Monday, September 1, 2008

The Design Philosophy Of The DARPA Internet Protocols

This paper gives the reader a perspective not always seen when regarding an architecture - not just the final result, but the reason that the design ended up that way. The initial goals for the creation of the packet switched network were responsible for many of the design decisions that shaped the internet. However, since the DARPA project was primarily based with regards to the military, those goals were quite different than those from a corporation might have been, placing durability first and accountability last.

In order to maintain durability given a network node was lost led DARPA to remove state from such nodes, instead transferring state information to the systems themselves. The project also evolved to support multiple types of service after it became clear early on that certain applications would prefer to forgo reliability in place of having speedy delivery and an occasional dropped packet, leading to the separation of TCP and IP, and the creation of UDP. From the very beginning, the purpose of this project was to join many different types of networks together, thus a minimum standard was set for a network to be able to carry connections. There were other goals, but they simply did not get the same attention as the primary goals, and many were only approached as an afterthought.

Since the original intent of this project was largely for military purposes, it is quite interesting to see the consequences of a military focussed protocol being used by the public. Accountability was really an afterthought of the internet, most likely because the military could assume that all partners would cooperate. However, in the consumer world, there is a much higher lack of trust, and potentially many more malicious types.

The writer also talks about the 'next' protocol and what considerations should be put into it, but at this point, can there ever even be a next protocol? IPV6 has been around for 10+ years and it has barely gained any ground in that time period. IPV6 is only one part of the protocol - if IP alone can not successfully be updated, then any bigger changes are surely going to be exponentially more difficult to introduce. We have already used these protocols for a lot longer than the original designers intended, and the future does not look any brighter.

This is a great paper because it gives some context to the protocols that are standard today. You can never quite escape your past, so you might as well find out why. After all, I just read that the width of train tracks is directly tied to the width of two horses!

No comments: