[Ns-developers] GSoC TCP Congestion Avoidance

[Tom Henderson]

On 3/31/10 4:56 AM, Alex Borisov wrote:
> Hi!
>
> I think I wasn't clear about my question. Sorry but I'm not fluent in
> English.
> I mean that in this post about TCP Validation project
> http://mailman.isi.edu/pipermail/ns-developers/2010-March/007593.html you
> give sources where
> Tahoe, Reno, and SACK TCP validation process is described. Reno TCP and SACK
> TCP aren't implemented in ns-3 yet. So these test suites should be
> implemented as part of TCP Congestion Avoidance project, isn't it? But Tahoe
> variant is already implemented. Is Tahoe test suite should be implemented as
> part of TCP Validation or Congestion Avoidance project?
>
> Thanks!

Alex,
My suggestion to see the related work was not necessarily to imply that 
we repeat exactly those tests, because as you observed, the ns-3 model 
is not the same right now.  We have a TCP Tahoe variant with fast 
retransmit.  I also suggested:
     While the goal would not be to add a new model but to improve an
     existing model, I expect that along the way, some small additional
     features of the TCP model would be added.  The benefit to
     pursuing this project is that the student would learn about
     sockets, TCP, and about how to test simulation models.
Adding some basic variants like Reno, Newreno, Sack could be put into 
scope.  Or, for instance, TCP timestamps.  The workflow that I have in 
mind is that the student would start by validating what we have, and 
then as it becomes clear that we cannot test for certain features 
because they are lacking, the student and mentor can discuss whether to 
spend some time to add a particular new feature and corresponding tests.

Tom