Sie sind auf Seite 1von 6

Discrete IP

Pars Mutaf pars.mutaf@gmail.com Sept 10, 2012 Abstract I define Continuous IP as a policy dictating to all entities on the planet to use the same IP version. I define Discrete IP as a policy allowing each entity to choose their IP version and design new ones. I discuss their differences. I conclude that Discrete IP is synonym to perfectness in terms of freedom of choice, economically more viable because it better respects the end-to-end principles. Discrete IP is better because it allows Continuous IP. Different entities can use their own IP version, even agree on the same one for 1000 years until one designs and transitions to a better one. Others may follow or not. Everything is possible and the question of IP transition is solved forever. We have no idea what time will show, and we do not have to. I don't present all the technical details. What is important is the concept at this point1. But I show that it is easy to design. Continuous IP

Continuous IP is a policy dictating to all entities on the planet to use the same IP version. Currently it is IPv4. The next generation is IPv6. Tunnel brokers currently allow to use both, but the real goal is worldwide transition to IPv6. Tunnel broker approach suffers from tunneling cost (160 bit per packet). But this is because it is assumed that everybody will transition to IPv6 and we will not have this cost anymore.

1 I stopped publishing in the traditional publishing system which is interested in all details of the work before even publicly discussing the proposed concept. We felt like primary school children in this system. We need a platform to openly discuss ideas before implementing them. Fear-based science has a very low signal/noise ratio. Better systems are possible, see for example http://www.content-based-science.org/

Discrete IP I define Discrete IP as a policy allowing each entity to choose their own IP version. This is illustrated in Figure 1.

Figure 1: Continuous vs. Discrete IP


Instead of having 1 Internet, we have many big and small Internets with random versions of IP. Big and small Internets join others.

Discrete IP can be simply designed as follows2:

Figure 2: Simple model of operation.


All nodes in all Internets are registered to the DNS. The following example design is illustrated in Figure 2. IP payload copier, IPPC, copies the payload in IPv4 packets to IPv6 packets, and vice versa. It is different from IP translators which translate header information. I discuss later how TCP can work over this scheme. 1. Node A doesn't speak IPv6. 2. Node A makes a DNS request, obtains the destination address D and the IP address of the IPPC to reach the address D. In this example, the IPPC address is B. 3. Node A sends the first packet to the IPPC in a special packet also indicating to the IPPC the destination IPv6 address. 4. IPPC creates state for the addresses A and D. 5. IPPC copies the payload found in IPv4 packets to IPv6 packets, and vice versa, between A and D. The next design challenge is to traverse multiple Internets. This can be achieved for example using multiple tunnels for the first packet of the session.

2 Better designs may be possible. This one is simple.

Figure 3: Discrete IP deployment example. All nodes are registered to the DNS.
Figure 3 illustrates a Discrete IP deployment example. In this example, we have the current IPv4 Internet (Internet 1), a new IPv6 Internet (Internet 2), and we have many small IPv4 and IPv6 Internets. Obviously, all these Internets will use the same IP address space. The problem is to locate the correct node. Using the DNS, the nodes 45.124.1.56 in Internet 3 and Internet 6 can be located as follows: DNS name1 DNS name2 45.124.1.56 IP address of the IP copier to Internet3 45.124.1.56 IP address of the IP copier to Internet6

We do not need to use NATs anymore, we can add new clients and servers by adding new Internets. I can install an IPv6 or IPv4 Internet at home and connect it to the IPv4 Internet. It will be one of the millions of Internets on the planet (mine will be Internet345611, for example). My nodes at home can be servers because they are registered to the DNS and reachable to whole planet.

Comparison Discrete IP seems to better satisfy the end-to-end principles: We do not touch the core network, i.e. the current IPv4 infrastructure which works. Those who wish, add additional infrastructure if necessary (IPv4, IPv6, etc.) We modify the end-hosts only (this is violated by the Continuous IP policy while transitioning to IPv6, because it requires modification to existing routers) We do not have worry about others' IP version preferences, we do not have to implement IPv6 for example

Some ISPs may not want to invest in transitioning to IPv6, others may be impatient to do it. China may want IPv6, Uruguay may not. Russian hackers can design their own IP version, it is their wish. A Saudi oil billionaire can buy a million node IPv7 Internet for his son's birthday studying computer science, even if the rest of the world does not recognize this IP version. It is their right. I can install any version of the Internet in my company. It is my right. Discrete IP may require some modifications to TCP and DNS. TCP assumes that source and destination IP addresses have the same IP version. We may need another type of identifier. Such host identification protocols already exist. These modifications can easily be made, and this is much better than worrying about the whole core Internet, i.e. others' IP version preferences, forever. This is the end-toend principle. We do not worry about the core network, we change the end nodes. If China deploys IPv6 but others do not, others will reach Chinese servers using tunnel brokers. Discrete IP removes this tunneling cost for every packet exchanged with IPv6 networks, and it can allow any IP version. These advantages did not come by chance. They came because the assumption that all nodes would transition to IPv6 was not made. Continuous IP can be more efficient at the IP layer. Multicast, Mobile IPv6, IPsec are possible. Discrete IP has some shortcomings if these protocols are necessary and cannot be implemented differently. But even in this case, Discrete IP is better, because it is so flexible that it also allows Continuous IP. For example, if we like

IPv6 multicast, we can use IPv6 and join other IPv6 Internets. Discrete IP is synonym to perfectness in terms of freedom of choice. Each entity can use their own IP version, even agree on the same one during 1000 years (Discrete IP implements Continuous IP, in this case) until one decides to use a better version of IP. Everything is possible. The real question is not How do we transition to IPv6?. The real question is Who wants what?. Discrete IP gives freedom of choice. It allows competition among different entities. For example an entity may come with a better solution than IPv6 multicast and others may follow. The question of IP transition is solved forever: Everybody do what they wish, and we see what happens. Internet designers can feel better because they will not have to decide for others anymore. Conclusion Discrete IP is better because it allows Continuous IP. Discrete IP allows for faster Internet deployment, because everybody has freedom of choice, it is more bandwidth-efficient than using tunnel brokers. We do not have to worry about others' IP version preferences. China can use IPv6 as they wish. The rest of the world can reach their servers efficiently, without needing to install IPv6, and transition to IPv6 if and when they wish. It can be used instead of NATs. We can install servers at home. It is economically more viable because it allows competition. New solutions are always possible. If the best solution is IPv6 everybody will use it. If it is not, others will use different solutions. If 1000 years later one designs a better IP, those who wish can transition to this new version. The question of IP transition is solved forever.

Das könnte Ihnen auch gefallen