[MLB-WIRELESS] Re: [fn-tech] Wireless network structure v1.3-PRE.1

[Simon J Mudd]

cliff at steam.com (Cliff Skolnick) writes:

> To be honest I think you are looking at wireless as an extension of wired
> networks, but hey we know wired networks.  Your stuff is a collection of
> standard best practises for wired networks, not really anything new.
> Revolutionary, and even evolutionary, wireless networks will have a whole
> new set of problems and need new solutions.

OK, fair comment, but up until now no-one has come up with _any_ real
design, or if they have I haven't seen it. I don't claim to be an
expert, but I do think that it's fine thinking locally of one or two
"AP"s connected together and it's quite a different thing to think
about how to reallly build a large WAN/MAN.

I'll accept criticisms and try and change the document if people think
I'm going the wrong way, but until now I've not seen anything better.

> I think it is a shame to assume that every node can reach some central
> server for IP address allocation.  This can be because you are really trying
> to link in a wired network of 5 systems that will have to go through a
> wireless router to get on the network.  Is everyone that wants to hook up 2
> systems going to run software to become a full fledge internal router in
> your setup?

I envisage the node's router in this way, yes. I see problems with
this, but haven't come up with a better soluction.

>  Or will all of them need to directly connect to the wireless
> network's internal router?  Or are you going to hack stuff with dhcp relays
> and proxy arp to glue stuff together in your network?

No "proxy services" I'm sure wouldn't work, or would need special software.

> Or will this box be a simple bridge in this case?  How about an
> example connecting an etherneted network of machines to the wireless
> network?

If you bridge different nodes then there'll probably be a much larger
amount of traffic going over the network especially with ethernet
broadcasts and stuff like that. If the "bridge link" between two nodes
doesn't interfere with the "client access" this may not be a problem,
but on a high traffic network this will probably break down(???)

> My vision is a community network user should be able to plug a box into a
> network on machines, a whole LAN, and have the thing take care of all
> machines.