[Ns-developers] ns-3 projects for gsoc

[Mathieu Lacage]

On Thu, 2008-03-20 at 15:50 -0400, Raj Bhattacharjea wrote:

> In regards to the "Linux kernel network stack integration" project
> mentioned, we've discussed the network simulation cradle before
> (http://research.wand.net.nz/software/nsc.php), and Tom had done some
> preliminary work (http://code.nsnam.org/tomh/ns-3-nsc/), although it
> looks like it hasn't been touched months.

I agree that these should be mentioned on that page. Feel free to edit
it.

> 
> Is the goal of this GSOC project to have a student start from scratch
> on a port of the Linux stack into ns-3?  That is probably duplicating
> a lot of work the NSC people have already done...we could broaden the
> scope of this project to mention/include NSC.

There are lots of issues to consider before choosing an implementation
path:

  - some of the nsc implementation is really scary. The globalizer most
notably.

  - the nsc implementation has solved some problems you probably don't
want to solve again.

  - what is your final goal ? Do you care about multiple kernel stacks ?
Are you interested in linux only ? Which version of the linux kernel ?

  - what kind of maintenance effort is going into nsc ? What is the
future of nsc ?

Some of these questions have simple answers. Others are much harder to
answer. Sam Jensen who is the nsc original author dropped by on irc
yesterday and we talked about these. 

In the end, the technical solution chosen will really be up to the
student who tackles this project (if there is one) but that decision
certainly won't be easy or uncontroversial. i.e., Depending on what your
final goal is, there are sufficiently many downsides to nsc that
starting again from scratch or forking it might be a sound decision.

my 2 cents,
Mathieu