discuss: Modifying the kernal for implementation of ad hoc routing protocol
Subject:
Re: Modifying the kernal for implementation of ad hoc routing protocol
From:
Glen Turner ####@####.####
Date:
21 Jan 2003 04:02:11 -0000
Message-Id: <3E2CC65D.6000102@aarnet.edu.au>
Abhinav Gupta wrote:
> I am a student on development of routing protocol for ad hoc networks.
> For implementation, I am planning to use laptops running Linux 7.3 with
> wireless capability (IEEE 802.11b). Could someone please give me
> pointers to modifying the kernal for the implementation of the routing
> protocol?
You might note that the Linux kernel only forwards packets.
Routing protocols and other topology discovery processes are
run in user space. For current user-space routing daemons
see Zebra and gated, although you'll need to ask nicely for
a gated license.
So you might not need a kernel modification at all, rather
just use of the kernel's "netlink" socket facility to
maintain the routing table.
The IETF are strongly promoting the seperation of "forwarding
plane" (kernel) and "control plane" (router daemon). See
the IETF ForCES working group, and especially
Netlink as an IP Services Protocol
<http://www.ietf.org/internet-drafts/draft-ietf-forces-netlink-04.txt>
Because of the long-standing philosophy of both the UNIX
community ("policy should not be in the kernel") and the
IETF community ("control plane and forwarding plane should
be seperated") that routing protocols do not belong in a
Unix kernel you've got stuff-all chance of getting a
routing protocol patch into Linus's, Dave's or Alan's Linux
kernel.
Have a look at the Zebra source code. In fact Zebra
provides a nice software platform for writing any
IP routing protocol. See http://www.zebra.org/.
This really isn't the right mailing list -- you want
some of the the IETF's mailing lists. Particularly
ans-research (ANS being "Ad-hoc network scaling")
forces
manet (Mobile ad-hoc networks).
Any new routing protocol is ging to have to meet the
forthcoming requirements from the Routing Protocol
Security WG. So you may as well join their mailing
list too.
Zebra also has a mailing list.
There are some people doing some interesting research in
"large flat networks" and in "ad-hoc networks". If your
work is going to have a tangible result (eg, it is more
than a semester project) then you could e-mail them nicely
asking for some introductions to other people doing similar
projects.
Similarly, most of the North Americans developing routing
protocols will appear at the IETF meetings. The next one
is quite close to you -- in San Fransisco 16-21 March.
The attendance fee for students is US$150 (if paid by 7
March), something your institution might be able to cover.
You don't need to stay at the IETF-arranged accomodation,
best to ask a friend in SF about alternatives, bearing in
mind that you don't want to be too far away as the Birds
of a Feather meetings can run quite late.
If you can't attend, then at least read the Internet Drafts
that might interest you and get a copy of the slides from
the WG meetings.
Staying across the IETF happenings in your area of research
will ensure that your research has the maximum chance of being
relevant, not reinventing someone else's work, and be an
approach discarded due to fundemental or practical flaws.
Best wishes,
Glen
--
Glen Turner (08) 8303 3936 or +61 8 8303 3936
Australian Academic and Research Network www.aarnet.edu.au
