Rethinking of the design of the Internet:
The end to end arguments vs. the brave new world
Blumenthal and Clark.
Review by Kiran Komaravolu

In this article, Blumenthal and Clark analyze the ^Óend to end^Ô argument and its applicability in today^Òs world. In the early days of the Internet, Saltzer et al., proposed the ^Óend to end principle^Ô which has gone on to become one of the fundamental design guidelines of the Internet. The end to end principle stated that for a function can be correctly implemented only with the help and knowledge of the applications standing at the end points of the communication system. In the case of the Internet the end point application was the application layer. Thus a function can be correctly implemented only with the knowledge of the application layer. This prompts more efforts and complexity to be dedicated to the correct implementation of the application layer.

This has been the general philosophy while designing the Internet. The core of the Internet provides and a very general data transfer mechanism with many different types of application making use of it. Ultimately it is the responsibility of the application to make sure that the data transfer/communication has occurred correctly.

However in today^Òs world it is not always feasible to follow the end to end guideline. Reasons being the expansive growth of the Internet resulting in an insecure world with untrustworthy clients, software and nodes. Also with the different kinds of networks which have been deployed, it may not always be possible to strictly adhere to the end to end principle.

As examples the authors cite the use of mail servers to eliminate spam email. Firewalls on gateways are used to protect clients from the rest of the world. Traffic filters mayve deployed to control access to different types of content. Content caches are used to reduce the burden on the network. NAT boxes are used to combat the shortage of Internet addresses. Gateways for wireless clients (cellphones using WAP) violate the end to end guideline by providing a proxy mechanism to improve network performance.

These examples are enough to prove that it may not be feasible to use the end to end principle under all circumstances. However the importance of the end to end rule has not diminished. Deploying the end to end rule still preserves the flexibility and openness of the Internet. An effort to use it should be made whenever possible.

Overall the article looks more like a conference talk rather than a research paper. The article as such does not present and new ideas. Also (not surprisingly) the paper does not have any kind of experimental work. It may be noted than some of the points presented in the paper were discussed in the class when the original paper on end to end arguments was presented. The article though makes interesting reading having been written by one of the original designers of the end to end principle. I would definitely recommend this article for someone to read.