[Ns-developers] Aodv source code and packet header structure

[Qihe Wang]

Hi Martin,

Your suggestion in previous post about changing packet structure is
something related to architecture changes, which would break the current
code base.  That kind of change would move ns-2 to ns-3.

An aodv error packet is a packet on its own.  It doesn't carry any user
data.  Error entries are its payload.  How does this violate the
semantics?  You may have a look at DSDV code, which have everything in
data_ member.

I think you can implement multiple versions and add a 'ver_' class member
which is settable at runtime, by tcl script.

Regards,
Qihe

On Thu, 21 Apr 2005, Martin Wenig wrote:

> Hi,
>
> i agree on the movement of the #defines i always disliked them at the
> begining of the aodv.h file. (Actually i had very similar problems when
> i tried to create a new AODV-Variant as you had when inheriting from it).
> Your second comment aims in the same direction as my suggestion to
> change the packet header structure. I also don't think that it is
> usefull to always have all headers in maximum size in every packet...
> But as a workaround for aodv, it would be possible to replace the
> statically allocated array with a dynamicaly created. This would then
> put the burden of correctly deleting it on the aodv-programmer. This may
> be especially complicated when an error-packet gets dropped.
> I wouldn't use the data_-member because the error-entries belong to the
> aodv-header. Of course there is no coding challenge in that but i think
> it would somehow break the semantics (again).
>
> BTW: Would you be in favor of a RFC 3561 compliant implementation or
> should the newest draft be implemented?
>
> bye,
> Martin
>
> Qihe Wang wrote:
> >Martin,
> >
> >This comment is related to AODV only.  I'd like to see two improvements
> >over the current AODV code structure.  One is moving all '#define's to
> >class members so that they are changable by inheritance (I'm adapting it
> >to Blueooth scatternet where the delay is much larger).  The other is
> >making it more memory friendly.  That is, instead of putting all error
> >entries in the header, move it to data_ if the number of error entries is
> >large.
> >
> >Regards, Qihe
> >
> >
>