Date: Tue, 22 Nov 1994 05:21:41 +0200
From: mitn...@erez.cc.biu.ac.il (Roman Mitnitski)
Subject: Any chance to see BSD networking on Linux?
Organization: Bar-Ilan University, Ramat-Gan, Israel.
X-Newsreader: TIN [version 1.1 PL9]
Hi. I guess it was duscussed many times, but...
There's lot of talking that Linux networking software is far from
being reliable, robust, etc etc. On the other hand, we have FreeBSD's
networking, which is told to be very stable and even fast.
I think I can understand why separate development of networking software
was started - due to infamous USL - BSD lawsuit(?)
Now, I think competition is great, why don't give BSD networking software
a chance in Linux? If net-2 (linux) will be better, it will survive...
now, when USL lawsuit is over....
Flame me if I wrong!
"Everyone is entitled to an *informed* opinion." -- Harlan Ellison
From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Any chance to see BSD networking on Linux?
Message-ID: < CzzA65.H19@info.swan.ac.uk>
Organization: Institute For Industrial Information Technology
Date: Mon, 28 Nov 1994 12:46:53 GMT
In article <19941122.061645.560783.NETN...@VM.BIU.AC.IL>
mitn...@erez.cc.biu.ac.il (Roman Mitnitski) writes:
> There's lot of talking that Linux networking software is far from
>being reliable, robust, etc etc. On the other hand, we have FreeBSD's
>networking, which is told to be very stable and even fast.
The FreeBSD one is derived from the original BSD code and is thus very BSD
compatible 8). There are small bits of the BSD code I would love to use
(basically the tcp proessing machine) but my legal advice was that I couldn't
mix BSD with GPL'd code in one program without infringing the BSD license
requirements. I'm not likely to get sued or anything if I did, its not likely
UCB would object but the consequences if they did are a little too drastic.
Now if someone could get Regents of University California to release the tcp*c
files in 4.4BSD lite without the requirement for accreditation on
documentation I'd drop the tcp state machine into the Linux code which is
basically sound apart from in this area. They have however never even
replied to my email.
I've spent a lot of time patching up the tcp layer knowing it needs
rewriting one day but hoping to put it off until after 1.2, Matt Dillon has
worked even more on it, and Charles Hedrick patched it to resemble tcp in
the 0.99.15x releases. We do need a new TCP, but the other layers I think
are best developed to completion. The fast linear buffers in the current
Linux code and the socket hashing algorithm are already superior to mbufs
and the one hit cache BSD uses.
Alternatively who out there feels like writing a tcp protocol layer 8)
One thing I am testing now is the BSD4.4 IP firewall code which was under
a much freer license and thus usable. This is working rather nicely now so
expect it soon. I've also got IP options working 90% of the time and a
unix domain socket family that does datagrams right. It's all a case of
testing time and trying to decide if something should go to Linus for 1.2
or wait. [Personally I'd like to see 1.3.x started now and 1.1.x mature
into 1.2 in parallel so that we don't get huge bottlenecks and testers
can let rip on new schedulers, the EIDE stuff, extra drivers, proper NFS
caching etc now].
> I think I can understand why separate development of networking software
>was started - due to infamous USL - BSD lawsuit(?)
Correct. The only BSD based piece is the header compression which is from
the RFC rather than the BSD source code.
>Flame me if I wrong!
I think partly right, partly wrong. Certainly not flamably stupid 8)
// Alan Cox // iia...@www.linux.org.uk // GW4PTS@GB7SWN.#45.GBR.EU //