On Thu, Jun 29, 2006 at 06:26:39PM +0530, Suresh Ramasubramanian wrote: > yeah. so build a whole new internet? you need to talk to jim fleming, i
Actually, I'm interested in protocols which would scale across GLYrs and micrometers alike. > dare say. > > geographical / country level allocation of IPs is a complete nonstarter, Geography has nothing to do with virtual lines that primates draw on maps. Geography has to do with shapes matter aggregates under own gravity and orbits, and visibility (vacuum is a lot more transparent than silica, and an excellent FIFO for very high data rates (~3 cm/bit for 10 GBit/s). I used to call it geodetic routing, I should probably have stuck with the term. Geography is too confusing. A nearest approximation in today's technology would be a dense WiFi mesh (routed as the crow flies), or a global satellite constellation that routes via line of sight laser (limited by visibility (planetary shadow and mutual acclusion, and distance). It is a reasonably interesting problem, how to encode the position of an orbiting body with a bit pattern intelligently. Are you aware of a Gray code that only has local bit flips, and no singularities (such a wrap around with binary counters, where a lot of bits flip)? Things are also nasty at the poles, so probably a different orbit arrangement in the node cloud is needed. > trust me .. though country level management of IP addresses through the > LIR (local internet registry) concept is - unfortunately - becoming > popular .. Satellites don't know any countries, so the only way they would know about ground monkeys is via key distribution. No key, no access. > It started off with Japan / Korea / Taiwan etc getting themselves a > "local distributorship" of IP space in their countries .. and this meant > some fairly useful things - people could get IPs by paying in local > currency, speaking with locals, in their own language etc. > > Then Korea, for example, had the "brilliant" idea of mandating that > people in korea could ONLY get IPs from the korean LIR, krnic .. not > directly from APNIC. So - effectively, country level allocation. But > still, at least nominally, under the APNIC / RIR framework and following > the same guidelines and policies in allocating IPs. A good protocol would refuse to route addresses which are "wrong". A good protocol has to make no distinction between routing and cut-through switching at relativistic speeds. You can't get the fabric of this universe to route information directly, but a very thin virtual layer can. > Sure there are issues with the current model but those are more in the > local implementation (and most "shortages" of IP space are fairly > artificial, mostly caused by mismanagement and poor allocation > practices). That old chestnut about MIT having more addresses than > china belongs on snopes.com by now. I never meant China, though https://www.ripe.net/info/internet-management/faq.html addresses that (used to be true, got changed). I don't know how much IPv4 address space Africa has, but it's immaterial, given that most of the rollout with be wireless/mobile, and IPv6 based. > The sort of decentralized model you're talking about and the sort of > "decentralized internet" you're describing is science fiction right now I've been trying to talk with people about this since 1996, or so, but it's too lunatic fringe. I don't expect a shift to this until nodes in space significantly outnumber nodes on planetary surfaces. Of course, given self-replicating automation, that can happen really quickly... > - but even there, what you describe re the actual management and > allotment of IPs is fraught with trouble (even more trouble than that > other kooky idea about alternate roots .. possibly less trouble than Jim DNS should have been P2P to start with. Of course, building a distributed abuse-proof cryptographic file store is rather hard, and it will take probably another 20 years at least, until one has been built. Assuming, the content mafia and TLAs won't put an end to it. > Fleming's "ipv8" if he ever manages to realize it) Hm, never heard about that. Have to check it out -- it's a really dumb day at work. > You wouldnt see it from the IETF probably - but you'd get laughed out of > nanog (say) were you to suggest it. I need to write this up someday, but given that I haven't figured out the details yet, it would be rather useless. (Last time I wrote it up it was 1996 or hereabouts, so it's really kooky: http://leitl.org/oldcontent/ui22204/.html/txt/8uliw.txt -- skip straight to the "Router" part). -- Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org ______________________________________________________________ ICBM: 48.07100, 11.36820 http://www.ativel.com 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
signature.asc
Description: Digital signature