2008-09-02 Readings - Internetworking Principles

END-TO-END ARGUMENTS IN SYSTEM DESIGN,

J.H. Saltzer, D.P. Reed and D.D. Clark

Summary: Articulates the end-to-end principle/argument, i.e. place functionality at the edge or end pts of the network instead of at the center. Provides examples to justify argument. Basic logic is that functionality is needed at the application level, and so should be provided at the application level to really guarantee the correct functionality. Redundant provision of the same functionality at the lower levels or in the network need not be completely "correct". Examples include reliable communication, delivery guarantee, secure transmission, duplicate msg suppression, FIFO delivery, transaction management, etc.

Background required: None.

Criticism & Discussion Points: Pretty long winded way to put forward a simple argument. Perhaps a cleaner approach would be the needs-definitions-premises-reasoning-conclusions-examples format. Examples in the paper are good. Would be good to talk about best way to gauge trade-offs for functionality in lower layers. Also should talk about deciding which functionality cannot be put end-to-end. Implication for future research - how much stuff to put end-to-end? what are the goals of the network and how does end-to-end help accomplish them? resistance to failure?



THE DESIGN PHILOSOPHY OF THE DARPA INTERNET PROTOCOLS,

David D. Clark

Summary: Outlines and discusses original goals of the DARPA internet, and discuss at length their implications on the internet architecture. The fundamental goal was "effective technique for multiplexed utilization of existing interconnected networks". Second level goals include survivability in the face of failure, accommodating different types of service, and interconnecting a variety of networks. Other goals including distributed management, cost effectiveness, low effort host attachment, and accountability are less effectively realized. Original internet a product of military thinking - minimalist design. Current use of internet - civilian use. Somewhat mismatched priorities.

Background required: None.

Criticism & Discussion Points: We need more historical paper like this - trace knowledge back to its beginning. The mismatched original design goals and current uses are definitely worth discussion. Perhaps of more important interest is design methodology. Perhaps military application projects tend to be able to get at the core fundamentals of a flexible, survivable technology, unhindered by any economic, business, or other non-technology constraints? What's the best way to invent a new tech - from corporations or from research or from the military?