Re: From Andy:

On Mon, 30 Sep 2002, Josh Howlett wrote:
[opinions on new distro at http://www.locustworld.com/]
> My opinion is that this is a brave effort, but I will happily bet any part 
> of my anatomy that this won't scale to 6000 nodes.  There are so many 
> things wrong with this approach it's difficult to know where to start, but 
> here goes:
> 
> 1) This ain't a mesh network.  Repeaters must associate directly with the 
> gateway node, and can't associate with each other.  Repeaters can't pass 
> data between themselves, and can't "chain" at all.  This limits the size of 
> the wireless "footprint", and increases the density of ISP connections that 
> you require.

Hmmm.  I think it would be required for the cells to allow at least one 
hop, preferably two or three.

> 2) There's no attempt (that I can see) to ensure that IP address space is 
> non-overlapping.  Hence, it's possible for two nodes to randomly select the 
> same /24 network.  The larger the network becomes, the more likely a 
> "collision" will occur (1 in 256 for the second node, with the probability 
> doubling whenever a new node associates).  Guess what this will do to the 
> routing.  Peer to peer communication will be flaky and will get 
> progressively worse as the network grows.

Ouch!  That could have dire consequences.  Surely it can't be that tricky 
to work out what subnets are taken, and go from there.  Even a database on 
the gateway would be a start.

> 3) You can't even roam between *repeater cells* without being allocated a 
> completely different IP address.  Which may or may not be conflicting with 
> someone else, d'oh.

That's a Bad Thing.

> 4) No authentication/encryption.  Yes it says encryption on the website, 
> but the keys on the tunnels are like this:
> 	YesAtTheMomentAllEncryptionKeysAreTheSame0
> 	YesAtTheMomentAllEncryptionKeysAreTheSame1
> 	YesAtTheMomentAllEncryptionKeysAreTheSame2
> 	etc...

Heh.  I expect it's in the pipeline.

> IMHO, this is really just a very complicated and broken system for NATing 
> an internet connection over wireless.

The author mentioned multiple gateways and redundancy.  What you have 
described seems to suggest that it supports neither.  Is this any 
improvement over the OpenAP distro at the same site.  If you're right, our 
best bet at the moment seems to be a hack of the OpenAP distro to run from 
compact flash and do routing via TCP/IP.  We are not dealing with 
thousands of nodes, so having to hard code routing on each one seperately 
is hardly an issue.

Andy


Bristol Consume Website: http://consume.andylaurence.co.uk/
Bristol Consume Discussion: http://consume.cyberprog.net/
Administrivia: To Subscribe send an e-mail to [email protected] with the subject subscribe. To unsubscribe send with the subject unsubscribe.
To post to the list just e-mail [email protected] or hit reply in your mail client.