Re: linux-kernel-digest V1 #4966

Robert Dinse (nanook@eskimo.com)
Wed, 22 Dec 1999 22:15:00 -0800 (PST)

On Wed, 22 Dec 1999 "Dwayne C . Litzenberger" <dlitz@cheerful.com> wrote:
>
> On Thu, Dec 23, 1999 at 01:31:42AM +0100, Dominik Kubla wrote:
> > On Wed, Dec 22, 1999 at 05:58:16PM -0600, Dwayne C . Litzenberger wrote:
> > > Does someone have Linus's take on why on earth he let something that
> > > should definitely be in user-space (the httpd) into the kernel? It see=
> ms
> > > absurd to me, and odd that Linus would allow it.
> >=20
> > Performance. The apache/khttpd combo beats the shit out of NT in most
> > benchmarks.
>
> Maybe so, but why don't we put Quake into the kernel, too? You would get
> great performance, but something like that simply belongs in user space. I
> don't think Linux deserves the benchmarks it gets if it cheats to get
> them. We shouldn't let politics allow us to employ bad design (If a
> registry ala Windows was a performance issue, would we use it? No,
> because it makes a mess of everything).=20
>
> An HTTP server is something that should be in a patch, not the main kernel
> tree. (Heck, software suspend and the PC-speaker driver should go in
> before khttpd does.)

I disagree. THE net application is HTTP server. Nothing is more critical
to a presence on the Internet than a robust effecient web server. It is SO
widely used that it deserves some kernel space if it buys a substantial
performance benefit. It's an OPTION, you don't HAVE to configure it in.

In the Linux/NT wars; the ability to provide a better performing web
server, not just in benchmarks, but in real world applications on the net is
going to be a major factor. I don't see embedded applications as "cheating".

One could just as easily argue that TCP/IP should be in user space not in
kernel space, and in fact I've seen some Sys 5r2 systems that DO implement it
as a deamon. But substantial performance benefits result from it being
incorporated in the kernel, and the same is true of the HTTP, and it's becoming
a common enough application that that protocol deserves kernel space. It's
usefulness is limited, but it at least gives you a way to serve static content
fast.

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