Why userspace rather than writing kernel space modules?
Ondrej Filip
feela at ipex.cz
Tue Jun 12 11:00:36 CEST 2001
On Mon, 11 Jun 2001, Pavel Machek wrote:
> Hi!
>
Hi!
> > I'm a newbie to BIRD who ran across it from freshmeat.net.
> >
> > Being a newbie, I have some newbie questions.
> >
> > How efficient is it to sync with kernel space routing tables if the
> > routing tables change rapidly?
>
> Efficient enough.
>
> > How often does a routing table change with a typical server
> > configuration and typical server traffic?
> > Lastly besides the portability, why implement a userspace daemon that
> > is required to sync with the kernel when kernel space modules could
> > be written to directly manipulate the routing tables themselves?
>
> Why write X windows when you could integrate them in kernel (like
> winNT do)?
>
> a) portability
>
> b) crash-proofness
>
> c) bird is swappable
>
> d) bird can use libc
>
> e) bird is preemptibly-scheduled. If it loops, nothing bad happens.
>
You are true. But if you build HW routers based on linux, the kernel crash
or the bird crash are the same problems.
Today's BGP adds about 100 000 items into routing table. It uses a lot of
memory and I'm not sure, that the transfer is so efficient.
> Pavel
>
Kind regards
Feela
More information about the Bird-users
mailing list