Patterns in Network Architecture: A Return to Fundamentals

I was involved in the design of the ARPAnet Protocols and worked on Internet protocols. For about a year I was the chair of of the ISO Presentation Layer committee. I have a lot of first-hand experience with where we are today and how we got here. I've just finished "Patterns in Network Architecture" and I'm impressed. While there are many things to argue over, the architecture Day describes sounds like it should work, and if it does work should be a vast improvement over the current ramshackle protocol structure. I haven't really done the hard work necessary to figure out if it really will work, and I don't expect to. And we know that the devil is always in the details. But I hope to hear sometime that an implementation has succeeded. Like many others in the field, John has strong opinions and axes to grind. Rather than a shortcoming, however, I think his strong opinions and sarcastic wit are a benefit; any reader can see clearly where John is coming from and can research the "other side" if it seems important. On page 79, Day writes: "... two types of protocols tend to alternate in architectures. The MAC layer does relaying and multiplexing, the data link layer does "end-to-end" error control; the network layer relays, the transport layer does end-to-end error control; mail protocols relay, hmm no end-to-end error control and sometimes mail is lost. ...we can make two observations: 1. Relaying always creates the opportunity for PDUs to be lost. Therefore, to guarantee reliability, there must always be an error-control protocol on top of a relaying protocol. 2. This would seem to indicate that there are really only three fundamental types of protocols: - Two data transfer protocols: Relaying and multiplexing protocols and error- and flow-control protocols with different policies - Application protocols" The rest of the book goes on to develop these thoughts, which in my view is a real breakthrough in clarity. This book ought to be used as a textbook! It is highly recommended to everyone working in the field. If only there weren't so many distracting editorial errors.

Working to uncover the "first principles" of networking, Day places the development of networking within its historic context, interweaving protocol design decisions with fundamental concepts of philosophy, mathematics, and computer science. It is not too much to say that this is the Godel, Escher, Bach of networking. Santayana said that "those who cannot learn from history are doomed to repeat it." Readers of this book will leave with a solid grounding the decisions leading to the current state of networking, and appreciate a provocative viewpoint on the way forward.

John Day has created a master technical work on the theory of networking for our consideration. This is the most instructive and intriguing book on the subject I have read since Uyless Black's "OSI: A Model for Computer Communications Standards". John's book is informative both on a technical basis and a sociological one. In it he explains much about the history of standards in both the Internet protocols and the OSI standards. I have been involved in standards work and have seen the dynamics he describes, and thoroughly enjoyed this telling of the history of how we got to where we are today. On the technical side, I think his recursive "one-layer" model is elegant (The Distributed IPC Facility, DIF). To me it encapsulates what we see happening in all layers -- that is they all seem, at some point, to borrow from the requirements of others to perform their services. I especially enjoyed the scalability and ability to tailor implied by separating mechanism from policy. Some other key elements were the emphasis that addresses much change from physical to logical at least once; that we need to distinguish between topological address, node address, and application address; that there is a continuum of function between connection oriented and connectionless messaging and how they can change roles from one DIF to another in the goal of achieving the desired Quality of Service (QOS). What matters most is that the Application can convey the needs of QOS along with a message so that the DIFs can affect the appropriate and optimal transfer. I heartily recommend this book for anyone working in applying network communications to new application areas and especially those involved in standards work. Thanks, John for a great read.

Students of network architecture and teachers assigning reading will find PATTERNS IN NETWORK ARCHITECTURE an excellent survey which takes a different approach to solving network architecture issues. Seven basic, still unanswered, questions identified during the ARPANET's development reveals protocols, patterns, and common conflicts. The result is an outstanding guide for any advanced networking computer library.

When it comes to the kind of people involved in computer networks, there are four different types; the architects, engineers, IT professionals, and the end users. The architects design, the engineers build and maintain, the IT professionals configure for the unique business purpose, and the users work on it. This book is written by an architect for architects (and engineers aspiring to be architects). I'm doing this review with the perspective of someone who works mostly as an IT professional but spends about 35% as an engineer. With many endeavors, it is easy to focus on the short-term with little or no emphasis on the long-term. John Day, as seen through this book, has both the unique experiences of designing and addressing very specific technical topics but also standing back and looking at how networks have evolved in perspective historically and where they need to go. This kind of work is indeed extremely important as our world becomes more interconnected every day, knocking down communication barriers and making more critical information available to people everywhere. We need to closely examine where and why the Internet has ended up where it is today so we can make the best long-term decisions for the future and that is exactly what John Day does in Patterns in Network Architecture. This is very technical book that brings detailed processes together through both history and theoretical patterns. I can see this book being used in educational environments concerned with network architecture design (103 level classes) and organizations that place a high amount of significance on practical theory. I'm giving this book a five because of the amount of detail it covers and the flow he keeps throughout the book. Most writers covering this type of information get lost in the logistics but I felt like I was engaged at a lecture (instead of studying after a lecture).