User:Beartham

I am Joseph "Bear" Thames, one of the originators of the mathematical software paradigm known as "MetaCalculus". I received my BS degree in Mechanical Engineering from Lamar University in Beaumont, TX in 1961, and my MS degree in this same major from Rice University, in Houston, TX in 1963. I started programming, using FORTRAN on an IBM 1620 at Rice to solve the gas-dynamics of a theoretical plasma rocket for my master's thesis, "Non-equilibrium Nozzle Flow of a Partially Ionized Gas". I have been developing software continuously since then, and I doubt I will stop during this lifetime, as I still enjoy it immensely.

I worked as a summer intern in 1962 on Apollo reentry aerodynamics at North American Aviation in Downey, CA, and I joined NASA in Houston in April 1963 as a propulsion engineer, just before I recieved my MS in June. Four years later, I was promoted to section head of the Primary Propulsion Systems Analysis Section there, responsible for the flight analysis of the Apollo spacecraft propulsion systems, SPS, LEMDE, and LEMAE. I had designed and developed key software that was employed to model these systems, used to mathematically reconstruct their inflight behavior by the method of state estimation (AKA data assimilation).

This software originated the technology of MetaCalculus. With its demonstration in the first unmanned flight of the Apollo spacecraft, firing the SPS engine over the Antigua tracking station in the South Atlantic, I received NASA awards for this work, which led to my promotion. Later in the program, the accuracy of this software in estimating engine performance of all three of these systems was instrumental in confidently gauging Apollo payload capability necessary to carry the Lunar Rover vehicle to the moon in later missions.

Soon after my promotion, and the Apollo 1 fire, I left NASA to continue pursuing the development of MetaCalculus at TRW. Later in 1972, I founded a spinoff, PROSE, Inc. to distribute its first commercial product, the PROSE language, in the time-sharing market. In 1975, an interactive version was introduced featuring interactive optimization via iteration breakpoints.

In 1978, a PROSE user group was formed, and with the waning of time-shared outsourcing as a viable market by 1980, its originator, Frank Pfeiffer, and others started lobbying the academic mathematics community to start campaigns to investigate and publish the breakthrough technology of automatic differentiation (AD), which became the Autodiff movement in 1991, focused on the underlying semantics of MetaCalculus.

In 1980, I joined the Aerospace Corporation with a computer-architect colleague, Arthur Speckhard, intending to use this non-profit think tank as an incubator to build new hardware architecture that could support calculus-based virtual machines like PROSE, in RISC microcode. We were motivated by the notion that PCs like the Apple would soon take the place of time-sharing for DIY modeling in R&D and education, while eventually obsoleting mainframe computers altogether. But the early PCs were far too primitive, especially with 16-bit segmented-memory addressing, to support the high interpretive-overhead of AD arithmetic in software, which was a prohibitive concern in those days. So we started designing a Writable Instruction-Set Computer (WISC).

I took a technology-management position in the Space Shuttle program office, and also managed the continued development of this experimental computer. My position was responsible for a mini-computer CAD lab for shuttle flight planning with NASA's Flight Design System (FDS). This lab supported Air Force "sortie mission" applications of the Shuttle, which was slated for major expansion with SDI missions and the polar-orbit shuttle launch facility, (SLC-6) at Vandenberg AFB. But the Challenger failure cancelled all this activity.

After four years of this computer development, the STS program-office was re-organized and I was working for a new boss, George W. Smith, who investigated our projects and discovered the computer we had built. As a former entrepreneur himself, he immediately decided to join Speckhard and myself to spinoff a new company. He had the backing of Aerospace Corporation management, and negotiated an exit agreement for us to form the spinoff, International Meta Systems, Inc. (IMS) He also arranged a budget for Speckhard, and an associate Lee Scantlin, to design a VLSI chip version of core RISC engine of the WISC architecture.

Over the next two years IMS was engaged in fabricating the new 32-bit RISC microcomputer chip (the IMS 3230) using the LSI Logic foundry; manufacturing the IMS "MAX-2" PC-AT ISA board "metasystem" (summarized here), which used the RISC as its core; and developing its first microcoded virtual machine, a FORTRAN 77 emulator, meant to become a Fortran Calculus VM, as the PROSE successor. A patent application for this Microprogrammable Language Emulation System design was filed in 1987.

Unfortunately the prototype chips only ran reliably at half speed as a result of a fabrication flaw. So the early Whetstone benchmark tests of the 32-bit FORTRAN VM (without floating-point hardware) showed only a factor of ten gain over similar benchmarks of the Intel 386 Fortran (with floating-point hardware)! Yet IMS was in a financial crisis at the time, and Smith (the CEO), who was not a software expert and had not been part of PROSE history, decided to abandon pursuit of scientific markets for the new computer, and instead pursue office automation by focusing on Smalltalk 80. As a result I left IMS, giving up half my founder's equity to protect my proprietary rights to MetaCalculus.

Thereafter, I immediately acquired a contract with Dupont, which enabled me to develop FortranCalculus 77 on Intel 386 computers, while still protecting my proprietary rights to the technology. I resorted to an AD compilation approach similar to MC1, piggybacking on FORTRAN compilers, eliminating the VM overhead which had motivated hardware development in the first place. Yet by 1990, I was now faced with GUIs, the new cosmetic veneer of computing, and soon after, client-server interaction and the web. Yet, with defense industry downsizing, venture capital virtually dried up in California, where I was based, and I soon ran out of my personal capital. So I had to either seek another full-time management job or do free-lance software contracting to upgrade and maintain my software skills. I chose the latter as the means of learning the new languages like Visual Basic, Perl and JavaScript to complement my FORTRAN skills.

Having to put MetaCalculus on hold, I spent the next decade in free-lance contracting and founded two ill-fated (under-funded) software ventures. In the first company, MetacyberNet, I created and patented a new web-tooltip sub-hypertext wiki generation system for unifying software source-code and documentation. We now call this new medium "Spiritext". I expected Spiritext would be eventually adopted for all programming languages in the future. Moreover it is a natural aid to MetaCalculus pedagogy and scientific collaboration, as illustrated by the previous link. I was also motivated toward this approach because it is a natural extension of language compilers via compiler-compilers, which we had perfected to a higher software-automation art called Metacybernetics at TRW and Aerospace Corporation. In fact, we had extended the metacompiler TREE-META to become the "virtual machine ('metacomputer') generator" for the WISC "metacomputer host" architecture. This was to enable automation by rapid-iterative prototyping (RIP) of new semantic-intensive "XGL's" on these machines

In 2004 when I could afford to renew work on MetaCalculus, I started working to assemble a cloud XGL automation laboratory as a Linux "Penthouse" with new Metacybernetics apparatus. Its intent is offer our source-code to the industry to usher a build-out of new modeling 5GL (e.g. MetaCalculus Fortran 95 dialects) 6GL (e.g. MetaCalculus Pythons), and 7GL (application icon-pallet GUI Languages, derivatives of 5GLs and 6GLs) appliances. In essence the 5GLs can become the new platforms upon which the 6GLs and 7GLs will be built. The first stage of this build-out will concentrate on creating GUIs unifying MetaCalculus and the derivatives of Forrester's System Dynamics stock-flow icon-pallet motifs, which can enable MetaCalculus to reinforce and add momentum to System Dynamics and System Thinking initiative already revolutionizing K-12 education.

The objective is to start a pendulum swing in computing back toward science modeling from which it initially grew with FORTRAN and BASIC. There are unique features of MetaCalculus that can catalyse this renaissance. Specifically, software technology since the FORTRAN era has become largely a recycling enterprise which is now heavily balkanzied (e.g. Unix into Linux into Andoid, etc.). MetaCalculus can turn this propensity into recycling old simulation applications into optimization, sufficient to turn the industry around and stimulate new economic growth that has been absent since the 1970s.