[Click] Using nsclick

[Jens Mueller]

Hi Mike,

> I haven't checked it out in detail, but I'm guessing your problem has to 
> do with the way the ns-2 side of nsclick handles address resolution and 
> packet encapsulation. The ns-2 side of things is expecting regular 
> wireline ethernet encapsulation on the packets it gets from click. Under 
> the hood, it has to map the ethernet addresses you're using inside of 
> click into the internal node addressing scheme that ns-2 uses so that 
> ns-2 can apply its MAC and PHY layer models and deliver packets to the 
> correct places.

I wrote a ns2 script to confirm this, which only uses ethernet frames. I
have two nodes one creating ping requests and the other replies. It
works as you suggested, but what confuses me, is that I don't need to
add a link between the nodes. And I don't see any packet transmissions in
nam, but packets are exchanged as I see them in the dump files created
by click.

> In order to do this, it has to know the encapsulation type so it can
> figure out where the approprate addresses are. If you use 802.11
> packet frames instead of regular ethernet frames it will likely get
> confused. If this is indeed the problem, a quick hack to get you up
> and running would be to add a layer of wireline ethernet encapsulation
> to the 802.11 packets your sending out. The ethernet type shouldn't
> matter, and copying the current hop/next hop 802.11 address fields
> into the source and destination fields of the ethernet packet should
> be OK. 

We wrote a click element, which does this job and now it works. But as a
drawback broadcasts are not shown by nam. Any ideas?

> Fixing it inside of ns-2 would also be an option, but only if
> you want to get involved in ns-2 internals.

You mean to fix it in the nsclick patch, right?
We also like to name wireless interfaces beginning with 'ath' and wired
interfaces with 'eth', but we don't find the right places.

> Another reasonable source of information about nsclick would be the
> archives of the old e-mail list (now defunct as far as I know) -- I'm
> not sure if that archive is up on the Colorado website or not. This
> page:
> 
> http://systems.cs.colorado.edu/index.php?id=13
> 
> also appears to have a little info on it. I think that nsclick is
> between official maintainers at the moment, though it's possible that
> someone at Colorado will pick it up at some point.

Yes, the old mail archive
(http://www.cs.colorado.edu/mailman/listinfo/nsclick-users) returns with
an page not found error.

Bye,
Jens