you
open (say) file "12345", you are looking at the address space of process
number 12345.  Writes into the file are writes into the address space.
There are ioctls for things like stopping and starting the process, sending
it signals, etc.  Access to the files is naturally subject to the standard
Unix file-permission system.  The whole thing is actually a figment of the
kernel's imagination, with the "directory" manufactured on the fly whenever
someone tries to read it, and operations on the "files" turned into the
corresponding operations on the processes.  Apart from being cleaner than
ptrace(), /proc is also faster.

[description of /n/face deleted]

In both cases, these odd filesystems look exactly like real ones, down to
things like "." and ".." entries in the directories.  You can use all the
standard Unix tools to operate on them.

At least one System V implementation of /proc exists inside AT&T, but it
hasn't made it into any released software that I know of.  I believe /n/face
is strictly a V8ism at the moment.
-- 
Mars must wait -- we have un-         Henry Spencer @ U of Toronto Zoology
finished business on the Moon.     {allegra,ihnp4,decvax,pyramid}!utzoo!henry




From ALAN at AI.AI.MIT.EDU  Mon Aug  3 02:34:10 1987
From: ALAN at AI.AI.MIT.EDU (Alan Bawden)
Date: Sun,  2 Aug 87 20:34:10 EDT
Subject: IP routing table
In-Reply-To: Msg of Wed 15 Jul 87 01:55 EDT from Alan Bawden <ALAN@AI>
Message-ID: <236463.870802.ALAN@AI.AI.MIT.EDU>

    Date: Wed, 15 Jul 87 01:55 EDT
    From: Alan Bawden <ALAN at AI>
    ...  Try out:

      :ALAN;REDRCT

    (If anyone can think of a good directory for this to live on other than
    ALAN, let me know.)...

It finally occured to me that the right directory for something like this
is the one named ".".  (Remember ":.;BOOT11" on the KL?)  So make that:

  :.;REDRCT

to fool with IP routing.