*BSD News Article 2905


Return to BSD News archive

Newsgroups: comp.unix.bsd
Path: sserve!manuel!munnari.oz.au!sgiblab!zaphod.mps.ohio-state.edu!cs.utexas.edu!uunet!brunix!brunix!wcn
From: wcn@cs.brown.edu (Wen-Chun Ni)
Subject: Re: AT&T vs. BSDI --> 4.3BSD-NET2 distribution requires AT&T license!!!
Message-ID: <1992Jul31.211148.14512@cs.brown.edu>
Keywords: AT&T 'Death Star' rises over BSDI's horizon [Tel. 1-800-800-4BSD
Sender: news@cs.brown.edu
Organization: Brown University Department of Computer Science
References: <l6nibgINNje6@neuro.usc.edu> <1992Jul21.152007.1126@news2.cis.umn.edu> <1992Jul30.174414.28488@kas.helios.mn.org>
Date: Fri, 31 Jul 1992 21:11:48 GMT
Lines: 72

In article <1992Jul30.174414.28488@kas.helios.mn.org> rhealey@kas.helios.mn.org (Rob Healey) writes:
>In article <1992Jul21.152007.1126@news2.cis.umn.edu> rodeen@buddha.ncc.umn.edu (Rick Odeen) writes:
>=I don't think this is a valid claim, Linus Torvalds developed the Linux
>=system in less than one year from scratch.
>=
>	Where did Linus get 99% of his MODELS for Linux? Ans: USL and BSD
>	UNIX. "We stand on the shoulders of giants..."
>
>	In other words, he used models for OS concepts that originated
>	in UNIX(tm). The system calls, the library calls, the utility
>	names, the program names, the memory models, the networking, the
>	file systems, the concepts of UID,GID,SUID,SGID, sticky bits,
>	mountable filesystems on a tree, etc. These are all basic
>	features of UNIX(tm) that he used when creating Linux. You'd
>	be VERY hard pressed to find OS concepts these days that HAVEN'T
>	passed through a UNIX(tm) kernel at some time in the past. For no other
>	reason than UNIX(tm) is the OS most researchers work with on
>	a day to day basis and what they tend to hack on. Even micro kernels
>	like MACH and probably NT borrow QUITE a bit from the UNIX(tm) system
>	in system call names and symantics as well as utility and program
>	names. Hell, even that pathetic MSDOG lifts I/O redirection and
>	directory command names from UNIX(tm), amongst other things.
>

I beg to differ. In my opinion, there are few original ideas really
coming from Ritchie and Thompson. *Unix* is a product of simplification.
The '70s are really the decade of simplifications: Pascal to Algol 68
and Unix to Multics. Please remember the initial platform Unix was
developed was only a 64K machine; you can't stand large amount of 
system overhead in doing fancy jobs like Multics did. The UID/GID/SUID
stuffs are merely *hacks* when we want to fit into a small system.
The "fork/exec" pair is also in the same class. The indistinction
between devices and files is only a simplfied idea of memory/file/..
objects used in Multics. It took computer scientists two decades
to fix the protection problems in Unix without success. The inter-
process commmunication under Unix is ugly compared with Multics.
So, why Multics failed? Yes, the hardware. Sure, Multics had its
problems, but we can't deny its innovative ideas and pioneering
works. Can MIT/GE sue Bell Labs for stealing the ideas from Multics?
Sure they can't.

Speaking originality, Corbato should have gotten his Turing award much
earlier than Richie and Thompson. The only machine, to my best
knowledge, targeting at Corbato's dream system was Intel iAPX 432,
which turned out to be a failure. The market won't accept such a chip
and such a system. The success of Unix can be ascribed to the freely
distributable source forms to the academic world before Edition 7.
Unix is popular doesn't mean that Unix is very original.

After the 80's, people are looking for a new way of integrating system
tools. The bad communication mechanism underlying Unix causes major
problems. Steve Reiss's great integrating software FIELD (which
derived HP's softbench) should be not that large if it's not developed
under Unix.

Please note that we are not voluntarily to use Unix in our college 
years. It was because the departments used it  as the basic environment
based on AT&T's *generous* offer. At last, we are hooked, and AT&T
can sue any of us for exposure to the Unix source (yes, I admit
that I studied the PDP 11/45 version of Unix ed 6 during my college
years.) We feel betrayed because of this.

Yes, Linus Torvalds did not invent new ideas from the academic
viewpoints. But he developed Linux with "learning 386" in mind,
and the playful toy turned out to be a great system *based*
on what we have *learned*.  Have you guys ever learned anything 
about IBM's MVS in class?




Wen-Chun Ni