Re: nam bug when using random loss modules

[Haobo Yu <haoboy@isi.edu>]

> Could you please explain what is the "snapshot". I checked for it on the
> ns homepage, but I couldn't find a mention of a snapshot.

They are daily extracted from the cvs repository of ns and nam. Available
at http://mash.cs.berkeley.edu/dist/vint/, look for *-current.tar.gz. 
 
> I would like to install the error model after the queue using lossmodel.
> Could you confirm that this Simulator::lossmodel would work for a duplex
> link?

Yes it works. You can find more information in the Error models section in
the most current nsdoc
(http://mash.cs.berkeley.edu/ns/ns-documentation.html). 

- Haobo
 
> Thanks,
> 
> -Hussein
> 
> > 
> > I think it should have been fixed in the current nam (and ns) snapshot.
> > Previously Simulator::lossmodel and SimpleLink::errormodule does not
> > generate any nam traces. You can manually add drop traces of error models,
> > however, nam does not show that correctly because of abnormal the
> > enqueue/drop order.
> > 
> > There are two solutions. First, if you are using ns snapshot you can use
> > Simulator::link-lossmodel to install the error model (which installs
> > an error model after the queue in a simple link), which allow any nam to
> > view the trace correctly. Second, if you still prefer installing an error
> > model before the queue in a simple link, you can download the current nam
> > snapshot, which contains a fix for that. 
> > 
> > - Haobo
> > 
> > On Tue, 9 Mar 1999, AL-HUSSEIN ABOU-ZEID wrote:
> > 
> > > 
> > > I've noticed that when you use nam with a model containing a two-state
> > > time-based loss model, the animation works correctly up till the first
> > > random loss, and after that it does strange things. One of these is that
> > > it shows a buffer size that is greater than the maximum buffer size
> > > allowed on the link (I am assuming this is a nam bug and not an ns bug).
> > > Another problem is that, after the first random loss, it might not show
> > > any packets getting through the link, as if the link remained in the bad
> > > state for the rest of the simulation time. However, this proves not true
> > > when checking the trace file, which implied a bug in the nam and not the
> > > ns.
> > > 
> > > I understand that the two-state and multi-state models are
> > > "under-developed" in the version I'm using (ns-v2.1b4a), so maybe someone
> > > would be interested in taking this problem into consideration in future ns
> > > releases.
> > > 
> > > Best Regards,
> > > 
> > >  --Hussein.
> > > ------------------
> > > Electrical Engineering Department
> > > University of Washington, Seattle.
> > > 
> > 
> > 
> 
>