Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[liberationtech] What's up with mesh?

Sascha Meinrath meinrath at newamerica.net
Thu Jun 23 09:00:23 PDT 2011


Hi all,

I'd originally planned to stay out of the discussion, but given a lot of the
questions that were posed to the list, thought I'd jump in.  Below are some
quick answers and links to primary sources on various facets of the discussion.

Generally speaking, mesh wireless has been both widely successful and little
understood.  Between 2000-2003, my development team worked with MIT Roofnet on
first-generation open source mesh Wi-Fi.  Roofnet was a prototype network, as
was the work we did as part of the cuwireless initiative (which became the CUWiN
Foundation).  Neither network offered service level guarantees since both
technologies were highly experimental -- thus, while Steve Weis's experience was
quite correct, it is based on technologies from a decade ago.

As for OLPC mesh, it was doomed from the start.   I still remember when we first
got a shipment of OLPC boards -- CUWiN was part of the original mesh development
team -- and realized that they'd used a Marvel chipset, which had no open source
driver.  When we requested the necessary reference docs, we were told that they
were proprietary information -- so open source developers couldn't develop for
OLPC.  Soon afterward, other developers ran into the same problem -- a fairly
good write-up of the problem is available at:

http://www.cmosnetworks.com/OLPC-MarvellIssue-MyWriteUpOfTheSituation.html

As for Shervin Pishevar's OpenMesh initiative -- I do hope it works out, but
haven't yet seen any meaningful information about the technologies they're
implementing.  While Shervin has gotten a good amount of press, I do worry since
I haven't found any technical specifications, repositories, or an active
developer community behind the initiative.

Matt Van Hoven's link to the Detroit Digital Justice Coalition's mesh initiative
(http://detroitdjc.org/wireless-mesh) is actually one of the collaborations that
OTI has helped coordinate and implement.  We'll be expanding the network this
summer and during the Allied Media Conference (happening this week --
http://alliedmedia.org), so for anyone who wants to see these technologies for
themselves, we'll be running some hands-on workshops this week.

Shaddi Hasan rightfully points out that "management overhead is much higher than
most expect" -- one of the key deliverables for the State Department supported
work we're undertaking is to improve auto-configuration on these systems.  One
of the key problems isn't whether the technologies work (they do), but that
they're not very accessible to non-techies.  Our broad goal is to get as close
to zero-conf (zero configuration) as possible.  That said, community wireless
networks are _extremely_ large-scale -- from thousands of nodes covering Athens,
Greece (http://awmn.net); to multi-layered mesh in Vienna, Austria
(https://map.funkfeuer.at/wien); to hybrid mesh/hub-and-spoke regional networks
covering the Djursland region of Denmark (http://djurslands.net); to the 13,000+
node network of networks throughout the Catalonia region of Spain
(http://guifi.net).  The folks running all of these networks are good friends,
so if folks have questions for them, I'm happy to make intros.  As I wrote two
years ago, open source mesh has been doing 80+ mbps over multi-KM links for
quite some time
(http://www.saschameinrath.com/2009/june/12/open_source_802_11n_big_breakthroughs_are_coming),
thus the throughput problem isn't usually the mesh itself, it's the Internet
uplink.

Griffin Boyce points out the importance of Intranet communications -- which is
exactly right!  The mesh networks we built in Urbana, IL did exactly this and
we've been calling for this type of technology for years now (see, for example,
http://newamerica.net/publications/policy/rise_intranet_era).  The NYT didn't
really cover the technologies involved in our work, but the ad-hoc mesh wireless
we've been building is, in fact, an Intranet -- thus, Internet connectivity,
while useful, isn't needed for network participants to communicate with one-another.

Charles Wyble has also pointed out that Atheros is currently by far the
front-runner for open source mesh wireless.  Their recent sale to Qualcomm has
left a lot of us quite concerned for the future of their relative openness,
however.  For those looking closely at the NYT "Internet-in-a-suitcase" picture,
you'll see several piece of Ubiquiti gear -- they are, in fact, pretty amazing
gear for the price point.  Once the FreedomBox Foundation gets their tech
functioning, that will be another really useful resource within a community
Intranet as well.  Meanwhile, other key groups that we've been working with
around the globe include the Serval Project (http://www.servalproject.org), Gnu
Radio (http://gnuradio.org) and the OpenBTS initiative
(http://openbts.sourceforge.net), FunkFeuer (http://funkfeuer.at) and the OLSR
crew (http://www.olsr.org), of course -- the Tor Project
(http://www.torproject.org), etc. -- all of whom have folks who are working with
us on Commotion (http://tech.chambana.net/projects/commotion).

Happy to answer any follow-up questions folks have,

--Sascha Meinrath
Director, Open Technology Initiative
New America Foundation




More information about the liberationtech mailing list