[Iccrg] benchmarking new tcp congestion control algorithms

[Douglas Leith]

> It seems to me that the scenarios should (i) be generic  
> (exhaustive?) enough to capture most common ones, and at the same  
> time (ii) be small enough in number that reporting results for  
> these tests does not take up 500 graphs and therefore encourages  
> folks to do more than just these tests.
>
> I love the idea of scripting these into ns.
>

For what its worth, I'd echo Mark and Wesley's sentiments.

Re ns scripting, some of the groundwork is already done (including  
html interfacing to see the results e.g. so they can be easily placed  
on the web).  Take a look at http://www.hamilton.ie/net/eval/ 
tcptesting.zip.  Its all pretty simple shell scripts, so very open  
and easy to extend.  As there's a clean interface to ns, its also  
easy to recycle this code to run experimental tests.

Doug