Re: [ns] Packet Header

[Lloyd Wood <l.wood@eim.surrey.ac.uk>]

On Fri, 20 Apr 2001, sandry@iol.it wrote:

> I have read the previous discussions on how ns usually doesn't simulate
> header overhead. My impression is that I have to enable the use of all
> headers of my interest

no. by default, everything is enabled, although you can now choose to
disable headers to save memory/increase simulation speed. (This does
not affect packet size.)

> and I have to give my agent the capability of
> fill the fields of the header that it have to use in each situation.
> But, if so, this is a great limit of simulator: in fact, I cannot
> simulate the gain of using compressed packets, because the simulator
> simulates that all packets have all fields of all headers I have
> enabled. In other words, I cannot see what happens when my agent sends
> compressed packets, because all packet has non only the fields of
> compressed headers, but also the fields of all other headers.
> I'm right?

alas, no.

There is no correlation in ns between having information in a struct
and resulting header size. That information is just there to be
associated with the packet, and the storage space it takes up in each
packet sent is precisely zero bytes.

(If we only stored the info each ns packet needed and a payload size,
we could determine the memory the packet info consumes, tweak for the
non-existent payload, and have our overhead, but this wouldn't be
'realistic' as far as e.g. IPv4 is concerned.)

How much realism do you want from a simulator?

Linux pppd implements VJ header compression; easier to modify that to
do the overly complex ROHC stuff on a serial link between two linux
boxes, I'd have thought.

L.

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>