I follow up this discussion a while and don't know why there are so many
concerns:
1st: it is a _module_ i.e. It does not bloat the kernel!
2nd: it would help embedded stuff very much
3rd: when you don't need it, don't load the module
It's that simple ;-)
Anyway, it's just another option
Let's finish this guy with a nice implemention and then we'll see if it
is worth including into stock kernel.
Marcus
Marek Habersack wrote:
>
> * Chris Smith said:
> > Dan Hollis wrote:
> > > Exactly. In-kernel http is lower memory profile, simpler, and faster.
> > > All are major issues in embedded systems.
> >
> > Hate to burst this bubble, but what is the use of a completely static web
> > server on an embedded device? None. If you can't use dynamic pages then
> > you can't report status, and if you can't handle some kind of CGI or
> Wrong. You can use Java applets to report status - that's what 3Com's
> devices do.
>
> > whatever else, then you can't do dynamic configuration. Which means you'll
> Oh, you can. If you're talking about configuration of the WWW interface -
> cookies are there for you, if you're talking about the configuration of the
> device itself, internal variables etc. are modified and stored on the flash
> media. Where's the need for dynamic pages?
>
> > need a CGI-capable web server running in user space anyway to do useful work
> Why?
>
> > in an embedded device. Thus, no memory savings. The extra memory for
> > in-kernel khttpd might even matter on an embedded device, so I doubt it
> > would get used anyway.
> I think it would get used - memory savings would be quite significant, as
> well as the time saved on context switches and the like.
>
> > Sure, you might be able to make things marginally faster -- just remember
> > there is a cost there, too. There's a reason that all processes don't run
> > at privelege 0, and that's because it makes the system less reliable.
> You are talking about an environment where there are a lots of user-space
> processes, many people connect to the server, many other servers talk to it
> etc. - none of this is true with an embedded system. And if you have a
> kernel with most of the services the given embedded system needs, asssuming
> they are well written, you won't have the problems you just envisioned. And
> try to stuff a full-blown httpd with CGI, FastCGI and whatsnot into a gas
> distributor with an embedded OS (and don't laugh, such things with Linux
> inside work in the Netherlands :)).
>
> marek
>
> ------------------------------------------------------------------------
> Part 1.2Type: application/pgp-signature
-
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/