[tt] NYT: Laptop With a Mission Widens Its Audience

[J. Andrew Rogers]

On Oct 9, 2007, at 12:13 PM, Eugen Leitl wrote:
> On Tue, Oct 09, 2007 at 02:50:21PM -0400, Premise Checker wrote:
>> Why do you think there still isn't a global mesh protocol, Eugen?
>
> It's not easy to do, and there's still no real demand for it.
> It might be a psychological issue -- ArpaNet set out with a
> very different fitness profile from a locally dense geomesh
> of wireless nodes, operating at relativistic speeds. What
> people currently do is hacking IPv4 on embedded nodes, which
> is way trapped on a local minimum.


IMHO, we will not see a global mesh-based protocol until there is an  
important application that makes it grossly suboptimal to design a  
protocol any other way; we have the technology, but no compelling  
need.  Economics will do the rest of the work.  From where I am  
sitting, such an application will need to have three properties:

1. Data sources that are naturally distributed across the mesh
2. Data sources where the physical and logical topology are largely  
congruent
3. Data sources that generate far too much data to backhaul or  
centralize economically

We have not had a ubiquitous data space that has all three of these  
properties.  The closest approximation might be P2P file sharing  
protocols, but that does not meet condition #2.  The nascent massive  
market that *will* quickly meet all three of these conditions is  
geospatial services and data sources, particularly once they are  
based on spatial data rather than reductions to raster.  This will be  
built on top of IPv4/6 networks but backbone networks have crudely  
mesh-like behaviors already, moderated by economics.


J. Andrew Rogers