On Thursday, March 27, 2003, at 05:13 PM, Jacques Caron wrote:
Hi,
Some quick comments in this discussion...
I believe there are two problems here. One is the scalability of the "viral mesh network", and the other is whether this network is an extension of the Internet, or an alternative.
The first problem is not easy. I don't know in detail how the mesh network boxes handle routing (haven't looked into AODV very much, just a quick glance), but I don't believe they suddenly found a "magical" solution to this problem (the AODV approach is interesting, but it certainly has limits -- how many nodes do you think your "viral" mesh network will have in one year? in two years? in five years? in ten years?).
unbounded ... What I'd like to have is to allow an unlimited number of meshes, that are a part of the internet (routing internet traffic ... yes I realize that's radical) so they would have to scale with the internet.
The only way I know to handle it is to split the problem into smaller ones. That's what was done in the Internet with the IGP/EGP split (with OSPF, IS-IS, EIGRP or even RIP in the first category, and BGP in the second), and even within IGPs (with OSPF areas, for instance), and other related tricks (BGP confederations...).
Yeah ... I can see that there are varying levels of administration to be considered. My idea of the "viral" internet is that there is no administration necessary at all... connecting a properly configured device to the mesh magically makes it a part of the internet in terms of having it's own IP (v6 ...) and being able to route internet traffic as well.
If that seems too far-fetched (which I'm not convinced of yet) then another level would be to enable CWNs to set up their mesh to have new nodes magically expand the mesh ... but not necessarily to route internet traffic, without more configuration and perhaps explicit involvement from the CWN.
So the "viral" mesh network will need some coordination if it intends to stay alive. The coordination might be automated somewhat, but it needs to be there. The WIANA system just proves it: you need to assign unique IPs to start with, and this is definitely a "central controller" in a much more direct way than anything on the Internet.
Definitely, what I know about WIANA, I don't like. The addresses it distributes AFAIK are all in the site-local address space so, it's a relatively small address space, and it means that there's NAT at the edges of the locustworld mesh.
Once you have solved that problem (say, with delegation of big CIDR blocks to WRIRs which will delegate to WLIRs which will assign IPs to nodes ;->), you might also add some way of assigning AS numbers from whatever space you might have picked, and have some automated aggregation and AS formation protocol to make all that happen transparently. Might be fun...
Gack... acronym soup. I know what an AS number is but I'll have to read up on CIDR, WRIR, WLIR ......
The second problem is whether this network is an alternative to the Internet, or an extension. If it is an alternative, do what you want: re-use the same IP address space, AS numbering space, and so on. But I doubt many people will be interested nowadays. You might even use the Internet to "bridge" parts of your network together, thanks to whatever VPN technology you may want to use.
If it is an extension, then it has to abide by its rules. And that currently means using properly assigned IP address space and ASes, and integrating in the current routing scheme, which means that either you are connected to one ISP with IP(s) assigned by that provider, or you use BGP to become multi-homed. This was not done to annoy the small and poor. This was done because that's the only way people found to enable the Internet to grow to its current size without breaking (and some people would contend that the target was barely met).
Since anyway the first problem implied there is some form of centralized control of the mesh network, this centralized control might get IP addresses and AS number(s) and BGP-enabled upstreams and all that is necessary to make it happen. And one can even imagine that nodes that are connected to the Internet otherwise use the VPN tunnels described above to get to one of the BGP-enabled central sites so that everything works.
My feeling is like you that it's better to be an extension of the internet rather than "I'll build my own internet". Part of the reason I think that's a good idea, is that I hope that there's a cross-pollination of ideas. Looking at where the IETF is at (with MANET and IPv6 in particular) and where the CWN movement is at (with reducing or eliminating the need for traditional ISPs, telcos, etc.), there is useful knowledge to be shared by both groups that will probably strengthen both.
BTW, that's my current setup here: I have public IP addresses on my network at home even though I'm connected via a cable provider that only assigns a single IP address. The magic of IPsec :-)
Do tell? :-)
So, the IETF can probably do a lot of things to help, but the IETF is actually all of the people who want to do it. If you have ideas of things to improve in protocols, new protocols to create, or whatnot, look through the dozens of working groups and hundreds of I-Ds out there, contribute to whichever might be working on it already, or submit an I-D or propose the creation of a working group. But the IETF works on protocols, not on concepts and revolutions :-)
OK. I posted a message to the MANET (mobile ad-hoc networks) WG list and got a response already.
simon
Just my 2 euro cents...
Jacques.
-- Jacques Caron, IP Sector Technologies Join the discussion on public WLAN open global roaming: http://lists.ipsector.com/listinfo/openroaming
-- www.simonwoodside.com -- 99% Devil, 1% Angel
-- general wireless list, a bawug thing <http://www.bawug.org/> [un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless
