High Frontier Computing 1958 to 1982 — Section-by-Section Summary
This is a detailed summary of High Frontier Computing 1958 to 1982, a history of numerical modeling and simulation in Texas, chiefly UT Austin and Houston, during the mainframe era. Its through-line is that Texas computing grew from practical needs in oil, aerospace, and defense, not from science-fiction imagery. The narrative follows a few breakthroughs — Simplex, ADI, SOR, subsurface modeling, and orbit determination — from Los Alamos through RAND and TRW and Houston's Exxon and Rice to the UT Computation Center.
The document is organized into nine numbered chapters and a conclusion. Chapter 1: Space Age Texas sets the political and institutional stage; Chapter 2: George Dantzig And Simplex and Chapter 3: David Young And SOR trace the two algorithmic currents that reached Texas from RAND and TRW. Chapter 4: Modeling The World, Chapter 5: Mary Wheeler And ADI, and Chapter 6: Bob Schutz And UTOPIA follow modeling outward from the subsurface to Earth orbit. Chapter 7: Galaxy In The Machine, Chapter 8: New Frontiers, and Chapter 9: Pioneers turn to the interactive, networked, and social computing of the CDC 6600 and DEC-10 era, and Chapter 10: Conclusion argues for preserving this history through code and people.
Chapter 1: Space Age Texas
World-class computing took root in Austin and Houston where Cold War imperatives, oil-industry demand, and shrewd political maneuvering converged, largely unnoticed at the time behind the louder Texas mythology. Lyndon Johnson steered federal money to the region, drafting the National Aeronautics and Space Act of 1958 and anchoring the Manned Spacecraft Center in Houston; his influence also underwrote UT Austin's 107-inch telescope at McDonald Observatory, completed in 1968, and a campus engineering revolution whose centerpiece was the underground Computation Center built for the CDC 6600, facing the LBJ Library across the East Mall. Byron Tapley's career — a bachelor's in 1956, the orbit mechanics program in 1961, department chair from 1966, and later the Center for Space Research — exemplifies a first generation of the Space Age active for more than fifty years.
Big Science And Engineering
After Sputnik, the 1958 National Defense Education Act and National Aeronautics and Space Act recast education funding as national security, militarizing its justification and creating both the modern research university and the federal student-loan system; UT's engineering advanced from desktop mechanical calculators to supercomputing within a decade. The era's faith in "big science" — centralized, computer-driven problem-solving across the Manhattan Project, the national labs, the ICBM and manned-spaceflight programs, and SAGE — gave way in the 1970s to an inward-looking, Earth-centered environmentalism, catalyzed ironically by Apollo 8's 1968 "Earthrise" photograph. A deliberate policy of knowledge transfer moved defense researchers back into universities: Harvard Underwater Sound Lab veterans such as Paul Boner founded UT's Defense Research Laboratory, later ARL:UT, and Tapley's ties to NASA's Jet Propulsion Laboratory produced the "Texas Mafia" of UT alumni, breaking the coastal monopoly on advanced scientific work.
IBM CPC At Humble Oil And UT Austin
The story began in chemistry, where Al Matsen secured grant funds for an IBM 650 in 1955 — the first real computer on campus, used for the Quantum Chemistry Integrals and Tables — and shared it so widely that a UT president once complained about a "computer center" that did not officially exist. In 1958 Humble Oil in Houston donated its IBM CPC, the very machine on which Rachford, Peaceman, and Douglas had created ADI in 1954; Matsen, a consultant to Exxon for over thirty-five years (the source of the "What possible use could you be to Exxon?" anecdote), and his students carried its components into Welch Hall by hand. The CPC was a step back from the 650 — an externally programmed electro-mechanical hybrid of 604 calculator, 402 accounting machine, and storage units in which the program existed only as a physical card deck — and was the same class of machine Dantzig had used for Simplex in 1952 and that descended from Feynman's Los Alamos pipelines. That same year David Young arrived from Ramo-Wooldridge to found the Computation Center — at first just himself, Robert Gregory, a secretary, and the 650 — and within eighteen months leveraged his reputation into a $400,000 NSF grant for a CDC 1604, then in 1966 the first $1,000,000 grant toward a CDC 6600. "David Young did it," as Gregory put it, making UT the first university with a 6600, ahead of Berkeley, Stanford, MIT, and Harvard.
CDC 6600
UT installed the $6 million CDC 6600 in the summer of 1966. As the fastest computer in the world — three times the IBM 7030 Stretch — it placed UT alongside Los Alamos, Livermore, and CERN. Designed by Seymour Cray and James Thornton and introduced in 1964, it offloaded operating-system housekeeping to ten peripheral processors so its 60-bit central processor, with ten parallel functional units and a scoreboard, could issue an instruction every hundred-nanosecond cycle and reach roughly three million floating-point operations per second; parallel design was in the center's genes, making it more national lab than typical university. The 60-bit word, with its 48-bit fraction and indefinite and infinity bit patterns anticipating IEEE 754, was the pragmatic answer to shock-wave simulations that overwhelmed 36-bit machines. Cray's obsession with signal delay dictated the machine's tightly packed, freon-cooled, four-wing form, which the underground limestone building was built to house. Young stepped down as director in 1970 to found the Center for Numerical Analysis and its ITPACK work, and the 6600 began a CDC-to-Cyber-to-Cray lineage whose "63 character set" independence seeded an unbroken supercomputing tradition.
Chapter 2: George Dantzig And Simplex
Dantzig's Simplex method for linear programming drove commercial computing and reshaped the oil industry. LP solved the refinery "blending problem," ending the "quality giveaway" of over-blending high-octane components. To run it, oil companies became early mainframe buyers, effectively subsidizing IBM's move to mass production; that spending funded the UT Austin and Rice computation centers and set up the "Texas model" feedback loop, in which energy profits underwrote university computing and the aerospace and semiconductor booms. The method grew from Dantzig's wartime work in the Army Air Forces Statistical Control Division, the "Whiz Kids," and a 1946 challenge to mechanize planning produced the 1947 LP formulation. His work founded Operations Research and Systems Engineering.
Simplex Linear Programming
The 1947 LP formulation and Simplex algorithm were "uniquely machine-ready," co-evolving with the first stored-program computers; the National Bureau of Standards' SEAC solved the first electronic LP problem in 1951. Simplex rested on a geometric insight: the optimum lies at a vertex of the feasible convex polytope, reached by iterating along its edges. Dantzig illustrated the brute-force alternative — assigning 70 men to 70 jobs by permutation would occupy "a solar system full of nano-second computers running from the big bang until the universe grows cold." Because early "computers" were rooms of desk-calculator operators, mostly women, the first programmers were often women such as Ida Rhodes. Dantzig implemented Simplex at RAND from 1952 with Orchard-Hays in octal, before Fortran.
RAND
Founded within Douglas Aircraft in 1946 and made an independent nonprofit in 1948, RAND was the first R&D lab dedicated to modeling and simulation, comparable only to Los Alamos and Bell Labs. It institutionalized systems analysis and game theory through von Neumann, and a 1954 study shifted US nuclear doctrine toward second-strike deterrence. As an early "hacker lab," it built its own JOHNNIAC in 1954 and produced the birth of symbolic AI under Newell, Simon, and Shaw and Paul Baran's distributed packet-switching (1964), of which he said "I built a cathedral." Much of this migrated to RAND's spinoff SDC and the SAGE Q-32, and thence to UT Austin through Forest Baskett.
IBM CPC At RAND
The "heroic age" of 1952 meant implementing Simplex on the IBM Card-Programmed Electronic Calculator (CPC), a hybrid of punched-card machines demanding constant human intervention, whose ancestry reached the punch-card machines run by Feynman at Los Alamos in 1943. Because the standard tableau was impossible on the CPC, Dantzig and Orchard-Hays reinvented the algorithm as the Revised Simplex Method (1953) and the Product Form of the Inverse, storing the inverse on cards to exploit sparsity — establishing the principle of computational economy that underlies high-performance computing. In 1958 UT inherited an IBM CPC from Humble Oil through Frederick (Al) Matsen, carried into Welch Hall by hand.
Exxon And Bill Drews
Dantzig's 1960s collaboration with Exxon's Bill Drews carried LP from defense into global industry; their 1962 Harvard Business Review paper "Mathematical Approach to Long-Range Planning" extended LP to multi-decade capital planning. Exxon formalized a Mathematics and Systems group in 1967, and its earth-science counterpart, Exxon Production Research on Buffalo Speedway in Houston, pioneered 3D seismic imaging and reservoir simulation and drove demand for Cray supercomputers, sharing with the aerospace labs a reliance on the finite-difference method and the "inverse problem." The UT-Rice-Buffalo Speedway triangle was porous, Exxon employing over 2,500 UT alumni.
Chapter 3: David Young And SOR
Three algorithms were born almost simultaneously in the early 1950s: SOR from Young, ADI from Rachford, Douglas, and Peaceman in Houston, and Simplex from Dantzig at RAND. All were tied to the IBM CPC or its precursors, and all established linear algebra as the tool for solving physical PDEs and optimization problems. Being pre-Fortran, these methods justified heavy investment in mainframes and hardware floating-point, and they marked a shift from artful human methods — Southwell declared relaxation unmechanizable — toward an automated science that could answer, at each step, "What do I do next?"
SOR Successive Over-Relaxation
SOR and Simplex ended the human-computer era. Discretizing a PDE onto a grid produces a sparse matrix, and direct elimination suffers ruinous "fill-in": a 100×100 grid yields 10,000 variables and 100 million storage locations, impossible on early machines. SOR instead sweeps the grid using only the non-zero coefficients, keeping memory proportional to the variables. Its "over-relaxation" factor omega extrapolates each correction; Young proved that for matrices with "Property A" the optimal omega gives an order-of-magnitude speedup. At UT, Young founded the Computation Center and the Center for Numerical Analysis, whose group produced ITPACK.
Ramo-Wooldridge And TRW
The mid-1950s ICBM program demanded fast PDE solving for atmospheric reentry. Ramo and Wooldridge pioneered a systems engineering built on digital simulation and in 1955 recruited Young to head the Numerical Application Section. Because Soviet rockets were far more powerful, the US compensated with extreme structural optimization — Ramo's "ten-to-one" structure-to-fuel ratio left the Atlas so light it would collapse unfueled — which meant solving huge sparse matrices, and Young's SOR was the critical tool. SOR-driven reentry simulation cut the Atlas from 460,000 to 240,000 pounds, and the resulting Fortran programs became the repository of US aerospace knowledge — a legacy Young carried to UT in 1958.
UNIVAC 1103A Scientific And Floating-Point
TRW's UNIVAC 1103 relied on fragile Williams-tube memory so temperamental that spark-plug noise from a passing car could flip bits and ruin a long SOR run. The 1103A of 1957 replaced it with reliable magnetic-core memory and added optional hardware floating-point, a shift from slow software emulation to single-instruction hardware 10 to 50 times faster. It opens an extended history of floating-point: Zuse's Z3 (1941) was the first binary floating-point hardware, anticipating infinity and NaN, though American doctrine favored fixed-point; the IBM 704 (1954) was the first commercial such machine, with Backus co-developing Fortran, and set a 36-bit standard; and before IEEE 754 (1985) there was no standard at all — a "zoo," in Kahan's term, whose worst offender, the IBM System/360's hexadecimal "wobbling precision," opened the market to the superior 60-bit machines from CDC and Cray. Young implemented SOR on the 1103 and 1103A from 1955 to 1958, both designed by ERA — later CDC — foreshadowing the scaling toward the 60-bit CDC 6600 (1966).
UT Austin And Numerical Analysis
Young moved to UT in 1958 to found the Computation Center, at first just himself, Robert Gregory, a secretary, and an IBM 650. Within eighteen months he secured a $400,000 NSF grant for a CDC 1604, and in 1966 the first $1,000,000 NSF grant toward a CDC 6600 — "David Young did it," as Gregory put it — making UT the first university with a 6600, ahead of Berkeley, Stanford, MIT, and Harvard. The period also saw numerical linear algebra mature: Wilkinson's backward error analysis rescued Gaussian elimination through partial pivoting, and the singular value decomposition of Golub and Kahan (1965) became a "Swiss Army knife" for ill-conditioned matrices.
Chapter 4: Modeling The World
Through the 1960s the engineering complex at Speedway and 24th reshaped itself, formally becoming the Department of Aerospace Engineering and Engineering Mechanics in 1968 and bringing Tinsley Oden in finite element methods and Byron Tapley in orbital mechanics and geodesy. Over the 1970s, modeling grew from isolated point-solutions to whole-field FDM and FEM simulations and to satellite geodesy of the Earth's geoid, plate tectonics, and mantle convection — a shift Oden framed as computational science becoming a "Third Pillar of Science" alongside experiment and theory. The unifying theme was that Texas's land and geology extended outward to orbit through computing, satellites, and laser ranging, with pioneers such as Tapley, Schutz, Duncombe, and Wheeler recognizing that satellite orbits were directly linked to the Earth's gravity, geology, and subsurface. The Navy's Transit system, first launched on 13 April 1960, revealed the Earth's asymmetrical shape and established the satellite as a "test particle" for probing gravity.
Transit Satellite Network And Ray Duncombe
The roots of satellite gravity modeling trace to Sputnik on 4 October 1957, when Ray Duncombe and Paul Herget of the US Naval Observatory worked around the clock to compute its orbit, Duncombe solving the path graphically by hand to produce the first American calculation of the Soviet satellite's track; he would later link UT's McDonald Observatory to the Naval Observatory, reflecting the deep maritime ties between astronomy, timekeeping, and navigation. Minitrack, the first US satellite tracking network — a passive radio-interferometry "fence" strung along the 75th meridian from Blossom Point to Santiago — was built for the 108 MHz IGY standard and had to be improvised onto Sputnik's 20 and 40 MHz signals in about two weeks, with crews lashing up dipole antennas from "2x4 end supports and lacing cord." Precise results leaned on Doppler analysis, the method that would mature into the Transit navigation system and, through UT's tracking work with CSR and ARL, feed the ITRF, ICRF, and IERS frameworks and the foundations of WGS 84 and GPS.
McDonald Observatory And West Texas
Laser ranging became central to space geodesy. Apollo-era lunar retroreflectors enabled centimeter-level ranging, and the faint lunar return demanded a large telescope — a key reason NASA funded the 107-inch telescope, which returned its first lunar signals on 19 August 1969 and remained the world's only routine lunar-ranging facility for over fifteen years. Satellite laser ranging developed in parallel through LAGEOS (1976, the first to measure tectonic plate motion, polar motion, and the Earth's mass). The geology mattered: McDonald sits on stable ancient craton bedrock whose dense trachyte anchors the 160-ton telescope. Laser ranging yielded Earth Orientation Parameters, North American Plate drift, and support for sea-level altimetry; the Moon recedes 3.8 cm a year.
Satellites To Subsurface
The relationship is reciprocal: a known ground position resolves an unknown orbit, and a known orbit resolves an unknown ground position. Minimizing orbital error shifted the emphasis from navigation to geodesy, treating the satellite as a gravitational test particle, which required "cannonball" satellites with tiny area-to-mass ratios such as Starlette (1975) and the dense LAGEOS. At UT, UTOPIA on the CDC 6600 performed precision orbit determination and inverted the data to map the Earth's subsurface density, treating the planet as a CAT or MRI scanner. Gravity modeling thus became a computational grand challenge alongside subsurface and weather modeling, strengthening the UT-Rice-Exxon ties that Mary Wheeler bridged.
Chapter 5: Mary Wheeler And ADI
Wheeler first met subsurface techniques in Young's UT numerical-analysis course. Dynamic PDEs model transient fluid flow in porous media, and the ADI method of Douglas, Rachford, and Peaceman at Humble Oil broke them into one-dimensional steps. At Rice her advisors were Rachford and Douglas, whom she called "the fathers of the field." She became the first tenured female professor in engineering at Rice, championed rigorous finite element methods, and later directed UT's Center for Subsurface Modeling. Her central contributions came in her 1971 dissertation: she proved that industry's cell-centered finite-difference schemes were simplified approximations of mixed finite element methods, and introduced the "elliptic projection" decoupling spatial from temporal error.
ADI And IBM CPC At Exxon Precursor Humble Oil
Exxon acquired its CPC in 1952 and used it heavily to create ADI, the Alternating Direction Implicit method of 1954, which decomposes multidimensional PDEs into sequences of one-dimensional problems; modern ADMM is essentially ADI. Card-manipulation tricks ran in parallel at RAND and in Houston, both descending from Feynman's colored-deck human-computer pipelines at Los Alamos in 1943. Humble Oil's hardware advanced from the CPC to the IBM 704 (about 1956), which freed the CPC for transfer to Matsen at UT in 1958, about when Young got SOR running on the 1103A.
Early History Of Reservoir Simulation
Before digital methods, reservoirs were studied with physical analogs in the 1930s and 1940s: water-coning models, electrolytic models of the East Texas Field, and resistor-capacitor networks in which voltage stood for pressure. Their cost, poor scaling, and inability to capture nonlinear multi-phase flow drove Rachford, Peaceman, and Douglas to digital FDM, and ADI (1954) bypassed the memory and stability bottlenecks of earlier schemes. FDM still struggled with complex geology, while early continuous FEM failed for fluid flow because it did not conserve mass locally — the gap Wheeler's work would close.
ADI And SOR
The two algorithms differed in both mathematics and culture. SOR came from academia — Young's 1950 Harvard dissertation — and was elegant, pure linear algebra; ADI was a gritty algorithmic hack born of commercial urgency, devised by Peaceman, Rachford, and Douglas in December 1953. Each found its home: ADI with IBM in oil and industry, SOR with UNIVAC and CDC in aerospace and the national labs. Their key contrast was tridiagonality: ADI splits a 2D problem into cheap tridiagonal 1D systems, whereas SOR grinds point-by-point. Each seeded a modern lineage — SOR the multigrid and preconditioned solvers, ADI the operator-splitting methods still used in climate modeling and CFD.
Subsurface Modeling
In the early 1970s FEM became a robust competitor to FDM, catalyzed by Rice's 1970 Department of Mathematical Sciences. Wheeler's 1973 "Pipeline in Permafrost" work on the Trans-Alaska Pipeline modeled the moving freeze-thaw boundary. The UT-CDC and Exxon-IBM divide was bookended by a shared IBM era in the 1950s and a shared Cray era in the 1980s, when both adopted Cray vector supercomputers — Exxon for larger 3D reservoir models, UT for the seismic imaging that helped find the Atlantis and Thunder Horse fields. Wheeler's later work turned to geologic carbon storage, injecting supercritical CO₂ into deep saline aquifers.
Chapter 6: Bob Schutz And UTOPIA
Schutz studied under Tapley from 1961 and became an expert user of UT's 60-bit systems for some twenty years. At NASA's Williamstown workshops, the 1969 report The Terrestrial Environment: Solid Earth and Ocean Physics anticipated GRACE and GOCE and recommended drag-free low-altitude satellites and satellite-to-satellite tracking. UT built UTOPIA, the University of Texas Orbit Processor, in 1974, and the Navy's Triad mission (1972) validated drag-free flight with the DISCOS system, a proof mass floated in a cavity while cold-gas thrusters chased it. Schutz and Tapley sharpened Transit orbit uploads with statistical filtering, driving the shift from Doppler to laser ranging and culminating in the Center for Space Research, founded in 1981 with Tapley as director.
Geodesy And Spherical Harmonics
Geodesy models the Earth's non-uniform gravity. Early Transit orbit errors reached about 2 km a day, while the 0.1-nautical-mile requirement demanded far better gravity models. The field rests on Laplace's equation and the spherical-harmonic expansion, which parallels FEM: both decompose complex reality into simpler parts, one through global basis functions and the other through a local mesh. The demands were steep — dense, globally distributed data, which TRANET's all-weather Doppler supplied; multiple orbital inclinations; and a ceiling above degree 100 set by atmospheric noise. UT was uniquely prepared for this by its 60-bit machine.
CDC 6600 And Cyber
Spherical-harmonic computation is expensive and unstable, involving thousands of correlated coefficients and ill-conditioned normal matrices; a single iteration could take 50 hours. In a historic marathon, Anderle's team solved 70 geopotential coefficients against 1,662 tracking datasets in about 50 hours on an IBM 7094, validating TRANET geodesy to 20 m. The models grew from WGS-72's roughly 500 terms to a modern degree and order 70, aided by Schutz's normalized Pines formulation free of polar singularities. The center's hardware advanced from the 6600 through the Cyber to the Cray X-MP (1985, the first Cray at a Texas university), which carried gravity models to degree and order 180.
McDonald Laser Ranging, Harlan Smith, And Pete Shelus
In the 1970s, combining satellite laser ranging with radar altimetry produced the first global high-resolution geoid maps, merging Baker-Nunn optical tracking and TRANET Doppler in the Goddard Earth Models. Satellite laser ranging, in use since 1964, improved from meters to millimeters; with it Schutz and Tapley made the first laser-ranging determination of polar motion and confirmed plate tectonics.
- Harlan Smith came to UT in 1963, chairing Astronomy for fifteen years and directing McDonald for twenty-six; he secured funding for the 107-inch telescope and served on the committee that helped launch Hubble.
- Pete Shelus was hired in 1970 to analyze lunar-laser photon returns and led the transition to the automated McDonald Laser Ranging Station on Mount Fowlkes; McDonald's Texas Minor Planet Program, running parallel to geodesy, calibrated Hubble's fine guidance sensors, and satellite laser ranging uniquely defines the geocenter and the Earth's mass.
UTOPIA
Data flowed from McDonald to Austin, processed by Schutz's code and embodied in UTOPIA, the University of Texas Orbit Processor. Its predecessor EARTHOD, in Fortran IV on the 6600, revealed localized short-arc Cartesian filters to be structurally inadequate. UTOPIA was the operational successor: unlimited global networks, batch least-squares with extended sequential filters including the square-root information filter, and multi-week arcs. It validated NASA's Geodyn and Germany's EPOS independently, as in the LAGEOS frame-dragging tests over eleven years. Schutz's fully-normalized spherical-harmonic formulation removed polar singularities and overflow, unlocking scalability from LAGEOS to SeaSat and TOPEX and eventually to 180×180 fields.
Chapter 7: Galaxy In The Machine
The late-1960s and early-1970s CDC 6600 era formed a distinctive shared community. Punch-card computing forced everyone, from aerospace researchers to undergraduates, into the same terminal rooms — an intellectual commons — with long batch turnaround where a single syntax error sent a job back to the start. A second unifying force was the computer game: CDC's own engineers stress-tested machines with diagnostic games on the 6600's CRT consoles, including Space Wars, a fast version of MIT's 1962 PDP-1 Spacewar. At UT this grew into a multi-generational family of Star Trek-inspired space simulations reaching from the 6600 to the early-1980s DEC-10, the thread that ties the section together.
National Science Foundation
The NSF funded the 6600 through its $1 million grant of 1966 and later helped introduce the DEC-10. Two initiatives broke the "Anastasio cycle" of powerful mainframes without good educational software: Project C-BE, launched in 1972, reached 75 professors and over 4,000 students, and Project CONDUIT tackled software transferability on a $450,000 NSF grant. The 6600 itself resisted timesharing — Cray answered a virtualization request with "No, I don't think so."
PDP-10, DEC-10, DEC-20
The acquisition of the HRC DEC-10 around 1975 was the decisive shift toward democratized, interactive computing. With native 7-bit ASCII, virtual memory, and high-concurrency timesharing, it decoupled interactive users from batch scientific loads and turned terminal rooms into proto-online social worlds — a lineage that culminated in the 18-player DECWAR. It subverted the "Big Iron" image and the batch-processing "priesthood" that Ted Nelson attacked in Computer Lib (1974). The PDP-10 sat at the heart of MIT and Stanford AI research and was rebranded the DECsystem-10 (1971) and DECSYSTEM-20 (1976, from BBN's TENEX), embodying a counterculture-meets-military paradox — MIT's ITS with no passwords, Stanford's bohemian SAIL — all ARPA-funded.
Chapter 8: New Frontiers
The Computation Center's frontier spirit was embodied by its director Charles Warlick, who held that "Texans do not undertake things in a small way." The Southwest Regional Educational Computer Network grew to 23 institutions, and in 1977 UT became the first ARPANET site in Texas. By 1973 the "Dual Dinosaurs" — the 6600 and 6400 sharing a filesystem, the 6400 running 128 timesharing ports for some 18,000 undergraduates — had made computing an instructional utility.
RAND, SAGE, Timesharing
Timesharing on UT's 6600 was a student-led initiative, chief among the students Forest Baskett. His inspiration came from a 1964 summer job at RAND's spinoff System Development Corporation (SDC), born of the SAGE project, where he used the AN/FSQ-32 (Q-32), a canceled SAGE prototype that ARPA repurposed for interactive-computing research in the spirit of "Man-Computer Symbiosis." Linked to MIT Lincoln Lab's TX-2 in 1965, the Q-32 achieved the first transcontinental data exchange between two operating systems, a direct ARPANET precursor. Frustrated by the 6600's batch OS, Baskett's team won Jim Browne's backing — "Hmm, that could be fun. Let's try" — and built the RESPOND timesharing service, which began in March 1967 and ran a decade.
Southwest Network
Timesharing and packet switching share a deep logic: both slice an expensive resource — CPU cycles or network bandwidth — to exploit the bursty nature of human-computer interaction. Davies designed packet switching in 1965 precisely because circuit-switched phone networks suited bursty traffic so poorly. By 1969 this lineage produced the Southwest Region Educational Computer Network, which grew from 9 to 23 institutions but exposed a content gap, pushing UT toward educational software. One highlight was SIMPLE, a conversational matrix language and direct ancestor of MATLAB; its creator Cleve Moler built MATLAB in the mid-1970s at the University of New Mexico.
ARPANET And Internet
The Q-32 foreshadowed the ARPANET's 1977 arrival. In the 1965 Q-32/TX-2 experiment, a user ran a Lisp program across the country, exposing circuit-switching's fatal flaws and convincing Larry Roberts to adopt packet switching; Bob Taylor, ARPA's IPTO director and a UT Austin graduate, had recruited Roberts. Wesley Clark's insight that "you've got the network inside out" led to the Interface Message Processor (IMP), the first network routers, with Telnet and FTP built on the Network Control Program. UT received its first IMP in 1977, but the HRC DEC-10 could not be the first host, since stock TOPS-10 did not support the IMP; instead a PDP-11/45 running hacked UNIX in the Daily Texan composing room became the UTEXAS host.
HRC And The DEC-10
The 1975 DEC-10 turned campus culture toward interactive exploration, sitting fittingly in the Harry Ransom Center near the Linguistics Research Center; its appeal was plain — compact, under $1 million, and reentrant compilers serving many users at once. It housed ICSCA — Don Good, Woody Bledsoe, and Jim Browne — doing program verification: their Gypsy Verification Environment and Certifiable Minicomputer Project mathematically verified ARPANET security code replicating BBN's Private Line Interface, an ancestor of IPsec and HAIPE. The DEC-10 was decommissioned on 31 October 1982, an occasion Dawson marked as "The Soul of an Old Machine."
Chapter 9: Pioneers
UT's 1970s computing belonged to a national hacker milieu — the Homebrew Computer Club, the Stanford and MIT AI Labs, Ray Tomlinson's 1971 email — sustained by ARPA's "conscious leniency" toward social lists and games. The Star Trek game had a rich lineage: Mike Mayfield's 1971 text version ran on a UCLA SDS Sigma 7, itself the very first ARPANET computer, and Dave Ahl's 1973 DEC BASIC port appeared in his 101 BASIC Computer Games, the first million-selling computer book. A separate UT branch ran from Hicks and Korp's 1973 BASIC variant on the 6600 through a Fortran rewrite, with "The Thing," to a two-player "WAR" around 1976 — the ancestry of the 1978 "ultimate version," DECWAR.
Teletypes And Cathode Ray Tubes
The campus geography of computing around 1970 to 1975 was scattered across obsolete buildings repurposed as terminal rooms, including Painter Hall (1972, shared with Physics and dubbed "Home of Star Trek"). San Antonio's Datapoint looms large: its Computer Terminal Corporation, founded in 1968 by Phil Ray (UT '57) and Gus Roche on Bob McClure's advice to build a "glass teletype," produced the Datapoint 3300 (1969), which replaced the loud ASR-33. The Datapoint 2200 (1970) was, in effect, the birth of the personal computer: its instruction set led to the Intel 8008 and thence to x86, leaving a little-endian legacy. The evolution ran on to the 1978 DEC VT100, though the everyday reality stayed constrained by 300- and 1200-baud dial-in.
Simulation And AI
Two worldviews stood in contrast. The classic assembly-and-Fortran symbiosis — inner loops in assembly, outer code in Fortran — defined the numerical-simulation era; DECWAR was among its last projects, while UTOPIA was nearly pure Fortran. Both were in effect large simulations: a satellite in a gravity field, or 18 people battling across a galaxy. By the 1970s a different AI spirit had taken root. The Certifiable Minicomputer Project (Don Good, 1975) built Gypsy, whose Message Flow Modulator became the first mechanically-proven software deployed in military security, passing all tests on the first attempt. The Linguistics Research Center, founded in 1961 by Winfred Lehmann, ran Lisp projects, including Siemens's revival of METAL, which Jonathan Slocum rebuilt in Lisp on the DEC-10 and delivered in 1984.
TECO-124, MACRO-10, FORTRAN-10, DECWAR
The climactic section first finishes the Lisp story: UT-LISP was a native Lisp 1.5 in CDC COMPASS assembly on the 6600, its three-pointer cons cell partitioning the 60-bit word into CAR, CDR, and CSR for semantic-network triples. DECWAR ran from BASIC (Hicks and Korp, 1973) through a Fortran rewrite to DECWAR on the DEC-10 (Hysick and Potter, guided by Clive Dawson), its name honoring the CDC WAR and MIT's 1962 Spacewar. The DEC-10's shared high-segment memory and inter-process locking let up to 18 players, each an independent job, share one "game board" in core memory, and several mainframe technologies converged in it. A custom TECO toolchain — TECO descending from Dan Murphy's 1962 PDP-1 editor and evolving into Emacs — auto-generated Fortran include files to align symbols with assembly memory offsets. MACRO-10 was a macro processor whose macros could override hardware instructions; Gates and Allen used it with the DDT debugger to write an 8080 emulator for Altair BASIC around 1975. FORTRAN-10 met assembly halfway, and pre-Fortran-77 bit-masking of Hollerith constants directly shaped DECWAR's design. The Romulan, evolved from "The Thing," had no job of its own; its behavior emerged from the player jobs, and it "spoke" through radio messages convincing enough that players took it for a real, malevolent person. Message handling needed MACRO-10 UUOs to lock shared memory, which bound the code to TOPS-10. DEC canceled the PDP-10 line in 1983, but DECWAR lives on through Obsolescence Guaranteed, a PiDP-10 replica running SIMH on a Raspberry Pi, and through Docker time capsules bundling SIMH, TOPS-10, and the reconstructed UT Source Distribution Tape.
Chapter 10: Conclusion
The mainframe-era history of modeling and simulation in Austin and Houston is worth preserving through both its code and its people, tracing the thread from the human computers of Los Alamos through Santa Monica and Houston to Austin by way of Simplex, ADI, SOR, subsurface modeling, and orbit determination. Because these developments were mathematical at their core, even the absence of original source code is no barrier: the machines and the exact methods are known, so the models can be faithfully reconstructed and run. The proposal is for Docker "time capsules" — self-contained, version-controlled environments bundling hardware emulation, operating system, tools, code, and documentation, with AI as an enabling partner — aspiring to be "a kind of Feynman Lectures of modeling." The proposed "heartbeat" health check is Feynman's Lectures on Physics Table 22-3, which produces continuous oscillation from pure arithmetic — a "sign of life" that fits in a few lines of assembly or Fortran.