Sunday, May 03, 2009

John's apocalypse




Several years ago (20?), I coined the term NET8 (pronoounced "Net theta" to mean network theory, later mutated to INTERNET SYSTEMATICS as I became more profficient with Metasystem Transition (+). this is how I perceived my attempt to understand the emerging the Net, that I happened to confront as a Computer Scientist in mid 80's (I blog my learning path slowly). Previous postings here indicate how the associated thinking developed along the practical task of building operational networks.

A couple of years ago, I read a paper by David Turner (my mentor into Functional Programming) on the significance of the theory of computing; the works initiated by Alonzo Church and Alan Turing that enable us to avoid the babel of Intel-computing vs Motorola-computing and Java-computing vs C-computing by leading us to a convergence of the corresponding concepts.

This somehow supported a further unfolding of my "naive" NET8 plan (a quest for fundamentals in computer networks).



A recent book by John Day "Patterns of Network Architecure" in 2008 by prentice hall came just in time that I was wandering about my next step.

First thing it does for me:


is to fill the gap to appreciate the case of OSI (due to my lack of the necessary experience and all round factual data concerning OSI). I know now that the "harm" I detected is confined to the OSI protocols but the OSI model is undeniably "good", for example the first to start the thread of separating the Application from the Protocol.

It also brought memories of MINITEL, CYCLADES, TRANSPAC, TELMAT SM90 my French initial-connection to Networking (due to Nicolas Malagardis of IRIA, INRIA).

It also gave me an immensely powerful insight to distinguish Engineering things from Science ones. I discovered many common things with the book like "General Systems Theory" (von Forester for John Day and Valentin Turchin in my case).

Now I can understand that nothing beats experience, I am a very simple seer on the same direction for the past 20 years as "seeking patterns" that understand the Network in Computer Science terms but I stayed at the periphery of the WITI problem (another coinage from me What Is The Internet) I never thought that the whole Net thing we have needs replacing. I somehow thought that the "onion layers" (an OS concept !) of the evolutionary attached Quanta of Automation (see previous postings) were the fittest that survived, optimal in the given circumstances for the Network ecology. John Day brings the realities that point to "quick fixes" (*) like the infamous NAT (Network Address Translation)

John Day is master and a seer (his term), he states:
In Computer Science we build what we measure


he sees too much Engineering and little if any Science on the Internet phenomenon because network things worked "off-the-shelf" from the start of the ARPANET unfinished Demo. There have been opportunities to rectify the situation but the inertia of success, vested interests and mainly lack of knowledge in the wider IETF community (!!!), "group-thinking can be dangerous" warns John Day, as an example he points to CLNS. (I did underline its prospects as a reporter in 1994 but only by accident and influenced by MTR's (M.Rose) ISODE package).

I could not have given such a focus to my research (my proud achievement in 1994 about "OSI can be harmful" apparently only tells half the truth). John Day says he did not even plan for a book but only for stock taking about lessons he had gained until then (2007-8) in order to understand Network independant of politics, religion and constrain of technology (i.e. in Science terms). Gradually he saw patterns forming that he worked out how to yield them.

In my case I thought of "theory" (again naively of course as I was extracting abstractions, not avoiding embarassment at times) because the Science style of Declarative Programming concepts conditioned my participatory observation. The above diagram shows a pattern I saw related to Network architecture, as an early attempt to codify my thoughts.

I will get to a comparative consideration between RNA (RINA) and NET8 but the most important issue that my mind is on the 6th in Boston where the PSOC meets (Louis Pouzin society, he invented datagrams), here is a note concerning how to clean slate the Net.

I wonder how many colleagues in Greece and Europe know that datagram is a French doing ? (pls leave a note if you can). In fact the French PTT (^) destroyed
CYCLADES the first datagram network !!! and developed TRANSPAC instead. I built the first
X.25 node link at 4.8K to it for Greece in 1987 collaborating with the Greek PTT (OTE).

I could not understand some negative comments the book got in Farber's list and waited for clarification. The list runs on a very tight self-controlled spirit so nothing came. But happily the Internet channel is amazing (Internet Protocol Journal) Jon Crowcroft of UCL wrote a very good review of the book for example he writes:

I found the book extremely readable and enjoyable, and although I might argue with some of the opinions in the book, I think that this is just more evidence that I should recommend the book to anyone interested in knowing why we are where we are in networking, and being better informed about where we should go next.

So my new discovery got a validation. I have followed such a course all along the NET8 path ie waiting for "proofs" because I realise my weak position to handle such
a big issue as Network Theory. This is why at times a give some sort of thumbs up sign to myself for encouragement.

(+) I found MST as a way out to explain to myself the fact that Functional/Applicative/Declarative model of computing did not conquer the world, as I had finished my research slot in that area. So I was looking for "killer aplications environments" to utilize the great works that C.Strachey had initiated. When I met ASN.1 (also LOTOS for a while) having entered networking for adventure in the mean time, it started ticking in the back of my mind that "the Net is a new kind of machine (SUN microsystems declared the infamous Net is the Computer phrase) that F/A/D model can be applied to". Next tick was with Active Networks of Farber so joined his previous IP maillist (F/A/D means Functional/Applicative/Declarative), the best spin-off of my Marco Polo like journey. Finally, the puzzle is solved this is why I use the title apocalypse above, Declarative networking exloits the separation of What from the How of protocol design. The "What" brings policy issues the "How" brings mechanism. So TCP and UDP for example in terms of "How" are the same protocol, in terms of "What" are different specification policies. Layering should shift 90 degrees left to separate things horizontally because fundamentally there is only one layer that recurses across different scope, bandwidth demands etc, John Day says. But recursion (ie Kurt Goedel's recursive functions) is the beginning of Computing, the ultimate of fundamentals. I went as far back (with my Automation Quantum as a fundamental concept of a model for the Net) to Turing Machine, so it is not a bad go I guess for an Net newbie
in 1999. Further, my VVNA - Virtual Von Neumann Architecture
does use the idea of a recursive phase in Net's evolution path (please see previous postings otherwise I seem cryptic here, in any case do query me ...).


(*) the phrase from BEATLES'LP Sergeant Pepper "I am fixing a hole where the rains gets in and stops my mind from wandering ..." is appropriate to the situation

(^) a peer to our greek OTE, I help plant ISP OTENET there in 1995 under the belief "first we take Manhattan from within", as the L.Cohen' song goes.