[Ns-developers] Pre-review request: csma module improvements (propagation delay and carrier sense)

[Vedran Miletić]

Datuma 1. studenoga 2011. 08:45 Mathieu Lacage
<mathieu.lacage at gmail.com> je napisao/la:
> I am not sure what you mean by the above but it is actually a feature
> to be able to relate the propagation delay to a real distance because
> this is what most csma/cd courses will mathematically demonstrate: the
> relationship between the upper bound on the length (given a uniform
> propagation speed) of a non-switched ethernet cable and the csma/cd
> protocol itself. So, we do want actual meters.
>

It is. But not in the way that I specify both separately, i.e. 500m
cable and 1 second propagation delay. If I make a model that expects a
user to specify cable length, I want propagation delay to be
calculated according to what I specified.

Otherwise, you will require students to calculate one or the other
manually or, worse, specifiy both as an arbitray value and get
propagation speed of 100 m/s.

I agree with Lloyd's proposal here:

> The correct thing to do is to define the refractive index for the channel, which gives the speed of light in
> the medium. Propagation delay across it then derives from set distances between nodes/speed of light.

Whether this is an academic issue or not I will let everyone to decide
for themselves.

> Beyond this academic issue though, the real issue is a soft eng issue.
> If a model needs to model positions and distance, it should use the
> ns-3 position models, not a model-specific thing.

You are completely right on this one. I will take a look. Thanks for a
pointer in the right direction.

Vedran