Looking for comparison data on network stack prowess

Looking for comparison data on network stack prowess

[David Ford]

<noflame=1>
I'm looking for some authoritative comparisons and discussions of the
current network stacks in *BSD and Linux.  I.e. NET4 in Linux and
whatever is most current in *BSD.

_PLEASE_ no flaming, no causing flamewar, nadda.

I am writing an article for Linux.com and I am attempting to debunk
longstanding fallacies on both sides of the camp.  I am aiming for a
truely neutral article which means I want to hear about the bad as well
as the good for both camps.

I am no master, and haven't played with *BSD in a few.  I would
appreciate any of you who can cooly speak their mind and provide
insightful information.

I am looking for:
    articles
    benchmarks
    commentary
    references
    etc..

Thank you,
-d

--
  There is a natural aristocracy among men. The grounds of this are virtue and talents. Thomas Jefferson
  The good thing about standards is that there are so many to choose from. Andrew S. Tanenbaum



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Looking for comparison data on network stack prowess

[Albert D. Cahalan]

David Ford writes:

> I'm looking for some authoritative comparisons and discussions of the
> current network stacks in *BSD and Linux.  I.e. NET4 in Linux and
> whatever is most current in *BSD.
>
> _PLEASE_ no flaming, no causing flamewar, nadda.
> 
> I am writing an article for Linux.com and I am attempting to debunk
> longstanding fallacies on both sides of the camp.  I am aiming for a

Start with the history: Net3 is not net3. Net2 is not net2.
Linux did not steal the BSD stack. I recall that Alan Cox
politely asked UCB to have it under the GPL, and was refused.

It amazes me how Linux can be accused of stealing BSD network
code while also being said to have poor network code... guess
that means we broke it?

Oh, BTW, BSD was _not_ the first OS with IP. The first was some
horrid mainframe thing. Sometimes, he who codes last codes best.

> truely neutral article which means I want to hear about the bad
> as well as the good for both camps.

Fair? Then this must be an equal-budget competition. SPECWeb99 is
just that, with "infinity" as the budget. I think it has to be noted
that BSDI has not accepted the challenge. That's just performance
though, which isn't a problem for most people.

I propose a feature and ease-of-use compatition. Each group gets
to suggest a few interesting routing and firewalling problems.
Maybe have a few Cisco and Microsoft fans try to stump us too.
Then each group tries to find a solution that is fast, reliable,
easy to understand, safe, easy to implement, and easy to maintain.
The use of non-standard tools and patches is tolerated, but it
greatly reduces your rank.

(solutions can be tossed into a HOWTO as well)

Example:
You have a home LAN with one fixed IP address on an ISDN line,
and another fixed IP address on a DSL line. You have domain names
that point to the ISDN line's IP, and you want to convert over
to using DSL exclusively. Connections initiated from outside ought
to go out the way they came in (with the right IP!), and connections
initiated from inside should go out the DSL line.

There are plenty of ways the groups could challenge each other:
PPPoE with dynamic IP assignment and decent firewall rules,
bridging all packets with a Cisco MAC address, IPv6-to-IPv4,
plain dial-on-demand into a strongly Microsoft-centric ISP,
VLANs, a VPN, AppleTalk and IPX support, ECN bit removal,
policy routing and bandwidth reservation...


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Looking for comparison data on network stack prowess

[Alan Cox]

> Linux did not steal the BSD stack. I recall that Alan Cox
> politely asked UCB to have it under the GPL, and was refused.

Start with the right history then

Ross Biro did the original Linux networking code. At the time the 386BSD code
was potentially useful but two things occurred

1.	I asked a real lawyer about mixing BSD and GPL code and got told the
	advertising clause was an additional restriction

2.	BSDI got sued, making the entire BSD codebase potentially contaminated

Someone in .de (Alas I forget their name now) actually did port BSD net/2 to
Linux.

FvK took over and then I took over and we had net2debugged (no relation to
BSD net/2) and then net/3 and net/4 over time.

The 1.0 networking code worked but certainly wasnt BSD grade, the 1.2 code
worked better but wasnt BSD grade. 2.0 was certainly on a par and 2.2/2.4
have added a lot of other stuff. *BSD has also not stood still.

You can certainly find cases where either is better.

> Oh, BTW, BSD was _not_ the first OS with IP. The first was some
> horrid mainframe thing. Sometimes, he who codes last codes best.

Humph. TOPS-10 is a beautiful OS.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Looking for comparison data on network stack prowess

[Andi Kleen]

Alan Cox <alan@lxorguk.ukuu.org.uk> writes:

> Someone in .de (Alas I forget their name now) actually did port BSD net/2 to
> Linux.

Matthias Urlichs iirc

He also later implemented a minimal STREAMS clone on Linux for his
ISDN stack. 

[and today Linux is reinventing non shouted streams with netfilter..]

-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/