wifi for 600, alex

Alexander Harrowell a.harrowell at gmail.com
Fri Feb 16 16:53:41 UTC 2007


On 2/16/07, JAKO Andras <jako.andras at eik.bme.hu> wrote:

Please don't forget that 802.11 uses the CSMA/CA protocol. All nodes,
> including the AP and _all_ the clients should hear each others'
> transmissions so that they can decide when to transmit (when the medium is
> idle).


Yes. But so long as they can all interfere with each other, you're still
going to pay a cost in informational overhead to sort it out at a higher
protocol layer, and you're still going to have the "electronic warfare in a
phone box" problem at places like NANOG meetings. 3GSM is the same - even
the presence of ~10,000 RF engineers doesn't prevent the dozens of
contending networks..

Essentially, this is a problem that perhaps shouldn't be fixed. Having an
open-slather RF design and sorting it out in meta means that WLAN is quick,
cheap, and hackable. Trust me, you don't want to think about radio spectrum
licensing. On the other hand, that particular "sufficiently advanced
technology is indistinguishable from magic" quality about it causes
problems.

Intentionally limiting the clients' TX powers to the minimum needed to
> communicate with the AP makes RTS/CTS almost obligatory, which may be
> considered a bad thing. Once again, in the ideal situation all nodes hear
> each other, at least from the CSMA/CA's point of view.
>
> Regards,
> Andras


I'm not sure that's ideal in my point of view, in so far as we're talking
about a point-to-multipoint network rather than a mesh. And why would anyone
ever want to use more power/create more entropy than necessary?

This argument sailed around in the early days of WiMAX, when people were
talking about running it in unlicensed 5.8GHz spectrum and "finally getting
away from the telcos and the government", until they realised that it's not
"big wi-fi" and isn't designed to cope with contending networks..

Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20070216/6c6a41ed/attachment.html>


More information about the NANOG mailing list