[Freedombox-discuss] Serval Mesh Extenders

Paul Gardner-Stephen paul at servalproject.org
Sun Jul 14 13:14:42 UTC 2013

Hi John,

On Fri, Jul 12, 2013 at 9:43 AM, John Gilmore <gnu at toad.com> wrote:

> > The idea is that it uses the UHF packet radio to mesh over greater
> > distances than is possible with Wi-Fi, the trade-off being lower
> bandwidth.
> >
> > In general, we find that the UHF packet radio has a range of about 10x
> that
> > of Wi-Fi when deployed indoors with omni-directional antennae.  This
> means
> > it has a range of about a block in a suburban or urban setting compared
> > with Wi-Fi's range of about one house or apartment.
> > ...
> > Extending the range in this way is a critical enabler for the adoption of
> > mesh communications because it removes the need for skilled installation
> > and lowers the required penetration rate from near 100% in a local area
> if
> > using un-aimed Wi-Fi to below 1%:
> My guess is that you are getting better range in the 915 MHz frequency
> block because fewer people are using that block.  Onmidirectional WiFi
> used to reach a whole block or more in San Francisco, before everyone
> had it (and aiming a dish at an omni would work all the way across the
> city).

There are a few factors involved.  One is, as you mentioned, the relatively
vacancy of the spectrum.
But we do also use lower bit rates (128kbit, giving +8dB versus even
802.11b/g @ 1mbit), and the propagation characteristics of 900MHz versus
2.4GHz give another +8dB to +9dB over ranges of a few kilometres. We also
use +24dBm TX power giving +4dB over typical wireless.  So all up, we are
about +20dB better than Wi-Fi.

> I'm speaking as someone who ran a small wireless ISP using 1-megabit
> WaveLan gear in the 1990s, and who ran a public WiFi access point from
> my roof in the early 2000s, fed as part of SFLan from across the city
> using a second WiFi dish.  And who brought Internet to Burning Man and
> fed it around the playa with WiFi on towers -- which worked over
> *miles* until everybody started bringing wifi phones and wifi laptops.
> And who tried using 802.11s-predecessor mesh on the One Laptop Per Child,
> which stopped working once you put 30 of them into a classroom.
> If you deploy a lot of these, they will stop giving you the advantage
> that you designed them for.  I would find it hard to recommend this
> strategy for widespread deployment.

Yes, I agree that increasing deployment density increases a variety of
problems, including the ones you have mentioned here.  Nonetheless, having
two different bands (900MHz and 2.4GHz) helps push that boundary a little
further.  But at the end of the day, we are happy to provide a solution for
small communities where it wasn't previously possible.

>         John
> PS: You don't need to go into the 900 MHz band to get longer range and
> lower bandwidth.  You can also get better range on WiFi merely by lowering
> its bandwidth, using the original 802.11'b' or 'g' protocols.

Quite right -- we are doing that already, and indeed going even lower than
802.11b/g allows as mentioned above.

> PPS:  "Mesh doesn't scale over radio." is the 5-word summary of my
> experience.  Merely making mesh work over wired connections is still
> a hard research problem that nobody has great solutions for.

Agreed. This is also why for all services except live voice we use
store-and-forward based protocols instead of trying to route packets,
because as you correctly point out, radio-based mesh packet switched
networks don't scale.

My overall position is that what we are doing most certainly has
limitations and constraints, but we will struggle and strive to push
against those limits as hard as we can to do the most we can with what we
have.  Perhaps in the process we will create something that helps people in
a variety of situations. Perhaps we will just make something that will be
an object lesson of failure in future. Either way we will have pushed the
boundaries and contributed something.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/freedombox-discuss/attachments/20130714/9706d6fa/attachment.html>

More information about the Freedombox-discuss mailing list