[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ns] wrong number of wnd_init_option 0
This problem was caused by Vegas not calling reset() in the right place
that the base tcp assumes. It's been fixed about half year ago (or at
least a problem with the same symptom).
- Haobo
On Thu, 7 Sep 2000, Lloyd Wood wrote:
> On Thu, 7 Sep 2000, Pablo Souki Mauricio Jiminez wrote:
>
> > When running a simulaton involving many nodes,
> > TCP/Vegas Cries out: "wrong number of wnd_init_option
> > 0". We know this variable is supossed to be one or
> > two, so we don�t understand where that cero comes
> > from.
>
> I've seen this too. It's puzzling; seems to be a problem with delayed
> binding of variables, since vegas is the only TCP derivative to call
> delayed binding routines again. (This is assuming that having
> increased initial windows in vegas makes sense and is intended.)
>
> if you look in tcp.h, you'll see that vegas has its own protected
> delay_bind methods for its own added variables. I bet this call in
> vegas' delay_bind_dispatch() has something to do with that zero
> cropping up:
>
> return TcpAgent::delay_bind_dispatch(varName, localName, tracer);
>
> tcp.cc's initial_window() says:
>
> else if (wnd_init_option_ == 2) {
> // do iw according to Internet draft
>
> that comment should be RFC2414 these days, I think.
>
> L.
>
> and wnd_init_option isn't called wnd_init_option_ in Tcl. Grmbl.
>
> <[email protected]>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>
>
>