Re: [ns] Trust on the simulator

[Sally Floyd <floyd@aciri.org>]

Daniela -

>I want just an opnion about the reliability of the NS simulator.
>
>If it has been available for so long and it has been used for so many
>people, if I just use it without making any changes, couldn't I
>guarantee the reliability of the results I get from it?

Nope, you can not "trust" the results that you get from the NS
simulator, without verifying for yourself that it is performing
the way that you are expecting it to perform.  You can trust the
validation tests, which illustrate specific behaviors (e.g., TCP
Brand X halving its congestion window after a packet drop).  But
you can't assume that there are no bugs at all, and you can't assume
the model implemented in the simulator matches your needs and
expectations.  

Just as an example, the one-way TCP models, that I use myself,
essentially assume that the header size is zero bytes, because the
number of bytes of data consumed from the application is the same
as the number of bytes of data that appear on the wire.  A historical
artifact from the earlier evolution of the simulator, but not too
good, if that fact happens to matter to your simulation, unfortunately.
And the two-way TCP models in Agent/TCP/FullTcp, which do take into
account header size in bytes, are not extensively validated in
other aspects.

Because of so many people who have used NS, you can probably trust
that the well-used modules don't have major bugs that would obviously
invalidate your results.  I wouldn't recommend that you trust any
farther than that, myself.  (I wrote the "Read this first:" section
at the top of the NS web page, and I would contend that it still
holds.)

- Sally
--------------------------------
http://www.aciri.org/floyd/
--------------------------------