[ILUG] Open wireless networks

[Brendan Minish]

> On Thu, 2011-10-20 at 09:45 +0100, Paul Scollon wrote:
I had approached the local council and commerical manager in Dundalk 
> about doing an open mesh network as part of Dundalk Town's 
> redevelopment, and got the green light for it. However, half way through 
> the planning stage they decided to buy a solution and pay a lot of money 
> for vendor support and a contractor to install, and tie authentication 
> into the local library (no membership, no access).


Paul,

I was involved some years back with a large outdoor mesh networking
project. 

I was also heavily involved in several community networks based on a
Co-Op model.  

Mesh a crap technology for internet access, frequently advocated by
network folks who fail to fully consider the limitations of the
802.11{a,b,g,n} technologies that it's generally built upon.
Issues (in no particular order and I may have left some out)

Hidden node. This is where node A talks to node B but node B can
overhear (& interfere with)  Node C talking to node D 
This problem is then made Much much worse because all the users are
potential hidden nodes too, but your Users collisions are also killing
the throughput between nodes. More collisions lead to more retries,
which can also collide. 

Multipath, nodes and users often won't have good radio paths between
them, there's more to achieving optimum throughput on a link than just a
good signal to noise ratio, if there is a lot of multipath the
throughput will be very poor but how does you mesh avoid using these
paths? 
How can the network 'tell' the end users device to connect to an
alternative node ?

No air interface QoS (a limitation of 802.11), QoS is important for
sharing a scarce resource (I.e end to end network bandwidth ) in a way
that allows latency/jitter sensitive applications to work well.

NAT, multiple layers of NAT.. The systems I have seen are designed to
deal with multiple uplinks, these are generally via a mix of providers,
each with a different public IP. 

Logging and network monitoring, this is perhaps solve-able but how do
you keep track of who has accessed the network and what did they do
whilst on the network. There are legal requirements that should be met. 


yes you can build a mesh that covers a large area well, what you can't
do is also scale that to reliably supporting a significant number of
users trying to access the internet in a bandwidth consuming manner.

Mesh networking DOES work for certain applications, E.g sensor networks,
it can also support low bandwidth applications but it's not a robust and
viable alternative to designing a network built from backbone Point to
point links that uplink access nodes.   

.brendan