Technology and Trends
 USENET Archives
  
From: ryen...@seas.gwu.edu (Richard D. Yentis)
Subject: freeBSD and Linux networking code (what is different?)
Date: 1995/05/21
Message-ID: <3pnmm5$i82@felix.seas.gwu.edu>#1/1
X-Deja-AN: 103025601
organization: George Washington University
newsgroups: comp.os.linux.networking


I have been using Linux for several years, but I have not used the
network code until very recently.

I am wondering what are the differences between the two versions of
BSD and the Linux networking code.  Are they all exactly the same?
are Linux and freeBSD the same but the other BSD is newer and/or
better?  Is there a historical tie between the code?  Can someone
point out who I might ask to find out?

Thanks,

Rich
-- 
Richard Yentis, Jr.			ryen...@seas.gwu.edu
Apt. 114				202-342-0952
510 21st St. North West
Washington DC 20006

From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: freeBSD and Linux networking code (what is different?)
Date: 1995/05/30
Message-ID: < D9EGyn.HHs@info.swan.ac.uk>#1/1
X-Deja-AN: 103496614
sender: n...@info.swan.ac.uk
x-nntp-posting-host: iifeak.swan.ac.uk
references: <3pnmm5$i82@felix.seas.gwu.edu>
organization: Institute For Industrial Information Technology
newsgroups: comp.os.linux.networking

In article <3pnmm5$...@felix.seas.gwu.edu> ryen...@seas.gwu.edu (Richard D. Yentis) 
writes:
>I am wondering what are the differences between the two versions of
>BSD and the Linux networking code.  Are they all exactly the same?
>are Linux and freeBSD the same but the other BSD is newer and/or
>better?  Is there a historical tie between the code?  Can someone
>point out who I might ask to find out?

NetBSD uses the BSD NET/2 code, from 4.3 BSD series kernels. This is not
stunningly fast, not stunningly clever, has some odd quirks but is
incredibly reliable due to its age and design.

FreeBSD uses the BSD NET/3 code, from 4.4 BSD. This has some pretty serious
bugs but has good multicast support and other niceities like TCP PAWS
support. Most but not all of the bugs are in unusual events. It has most
of the odd limitations of BSD NET/2 as well - eg not being able to use
RAW sockets to listen to a protocol the kernel understands.

Linux uses its own scratch written code. As of 1.2.x the code is as reliable
as BSD (imho), and I run things like a big web server on it. Its also got
some odd quirks and as few bugs and limitations. In theory it has the scope
to be faster on PC hardware than the others, but optimisation to really
make it rip is a 1.3.x project

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  iia...@www.linux.org.uk   //  GW4PTS@GB7SWN.#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''
Redistribution of this message via the Microsoft Network is prohibited

From: pet...@telerama.lm.com (Peter Berger)
Subject: Re: freeBSD and Linux networking code (what is different?)
Date: 1995/06/01
Message-ID: <3qk3og$8bp@ivory.lm.com>#1/1
X-Deja-AN: 103630660
references: <3pnmm5$i82@felix.seas.gwu.edu> < D9EGyn.HHs@info.swan.ac.uk>
organization: Telerama Public Access Internet, Pittsburgh, PA USA
newsgroups: comp.os.linux.networking

In article < D9EGyn....@info.swan.ac.uk>,
Alan Cox < iia...@iifeak.swan.ac.uk> wrote:
>In article <3pnmm5$...@felix.seas.gwu.edu> ryen...@seas.gwu.edu (Richard D. Yentis) 
writes:
>>I am wondering what are the differences between the two versions of
>>BSD and the Linux networking code.  Are they all exactly the same?
>>are Linux and freeBSD the same but the other BSD is newer and/or
>>better?  Is there a historical tie between the code?  Can someone
>>point out who I might ask to find out?

http://www.freebsd.org
http://www.netbsd.org

>NetBSD uses the BSD NET/2 code, from 4.3 BSD series kernels. 

This statement is completely wrong, on both counts.  NetBSD is completely
4.4BSD Lite.

>This is not
>stunningly fast, not stunningly clever, has some odd quirks but is
>incredibly reliable due to its age and design.

"Fast" and "Clever" are relative terms, and the 4.4Lite code is certainly
both, compared to ... er, other, less well engineered networking 
implementations.  

>FreeBSD uses the BSD NET/3 code, from 4.4 BSD. 

There is no such thing as BSD Net/3.  This is the 4.4Lite networking code.
You know ... the stuff that all of those RFC's are based on?

>Linux uses its own scratch written code. As of 1.2.x the code is as reliable
>as BSD (imho), and I run things like a big web server on it.

I question your definition of a "big web server".

The proof of the pudding is under the crust.  If Linux networking code is 
more (or even as) as reliable as BSD, you should certainly be able to come 
up with a few examples of Linux machines performing under loads as 
rigorous as, say, ftp.cdrom.com (running FreeBSD 2.0) or the simply huge 
number of midsize ISP's running BSD/OS 1.1 or 2.0 from BSDi.  As of yet, 
I haven't heard of any.

>Its also got
>some odd quirks and as few bugs and limitations. In theory it has the scope
>to be faster on PC hardware than the others, but optimisation to really
>make it rip is a 1.3.x project

"It doesn't work yet, but when it does, watch out, world!"

I, frankly, don't know much about the Linux networking code except that
very few seriously loaded providers are using it.  However, it is worth
noting that the BSD networking code pretty much -IS- the Internet, 
and taking gratuitous (and inaccurate) shots at the other free Unices
as using "old" code is simply tacky.


-- 
"Actually, you just think that's a telephone.  Really, it's the alarm
that rings whenever I get out of my chair."  		-- E.S.
Peter Berger. System Administrator, Telerama Public Access Internet
http://www.lm.com/~peterb	    Serving Pittsburgh since 1991.

From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: freeBSD and Linux networking code (what is different?)
Date: 1995/06/12
Message-ID: < DA2AFK.LE@info.swan.ac.uk>
X-Deja-AN: 104324002
sender: n...@info.swan.ac.uk
x-nntp-posting-host: iifeak.swan.ac.uk
references: <3pnmm5$i82@felix.seas.gwu.edu> < D9EGyn.HHs@info.swan.ac.uk> 
<3qk3og$8bp@ivory.lm.com>
organization: Institute For Industrial Information Technology
newsgroups: comp.os.linux.networking

In article <3qk3og$...@ivory.lm.com> pet...@telerama.lm.com (Peter Berger) 
writes:
>This statement is completely wrong, on both counts.  NetBSD is completely
>4.4BSD Lite.

The NetBSD I have is 4.3 based. OK so its very out of date and I'm glad
to be put right.

>"Fast" and "Clever" are relative terms, and the 4.4Lite code is certainly
>both, compared to ... er, other, less well engineered networking 
>implementations.  

4.4BSDlite is _full_ of networking bugs. Connect/bind of raw socket 
is broken, you can't receive RAW socket UDP, TCP. You only raw receive some
of ICMP. The last one I looked at had errors setting sequence numbers when
it does the unofficial TIME_WAIT->SYN_RECV jump.

>>FreeBSD uses the BSD NET/3 code, from 4.4 BSD. 
>There is no such thing as BSD Net/3.  This is the 4.4Lite networking code.
>You know ... the stuff that all of those RFC's are based on?

On the contrary the 4.4lite code is customarily referred to as NET/3 by
many BSD people, Stevens included. Pity you are out of touch with
yourselves.

>>Linux uses its own scratch written code. As of 1.2.x the code is as reliable
>>as BSD (imho), and I run things like a big web server on it.
>I question your definition of a "big web server".

>The proof of the pudding is under the crust.  If Linux networking code is 
>more (or even as) as reliable as BSD, you should certainly be able to come 
>up with a few examples of Linux machines performing under loads as 
>rigorous as, say, ftp.cdrom.com (running FreeBSD 2.0) or the simply huge 
>number of midsize ISP's running BSD/OS 1.1 or 2.0 from BSDi.  As of yet, 
>I haven't heard of any.

Web servers taking 100-200,000 hits a day. FTP sites like ftp.idsoftware.com
- all those DOOM downloads, garbo.uwasa.fi. I can find you plenty of small 
ISP's running it. 

>>Its also got
>>some odd quirks and as few bugs and limitations. In theory it has the scope
>>to be faster on PC hardware than the others, but optimisation to really
>>make it rip is a 1.3.x project

>"It doesn't work yet, but when it does, watch out, world!"

It works nicely thank you 8). It doesn't have PAWS as 4.4lite does, and its
rtt estimator flakes out when you try and run 100Mbit/second. There are
things that need improving in all the stacks, I don't mind admitting the
Linux ones. There are Linux boxes as military DNS/mail hubs, Linux boxes
collecting data all over antartica, Linux boxes in belize....

Much of it is priorities. I considered fast appletalk phase 2 kernel
support, polished IPX, AX.25 and NetROM, IPIP tunnels, load balancing
across SLIP links etc more important than the polishes over the TCP layer
and adding VJ and friends clever little tweaks to get UDP over ATM up
over 80Mb/second checksummed. TCP polishing is the next logical step.
The end result should burn BSD off for speed and be more efficient. 
That bit is your "but when it does..."

>I, frankly, don't know much about the Linux networking code except that

Ah the traditional usenet 'I don't know much about, engage keyboard,
disengage brain'. ;) [large humour detection hint flag]

>very few seriously loaded providers are using it.  However, it is worth
>noting that the BSD networking code pretty much -IS- the Internet, 
>and taking gratuitous (and inaccurate) shots at the other free Unices
>as using "old" code is simply tacky.

BSD isn't 'old' code. Well no more than Linux in that its not using all the
clever work by people like Van Jacobson. Nowdays all the internet backbones
are held together by Cisco, 3com, BBN and similar TCP/IP stacks. Also 'old'
in the real world as opposed to academic world is not an insult, its 
reassuring.

BSD based networking for historical reasons holds much of the rest, although as 
things like Solaris 2.4 spread the grip will gradually loosen. You might want 
to ask why if the BSD networking is so perfect, and free for commercial users
that Microsoft wrote their own TCP for Windows/NT.

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  iia...@www.linux.org.uk   //  GW4PTS@GB7SWN.#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''
Redistribution of this message via the Microsoft Network is prohibited

From: n...@trout.sri.MT.net (Nate Williams)
Subject: Re: freeBSD and Linux networking code (what is different?)
Date: 1995/06/17
Message-ID: <3rtmve$brh@helena.MT.net>#1/1
X-Deja-AN: 104604711
references: <3pnmm5$i82@felix.seas.gwu.edu> < D9EGyn.HHs@info.swan.ac.uk> 
<3qk3og$8bp@ivory.lm.com> < DA2AFK.LE@info.swan.ac.uk>
organization: SRI Intl. - Montana Operations
reply-to: "Nate Williams" < n...@sneezy.sri.com>
newsgroups: comp.os.linux.networking

In article < DA2AFK...@info.swan.ac.uk>,
Alan Cox < iia...@iifeak.swan.ac.uk> wrote:

>BSD based networking for historical reasons holds much of the rest,
>although as  things like Solaris 2.4 spread the grip will gradually
>loosen. You might want  to ask why if the BSD networking is so perfect,
>and free for commercial users that Microsoft wrote their own TCP for
>Windows/NT.

Alan, if there is one thing we all know, is that Microsoft is the King
of Not Invented Here.  The Microsoft Motto "If it can be down, we can do
it better."

When you write your own implementation, you are free to 'tune' it for a
particular machine and/or application.  This means you can 'break' it
accidentally (see WNT newsgroups for TCP/IP breakage) and claim it was a
bug.  Now, I'm not saying they intentionally break TCP/IP, but as you
pointed out priorities are more important to make it work with uSoft
products than to make it work with the rest of the world.

And, you're statement that "you'll *burn* the BSD networking code soon"
is laughable.  As with any of the 'my car is faster than your car'
arguements, you assume your competition is sitting still while you get
the kinks worked out of your system.

The attitude that "Why use your code when I can do it better than you
can" is the most annoying feature of both Microsoft and many free OS
developers.  If it works, use it rather than spending 2 years of your
life building/fixing something that still isn't as robust as the BSD
TCP/IP code.  If you spent all of the effort fixing the brokeness in the
original code, *everyone* would be better off as both Linux and 4.4BSD
crowds would have a much better protocol stack.


Nate
-- 
n...@sneezy.sri.com    | Research Engineer, SRI Intl. - Montana Operations
n...@trout.sri.MT.net  | Loving life in God's country, the great state of
work #: (406) 449-7662 | Montana.  Wanna go fishing?  Send me email, and we'll
home #: (406) 443-7063 | setup something.

From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: freeBSD and Linux networking code (what is different?)
Date: 1995/06/29
Message-ID: < DAxtx6.Evn@info.swan.ac.uk>#1/1
X-Deja-AN: 105270453
sender: n...@info.swan.ac.uk
x-nntp-posting-host: iifeak.swan.ac.uk
references: <3qk3og$8bp@ivory.lm.com> < DA2AFK.LE@info.swan.ac.uk> 
<3rtmve$brh@helena.MT.net>
organization: Institute For Industrial Information Technology
newsgroups: comp.os.linux.networking

In article <3rtmve$...@helena.MT.net> "Nate Williams" < n...@sneezy.sri.com> 
writes:
>And, you're statement that "you'll *burn* the BSD networking code soon"
>is laughable.  As with any of the 'my car is faster than your car'
>arguements, you assume your competition is sitting still while you get
>the kinks worked out of your system.

Well Van Jacobson told the world to do checksum & copy in the early 1980's
and BSD is basically standing still compared with commercial BSD based stack
people like SGI. Solaris 2.4 has faster networking than BSD even when it
is using streams as a core. If BSD keeps on moving and there is an 
incentive to competition for both parties so much the better.

>The attitude that "Why use your code when I can do it better than you
>can" is the most annoying feature of both Microsoft and many free OS
>developers.  If it works, use it rather than spending 2 years of your
>life building/fixing something that still isn't as robust as the BSD
>TCP/IP code.  If you spent all of the effort fixing the brokeness in the
>original code, *everyone* would be better off as both Linux and 4.4BSD
>crowds would have a much better protocol stack.

Two fundamental errors in your argument

1.	The BSD licensed code arguably cannot be used in Linux otherwise
I'd have used the BSD TCP layer with the Linux lower layers. The regents of
UCB didnt even bother to answer my letter.
2.	Anyone can pick up the BSD code improve it and keep the improvements
secret. I work on the Linux TCP as a free software project. If you want me
to work on non GPL projects like BSD 4.4 TCP you can email me on on my
commercial address and discuss rates.

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  iia...@www.linux.org.uk   //  GW4PTS@GB7SWN.#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''
Redistribution of this message via the Microsoft Network is prohibited

From: f...@oberon.pps.pgh.pa.us (John Fail)
Subject: Linux vs. FreeBSD
Date: 1995/05/30
Message-ID: <3qfhhv$7uc@titania.pps.pgh.pa.us>#1/1
X-Deja-AN: 103496139
organization: Pittsburgh Public School District
newsgroups: comp.os.linux.advocacy


What are the differences between Linux and FreeBSD?  What make one
better than the other?  I've always wondered . . .

-john
p.s.I run Linux because it seems to be supported more on the Net, and
also I got DOOM for it.

From: p...@zeus.fasttax.com (Phil Howard)
Subject: Re: Linux vs. FreeBSD
Date: 1995/05/30
Message-ID: <3qfotu$els@zeus.fasttax.com>#1/1
X-Deja-AN: 103496150
references: <3qfhhv$7uc@titania.pps.pgh.pa.us>
organization: fasttax.com
newsgroups: comp.os.linux.advocacy

f...@oberon.pps.pgh.pa.us (John Fail) writes:

>What are the differences between Linux and FreeBSD?  What make one
>better than the other?  I've always wondered . . .

I believe you will find Linux leans to POSIX and FreeBSD leans to BSD.
It depends on what you want in a system.  Both work well as far as I
know.  Linux appears to be the leader for reasons that are probably
not all that related to technical details.  Sometimes the technically
superior wins (which IMHO Linux is over every UN*X out there) and
sometimes not (DOS/Windows, VHS, etc).  If the world judged things
on technical superiority, the world would be different and salesmen
would be unemployed.

Among the reasons I chose Linux, which I did before it had the big lead
over FreeBSD (and I've always preferred BSD over SYSV) were the copyright
clauses and the committment to POSIX (I don't like many things in POSIX,
but standards usually do have to be compromises).  I like the variety of
filesystems it can support (I could go add my own for instance).

Today Linux appears to be rock solid and stable (unless you are tracking
the odd kernels, but that's to be called an adventure anyway).  I'll have
to buy a UPS in order to keep from rebooting it every now and then.  And,
distribution packages like Slackware work quite well.


>p.s.I run Linux because it seems to be supported more on the Net, and
>also I got DOOM for it.

There are tons of programs on the net for DOS/Windows.  You can get DOOM
for DOS/Windows.  So why do you REALLY use Linux?

-- 
Phil Howard KA9WGN      | When citizens fear government, we call it tyranny
Unix/Internet/Sys Admin | When government fears citizens, we call it freedom
CLR/Fast-Tax            | It really says "...the right of the PEOPLE to..."
p...@fasttax.com        | This is not my only SIG.  My other one is a P226.

From: s...@amber.epcc.ed.ac.uk (Scott Telford)
Subject: Re: Linux vs. FreeBSD
Date: 1995/06/02
Message-ID: < D9K30I.F9t@dcs.ed.ac.uk>#1/1
X-Deja-AN: 103630030
sender: cn...@dcs.ed.ac.uk (UseNet News Admin)
references: <3qfhhv$7uc@titania.pps.pgh.pa.us> <3qfotu$els@zeus.fasttax.com>
organization: Edinburgh Parallel Computing Centre, University of Edinburgh, UK.
newsgroups: comp.os.linux.advocacy

Phil Howard (p...@zeus.fasttax.com) wrote:
: f...@oberon.pps.pgh.pa.us (John Fail) writes:

: >What are the differences between Linux and FreeBSD?  What make one
: >better than the other?  I've always wondered . . .

: I believe you will find Linux leans to POSIX and FreeBSD leans to BSD.

NetBSD is pretty much POSIX.1 compliant, and I expect FreeBSD to be
similar. POSIX.1 compliance is no big deal these days.

--
Scott Telford, Edinburgh Parallel Computing Centre,        
University of Edinburgh, Mayfield Rd, Edinburgh, EH9 3JZ, UK.(+44 131 650 5978)
we came in?....................................................isn't this where

From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Linux vs. FreeBSD
Date: 1995/06/05
Message-ID: < D9p65M.FGu@info.swan.ac.uk>#1/1
X-Deja-AN: 103831594
sender: n...@info.swan.ac.uk
x-nntp-posting-host: iifeak.swan.ac.uk
references: <3qfhhv$7uc@titania.pps.pgh.pa.us> <3qfotu$els@zeus.fasttax.com> 
< D9K30I.F9t@dcs.ed.ac.uk>
organization: Institute For Industrial Information Technology
newsgroups: comp.os.linux.advocacy

In article < D9K30I....@dcs.ed.ac.uk> s...@amber.epcc.ed.ac.uk (Scott Telford) 
writes:
>NetBSD is pretty much POSIX.1 compliant, and I expect FreeBSD to be
>similar. POSIX.1 compliance is no big deal these days.

POSIX compliance is a big deal to win contracts and to be allowed to tender
for things. I imagine CD-ROM vendors of Linux/NetBSD/FreeBSD are in the end
going to get involved in this. The problem is the cost.

Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  iia...@www.linux.org.uk   //  GW4PTS@GB7SWN.#45.GBR.EU  //
 ``----------'`----------------------------'`----------------------------''
Redistribution of this message via the Microsoft Network is prohibited