[Ns-developers] [ns3] a synchronous socket/posix API

[Sam Jansen]

On 05/04/2008, Mathieu Lacage <mathieu.lacage at sophia.inria.fr> wrote:
>
>
> On Sat, 2008-04-05 at 18:21 +0100, Gustavo Carneiro wrote:
>
> >
> > So, very good intentions, but the problem is more complex to solve
> > fully. :-/
>
>
> I just put in something which works, is simple and never claimed I
> solved the problem fully. Feel free to add more fancy delay calculation
> models if you are interested in this by adding extra subclasses of this
> base class.
>

I think it is fine for a basic implementation, and might well be of use to
many people.

However, could the default be for a "no delay" model?

It seems to me that elsewhere in the simulator, and in general with network
simulation, the assumption is made the CPU time is not modeled. If a user
wants to model processing time, this should be a conscious decision, and the
user should be aware of which model they are picking. Silently modeling
delay based on actual recorded wall time here also makes the simulations
non-deterministic by default, which sounds like a bad idea to me.