If IP addresses were ~free, wouldn't it make sense for each program to just have it's own ?
@cjd you mean like, using the last ~2 octets of IPv6 address like we currently use TCP/UDP port number? With well-known suffixes for well-known protocols, and random high suffices for outgoing traffic, etc?
If cjdns then you don't get to pick the address you will get, but not a big deal, when you call bind you use whatever port number identifies the protocol and you get an address unique to your app..
@cjd smh obviously cjdns has some limitations. I was thinking what we'd want in an ideal world.
What routing-protocol-agnostic addressing scheme would be desirable.
In an ideal world, what kind of central authority would exist to manage the issuance of ip addresses?
Ok sounds good, you keep praying and I'll keep building things that can potentially work, let me know how that goes..
@cjd the idea is to assume we have a perfect implementation of some things, and then see how we can architect or implement other things.
Obviously the ideal world implementation won't always translate 1:1 to the real world, but it may give us some hits as to what we may want to approximate, or asymtpotically approach, in the long term, without getting caught on short-term benefits.
We have RIRs and BGP now.
We may have cjdns and pkt tomorrow.
But they won't be good forever, and if we don't plan ahead we'll end up with x86-esque mess of legacy baggage.
Not to mention that cjdns reeks of "I'm the only one on the network, explicitly incompatible with any other routing protocol"
I'm thinking that in ip-is-the-hash-of-your-key land, this becomes possible. So when cjdns becomes the dominant paradigm, we might want to rethink ip addresses in general...
Server run by the main developers of the project It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!