The Petrochemistry of Software

Resourceful as ever mrs. motz pointed us to an interesting presentation a certain Mr. John Day recently held at the "future of networking" conference in Budapest.

John D. Day (born 1947) might not be as well known as Vinton Cerf, Bob Kahn, Larry Roberts or even Jon Postel - really the most important of them all - but like all of them he is a computer scientist, an Internet pioneer and a historian. He has been active in the development of the communication protocols of Internet and its predecessor ARPANET since the 1970ies and has also also been involved in the design of the Open Systems Interconnection Model (aka OSI Reference Modl.

Day is the author of the 2008 book Patterns in Network Architecture: A Return to Fundamentals and the RFC documents RFC 520, RFC 728, RFC 731, and RFC 732.

A critique of the Internet architecture at a historically and technically fundamental level

You will find a pdf of this presentation of Mr. Day's on Future InterNet Design (FIND) at Forum Athena here. That one is for all practical purposes the same as the one he used in Budapest.

Mr. Day is a very classical sort of technology pioneer. Having been there and having co-designed the t-shirt, he knows exactly where the quarrels were and which decisions (in his opinion) went wrong and which problems of today's Internet are connected with which of those decisions. So nowadays he travels the internetworking congress circuit and pokes his fingers into all the wounds, hoping this will result in a real wakeup call.

Everybody knows, we here at tinytalk plc. are buffs for computing and networking history. If you belong to those who know us a bit better, dear reader, you might also know that we are glad about demystification. And though having moved our HQ to Jamaica we also adore any hint at points where we can see that important computing and networking systems still contain technology invented in Europe and not in California or Massachusetts and who the guys were who developed and provided it.

Listen to a short sample of Mr.Day's Budapest talk here. It is an mp3-audio, also provided by Mrs. motz at motz.antville.org.

Moore's law has made us sloppy.

But enough of the preliminaries. Now we'd like to look at just some of Mr. Days concise aphorisms on why software (especially networking code) is still as inmature as it is: "Moore's Law has allowed us to ignore important questions, it has made us sloppy".

After thorough examination and analysis of this stunning proposition the tinytalk tech stuff gave the following variant as a hopefully even more correct alternative: "The phenomenon that Gordon Moore observed (called a law by Carver Mead) in conjunction with the current mode of production and distribution has forced us* to decide for solutions that were mostly not that bad but had no answers for a lot of important questions".

The meaning of these two variants is relevantely different. Whereas the first one implies that sloppy idiots took the wrong bends in the past and did not listen to better informed people (like Mr. Day maybe). The second one points to a set of circumstances that forced non-ideal decisions that got things ahead while paying a price, 'cause, as we used to say, there is no such thing as a free lunch. The same goes for the sometimes mysterious success of programming languages and the strange relapse of software practice into the primitive when microprocessors came along. It is just that sloppiness was forced but not the point.

*Society in general and the software academia and industry in particular

An echo elsewhere

The proposition of "Moore's law driving the computer industry" like fossile fuel driving the economy at large and by its "cheapness" (eq. easy availability) preventing "better solutions" finds a nice echo in Steven Wozniak's hope that software will get better, once the run of Moore's law hits the "atomic barrier". As the Wired Archive tells us:
But his [Wozniak's] hope for the future lies with the end of the curve of Moore's Law. Thinking about when designing personal computers might be fun again, Woz projects himself into a time when advances in chip design will have run up against some hard physical limits. Finally, the box will be defined, and, rather than rushing new hardware out the door every six months, we can all take a breather and ask, "How should the software work with the human being?" To him, the current system is driven solely by new generations of hardware that ship too fast, leaving no time for careful study.

Here's a question from an e-mail to Mr. Wozniak: Woz, I would like to hear your opinions on modern computer directions. Are we where you expected to be by now? What do you see that is exciting for the future?

And here's WOZ's answer: Too vague for me. Too unpredictable too. My latest hopes are for humanistic software but it may take until about 2020, assuming that Moore's law ends for atomic reasons around 2012.

A preliminary observation

Well, somehow the run of Moore's Law (actually an observation that soon became a self fulfilling prophecy) will not end in 2012, as far as we can tell now, but we we will probably see the day in this very lifetime and not only in our next one. However that be, Dennard scaling has ended in 2006 and with it a lot of what was driving the development commonly called Moore's law. Ever more logic gates, multicore and automatic optimization is what's driving progress now.

Post scriptum This is - as so often here on these pages - mere speculation and prejudice - but somehow we cannot get rid of the feeling that our Mr. Day's rather acrid view of the internet architecture stems from a sort of insultedness that although the 7 layer OSI burger, in the creation of which Mr. Day played a large role, had support from the DoD to the complete communication elite of the EC (now EU) never really got its feet moving on the ground.

It reminds us a bit of how Mr. Nelson always trumpeted how wrong the designers of http and html got hypertext. All these guys while being adamant about science have their blind spots and have not studied complex biological systems enough. Still all the fundamentals and Woz give us hope that Moore's law and its ill effects on software will end its run in the foreseeable future. What then?

People get ready there's a train a-coming,
you don't need no baggage just a-get on board.
All you need is faith to hear the diesel humming,
you don't need no ticket no no just thank the lord.

... plink