[LINK] Indonesia Overtaking Australia with Wireless Internet

[Richard Chirgwin]

On 8/06/11 8:40 AM, Tom Worthington wrote:
> Jan Whitaker wrote:
>
>> I'm curious. Is there any work going on to increase the amount of
>> throughput and multiplexing within a frequency allocation? We keep
>> talking about the way fibre expansion can happen with better end
>> equipment replacement. Are there similar advancements in the R&D
>> centres for wifi as well? ...
> Yes, there are people researching how to increase wireless capacity, at
> ANU, NICTA, CSIRO and other research organisations in Australia. This
> research is mostly into how to use the greater signal processing
> available, multiple antennas and more, shorter range, base stations:
> Here are some recent papers:
> <http://scholar.google.com.au/scholar?as_q=wireless&btnG=Search+Scholar&as_epq=australian+national+university&as_occt=any&as_ylo=2010&as_sdt=1&as_subj=eng&as_sdts=5&hl=en&as_vis=1>.
>
> Fibre optic cable will still have fundamental advantages over
> wireless: the signal is confined to the cable making interference less
> of a problem and light has a higher frequency than radio frequencies
> used for wireless, so it can transmit much more information.
>
> I get invited to the ANU/NICTA/CSIRO wireless seminars:
> <http://www.google.com.au/search?q=site%3Acecs.anu.edu.au%2Fseminars+wireless>.
> But I do not understand much of what is discussed. One area which has
> not been explored much is mesh networks, where the handsets themselves
> relay the signals.
Mesh networks have been explored, and deployed. What we've learned in 
practice:

1. Latency is hard to manage. In any unmanaged mesh, you can't predict 
the number of hops between pairs of devices - or between one device and 
the Internet. Not only may latency be very long (multiply the air 
interface latency by the number of devices), it changes over time 
(jitter). For an app like voice, you need predictable and low latency. 
So mesh networks are limited in their applications.

2. In a mesh, you either share spectrum (reducing the capacity available 
to individual devices), or need sophisticated (therefore expensive and 
probably proprietary) radio management.

3. A mesh network is also subject to whether anyone else has a right to 
use the spectrum involved. Would anyone welcome their home WiFi not 
working because of overcrowding by a citywide mesh of handsets?

Look at the problem from the point of view of handsets relaying signals.

- the radio used for relaying signals has to be separate from the one 
communicating with the base station. OK: use 3G for on-network traffic 
and WiFi for mesh traffic; see (2) above.
- handsets need the processing power and memory to act as routers rather 
than as endpoints.
- user traffic needs to be encrypted so as not to be published to 
everyone on the mesh, adding more overhead to the handset.

This is rather a lot of work to create a network which will strip out 
the "advanced" features of an iPhone or Android handset (no video or voice).

I can imagine good applications for mesh networks - for example, 
low-value, undemanding machine-to-machine applications. An electricity 
smart meter in densely-populated areas is a very good example: each 
meter is going to be within reach of the next one, there's no reason not 
to encrypt the data, and its bandwidth and latency demands are quite 
modest.

But mesh has been idealised (for example, the recent Mark Pesce book 
gives mesh networking the full science-fiction treatment with no 
consideration of the technology) beyond its capabilities.
> What I find more interesting are human factors ways to increase the
> responsiveness of a hand held device, by providing a simplified
> interface and summarised documents suited to a smaller screen and a more
> distracted user. A good example is SMS, where the lack of length,
> formatting and multimedia are see as features, not limitations. Images
> and videos can be sent at lower resolution for mobile users and simpler
> versions of web pages and documents can be  provided.
"Can be", but regrettably, are not. The aim of the content push is 
exactly the opposite: embed the content into the apps into the device, 
and make them richer and more complex. This seems to be popular among 
users, but it works against doing things to minimise the waste of spectrum!

> Also the location
> and current activity of the person can be derived automatically and used
> to filter information offered.
Which I argue is a good reason *not* to use such services. But that's a 
view confined to an apparent minority of people who (a) don't like being 
advertised to, and (b) don't like the privacy invasions implicit in such 
services. Once that privacy invasion extends beyond the owner of the 
device - as in Google / Apple / Microsoft recording any WiFi hotspots 
their devices see, regardless of ownership - it's clearly become 
unreasonable, intrusive and creepy.

RC
> Metadata can be used as a form of extreme data compression for mobile
> users: rather than offering a large document which will take a long time
> to download and the person will probably not want to read anyway, a
> summary can be provided instead.
>
>