[MLB-WIRELESS] Re: Node x is over this way -was- Applications

Ben Anderson a_neb at optushome.com.au
Wed Mar 20 05:18:26 EST 2002


> If i follow this, the GPS position is just a convenient way to
> establish the location of the node *in the network*, and at times
> it's even a bit counterproductive, in the cases where geography and
> network topology don't actually concur.  eg having to bounce south to
> go north.

Umm, I was proposing gps co-ordinates as the actual routing metric use to
migrate packets towards the destination node, with 'shortcuts' through the
network defined, and propegated such that each node knows how to go
backwards to go forwards.
Limiting the amount of broadcast traffic is very important.

> If it's actually the position in network space that matters, would it
> not be possible to encode the concept of "he's over there" in network
> topology terms?

The position in network space is a problem, because then you run into
gnutella like scaling issues, where the vast majority of the traffic is
routing traffic.  And it's even worse than the gnutella problem, because
we're talking about implenting it on broadcast-only mediums.

> Please excuse me if this is one of those questions that was answered
> in like 1670 by an arthritic frenchman with a slide rule.

It isn't too far dissimilar to the travelling salesman problem ;)


> hmmm.
>
> what if each node was assigned a unique prime number, and it is
> multiplied by  the primes of each node it can see...  and then those
> numbers are in turn multiplied and passed on and  ummm....  then
> would it be possible that each node around me has a large but single
> number that describes the network topology around it?  then if brian
> has previously sent me his prime, i just have to divide the number
> stored at x nodes around me by his prime.  send his packets to any
> node that has his prime as a factor, and they pass it on and so on....

It's possible, sure.  But the number gets **rediculously large** as you
scale to large numbers... Just think about a million different prime
numbers, all multiplied together...  and now propegate that around the
network every 10 seconds to deal with nodes roaming around.  Run out of
bandwidth, don't we ;)

> i realise that soon you're getting into large primes, but as its not
> necessarily a security thing, could we start with 1,2,3,7,11,13
> etc... rather than 128 bit primes?  is it essential that network
> _location_ be that secure, as opposed to content and _personal_
> identity?  besides, we don't have to factorise for unknown primes,
> just simple division

Even if the guy in the black mask doesn't know my real name, I don't want
him kneecapping me as I go to check the snail-mail.


> or can we exploit six-degrees here, ie "has anyone seen brian today?"
> (echo echo echo) give up.  if i can see x nodes, and i ask them, and
> none of them can see brian, and they ask y nodes, then kill the virus
> after six generations, then the node nearest me says "yeah i know
> which way he is", wothout actually passing his location back to me.
> then there's a kind of virtual channel back to him, where each hop is
> concealed from the hops either side..

This is very similar to the way mobilemesh do it..  broadcast existence
using a 'hello' type strategy to the whole network, and stop passing the
hello on after x hops.  Which means nodes beyond x don't know of that nodes
existance, and can't talk to that node.
That might be fine for mp3 trading, but it's pretty crappy for general
communications etc.


> I'm probably repeating stuff that's already been said, except in
> layman's terms, and/or wrong.  or i need a first year network theory
> book to read.  and i'm grasping for the shape of something i can
> almost see, which may or may not be possible even if i do grasp it.
> on the other hand if i'm getting somewhere please let me know.

Hopefully I haven't rained on your parade enough to discourage you...  Some
interesting ideas, the primes one actually had me thinking for a few minutes
:)  Keep 'em coming :)

> anyway that'll be 2 cents thanks - feedback welcomed

Feedback granted.

> clae.
>
> ps mojo =  bandwidth as currency?

Not quite...  it makes sense to reward people who run proxy servers with
mojo as well, as they can arguably add more bandwidth availability to the
network by caching stuff.
Mojo is just some magic number in a database somewhere (though I want it
distributed, with strong encryption to protect the little cookie like
objects).  The more useful infrastructure you provide to the network, the
more you're rewarded by right to use others infrastructure at higher
priority.
The mojo idea also has other benefits as it makes a lot of DoS issues just
go away.  Also helps to maintain QoS, depending on the metrics used to
govern mojo distribution.  A lot of simulation I suspect will be very useful
in discovering good metrics for distributing mojo.
Hopefully that last paragraph doesn't confuse you too much...  Point out
bits if you think I need to simplify, as I'll write a wiki page soon and
making it readable to 'the masses' would be beneficial.


--
To unsubscribe, send mail to minordomo at wireless.org.au with a subject of 'unsubscribe melbwireless'  
Archive at: http://www.wireless.org.au/cgi-bin/minorweb.pl?A=LIST&L=melbwireless
IRC at: au.austnet.org #melb-wireless



More information about the Melbwireless mailing list