[Ns-developers] Visualization library

[Gustavo Carneiro]

On 17/03/2008, Jeremy <jnorman at mines.edu> wrote:
>
> Sorry it took me so long to respond, it's spring break  for us and I was
> out
> of office.
>
> Just to clarify what we are hoping to have occur is to have iNSpect
> distributed with ns-3 as a standalone piece of software. That being said,
> for now we are planning on focusing on continuing development of iNSpect
> as
> a standalone package rather than switching our version control system.
> However, anyone who would like a copy of our working code, I'd be happy to
> send it to you directly or if you would like to be more involved, we can
> figure out a solution.


>From the videos, iNSspect looks promising.

ns-3-viz is only less than 400 lines of code, right now.  It will probably
grow some day, but still it is a rather "lightweight" visualizer.  iNSpect
looks more "enterprise-y" code.  It is not clear to me how does iNSpect
integrate with NS-3 code.  Trace files, or?...  If iNSpect links directly
with the NS-3 library, keep in mind that that code needs to be GPL.

About using OpenGL, harware accelerated OpenGL can be a good thing, and can
make the iNSpect very performant.  On the other hand, one of the drawbacks
of OpenGL is that when it is not hardware accelerated it can also be
extremely slow.

ns-3-viz, on the other hand, uses goocanvas, which is based on the portably
Cairo 2D graphics library.  Cairo can also be hardware accelerated.  On
linux, besides a simple software renderer, it can use X RENDER extension, or
OpenGL (glitz).  On Windows Cairo uses GDI+.  Goocanvas is available in
recent linux distributions, and can be compiled for other systems
(goocanvas' configure script checks for win32, so I assume it can also be
compiled for win32).

One big advantage of Cairo over OpenGL is the ability of directly exporting
graphics to a vector format, such as PDF, EPS, or SVG.  For researchers
trying to produce high quality figures for papers this could be a real time
saver...  On the downside it doesn't support 3D, but I don't think in this
case 3D is important, and in any case 3D scenes can be transformed into 2D
graphics (projection).

To conclude, I would say that iNSpect and ns-3-viz only partially overlap.
Different feature sets, different levels of integration with NS-3.  I still
think NS-3 would benefit from inclusion of ns-3-viz.

Thanks,

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert