Path: gmd.de!urmel.informatik.rwth-aachen.de!
newsserver.rrzn.uni-hannover.de!hrz-ws11.hrz.uni-kassel.de!
news.th-darmstadt.de!math.fu-berlin.de!zib-berlin.de!netmbx.de!
Germany.EU.net!EU.net!howland.reston.ans.net!paladin.american.edu!
darwin.sura.net!newsserver.jvnc.net!igor.rutgers.edu!geneva.rutgers.edu!
hedrick
From: hedr...@geneva.rutgers.edu (Charles Hedrick)
Newsgroups: comp.os.linux.misc
Subject: Re: Linux vs. NetBSD
Message-ID: <Jan.23.21.30.58.1994.1280@geneva.rutgers.edu>
Date: 24 Jan 94 02:30:59 GMT
References: <2hosie$4ts@homea.ensta.fr> 
<2hv1h9$gt9@olivaw.apanix.apana.org.au> <RV.94Jan23194254@lego.cs.brown.edu>
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 63

r...@cs.brown.edu (rodrigo vanegas) writes:

>But BSD source is free isn't it? Why don't the linux network code
>developers simply copy it out?

I think the real answer is that it's more fun to write a new
implementation. However there are also a couple of rational
justifications:

- the BSD code is now fairly old. Some people have claimed that
a new implementation could get better performance, particularly
in a Linux environment. (There are some differences in 
the kernels that may affect how one would interface networking
code to it.)

- there is still a suit in progress covering the free BSD code.
While I am reasonably sure that the networking portion
of the code is not at risk, courts have in the past made
decisions that I consider irrational. It's not entirely
unreasonable (though perhaps it is bordering on paranoia)
for the community to want an alternative to the BSD code.

Note also that you can't just take the BSD modules and insert them
into the kernel. Some adaptation would be necessary to make them fit
into the Linux kernel. Whether that adaptation would be more or less
work than fixing the existing Linux networking code is a matter of
debate.

I personally think that neither of the reasons above is very strong.
Berkeley has put an incredible amount of work into design and
performance, and in order to do better you'd at least have to start by
understanding their work. (I say this despite the fact that I get
roughly twice the throughput for FTP's involving Linux than netBSD.
However it's hard to believe that this is a basic problem with the BSD
code. Its basic capabilities are well known from other ports. In
particular, it can saturate an Ethernet with a single TCP connection
using less processing power than our Linux systems have.) The
networking portion of BSD had been distributed widely well before the
suit, and as I understand it ATT had agreed that that portion of the
code was OK to distribution. So I don't think there's any real legal
issue.

The real answer is more practical: no one has done the necessary work.
There was actually a port of the netBSD networking code into Linux. I
tried it briefly. I thought it was very promising. However there
were enough loose ends that the Linux code worked better. I think the
community would have been better off to clean up that port than do
something new. However the direction of Linux is determined by what
people are willing to put time into. If everybody working on the
networking code is more interested in working on something new, it
doesn't matter if that makes sense or not -- 1,000,000 people could
want a BSD port, but if they're not willing to put their coding where
their mouth is, it doesn't make any difference. I don't know what
Linus would do if different people sent him both a new TCP
implementation and a fully functional port of the BSD code. He might
be inclined to stick with the GPL'ed code, or he might not. But he
hasn't been presented with that choice yet. I believe the current
Linux TCP can be made solid. Since I don't have time to do my own TCP
for Linux, that's what I working on. And I have just enough residual
paranoia about the legal system that I'm not sure that's a bad thing.
I can make excellent arguments that no possible outcome of the suit
could interfere with distribution of the BSD network code, but
lawyers just don't think like normal human beings.

Newsgroups: comp.os.linux.misc
Path: gmd.de!xlink.net!zib-berlin.de!netmbx.de!Germany.EU.net!EU.net!
howland.reston.ans.net!pipex!uknet!cf-cm!cybaswan!iiitac
From: iii...@swan.pyr (Alan Cox)
Subject: Re: Linux vs. NetBSD
Message-ID: <1994Jan24.182958.7042@swan.pyr>
Organization: Swansea University College
References: <2hosie$4ts@homea.ensta.fr> 
<2hv1h9$gt9@olivaw.apanix.apana.org.au> <RV.94Jan23194254@lego.cs.brown.edu>
Date: Mon, 24 Jan 1994 18:29:58 GMT
Lines: 13

In article <RV.94Jan23194...@lego.cs.brown.edu> r...@cs.brown.edu 
(rodrigo vanegas) writes:
>But BSD source is free isn't it? Why don't the linux network code
>developers simply copy it out?

Because AT&T appeared to have different potential opinions, and because we
had a chance to do better. Primarily though because Ross Biro sat down and
started writing a TCP/IP stack. At the end of the day we will have a better
network environment than generic BSD NET-2 and it will be GPL protected.

Alan
iii...@pyr.swan.ac.uk

Newsgroups: comp.os.linux.misc
Path: gmd.de!newsserver.jvnc.net!darwin.sura.net!howland.reston.ans.net!
vixen.cso.uiuc.edu!sdd.hp.com!think.com!spdcc!merk!rmkhome!rmk
From: r...@rmkhome.com (Rick Kelly)
Subject: Re: Linux vs. NetBSD
Organization: The Man With Ten Cats
References: <2hosie$4ts@homea.ensta.fr> <2hv1h9$gt9@olivaw.apanix.apana.org.au> 
<RV.94Jan23194254@lego.cs.brown.edu> <1994Jan24.182958.7042@swan.pyr>
Message-ID: <9401302227.29@rmkhome.com>
Reply-To: r...@rmkhome.com (Rick Kelly)
X-Newsreader: TIN [version 1.2 PL2]
Date: Mon, 31 Jan 1994 03:40:56 GMT
Lines: 21

Alan Cox (iii...@swan.pyr) wrote:
: In article <RV.94Jan23194...@lego.cs.brown.edu> r...@cs.brown.edu 
(rodrigo vanegas) writes:
: >But BSD source is free isn't it? Why don't the linux network code
: >developers simply copy it out?
: >
: Because AT&T appeared to have different potential opinions, and because we
: had a chance to do better. Primarily though because Ross Biro sat down and
: started writing a TCP/IP stack. At the end of the day we will have a better
: network environment than generic BSD NET-2 and it will be GPL protected.

Just remember, "better" is subjective.

If you are going to reinvent the wheel, just be ready to test it against
the approximately 300+ versions of TCP/IP, NIS, and NFS that are out there
now.

I deal with this everyday. It's a swamp.

-- 

Rick Kelly r...@rmkhome.com r...@bedford.progress.com

Newsgroups: comp.os.linux.misc
Path: gmd.de!newsserver.jvnc.net!howland.reston.ans.net!pipex!uknet!
cf-cm!cybaswan!iiitac
From: iii...@swan.pyr (Alan Cox)
Subject: Re: Linux vs. NetBSD
Message-ID: <1994Feb1.170209.5016@swan.pyr>
Organization: Swansea University College
References: <RV.94Jan23194254@lego.cs.brown.edu> 
<1994Jan24.182958.7042@swan.pyr> <9401302227.29@rmkhome.com>
Date: Tue, 1 Feb 1994 17:02:09 GMT
Lines: 23

In article <9401302227...@rmkhome.com> r...@rmkhome.com (Rick Kelly) writes:
>If you are going to reinvent the wheel, just be ready to test it against
>the approximately 300+ versions of TCP/IP, NIS, and NFS that are out there
>now.

People already have been testing it. One thing Linux has is an army of
alpha testers. Mostly its now obscure issues like NCSA telnet with its
mss > window. Rik Faith has already done a lot to the NFS side to cope
with the numerous Sun and other bugs.

>I deal with this everyday. It's a swamp.

Yep.. but we are still wading onwards and have no intention of stopping.
I've traced several bugs to broken tcp stacks now - but almost every bug
we've found has been a Linux one normally an 'assumption'.

Anyway if we wanted an alternative networking layer something like the Mach
net server code would bne much more appropriate than the early 80's 
mentatlity of BSD. The BSD code isn't wonderful by design - its pretty good
by design and better than the current legacies imbedded in Linux NET2[DE].
It's more reliable through testing - nothing else.

Alan