SIMSCRIPT

SIMSCRIPT is a discrete-event simulation language, conceived in 1962 by Harry Markowitz and Bernard Hausner at the RAND Corporation, with Herbert Karr writing the manuals that taught people to use it. It is one of the oldest programming languages still in production use, and almost certainly the oldest one whose surviving production users are wargames.

Its central contribution is not a syntax but a way of describing the world. Before you write a line of executable SIMSCRIPT you write a preamble declaring what exists: entities (things), their attributes (properties), and sets (collections that entities own and belong to). The passage of time is handled by events - instantaneous occurrences, each with a routine that updates the state of the system - which the language’s built-in scheduler holds in a future events set and fires in timestamp order. Simulated time jumps from event to event rather than ticking. That triple of entity, attribute and set, together with an automatic timing routine, was the language’s whole argument, and it turned out to be durable enough that the September 2025 edition of the JTLS-GO Executive Overview still explains it to new users almost word for word as Markowitz did.

History and origins

Markowitz came to RAND in the 1950s having already written the paper on portfolio selection that would eventually win him a Nobel Memorial Prize. At RAND he worked on large-scale simulations - manufacturing, logistics, military operations - and ran into the problem that motivated the language. Simulations of even moderate complexity took months to program in FORTRAN, and the whole point of a simulation study is that you change the model repeatedly as the preliminary results come in. RAND’s own abstract for RM-3310 puts it in exactly those terms: SIMSCRIPT “reduces the time needed to program simulations of even moderate complexity and provides increased flexibility in modifying such models in accordance with the findings of preliminary analysis and other circumstances.”

The division of labour was clean. Markowitz supplied the world view - entities, attributes, sets, events. Hausner, a RAND programmer, built the implementation: a preprocessor that read SIMSCRIPT and emitted FORTRAN for the IBM 7090. Karr, a friend Markowitz brought in, wrote the manual. The result was published in 1962 as RAND Research Memorandum RM-3310-PR, with a companion technical appendix as RM-3813, and as a Prentice-Hall book the following year.

SIMSCRIPT I is also where the buddy memory allocation algorithm came from: Markowitz devised it to manage the language’s dynamically created and destroyed temporary entities, and it went on to become a textbook allocator in its own right.

Because the work was done under an Air Force contract, the language went into the public domain. In July 1962 Karr and Markowitz left RAND and founded California Analysis Center, Inc. in Santa Monica - a company whose first office, in CACI’s telling, was a phone booth and a park bench - to sell training and support for a language nobody owned. The name became Consolidated Analysis Centers, Inc. in 1967, and CACI outright in 1973; the company that started as a SIMSCRIPT consultancy is now a multibillion-dollar defense contractor. In 1965 CACI released SIMSCRIPT I.5, a real compiler for the IBM 7090/7094 rather than a FORTRAN preprocessor, which by CACI’s account made it one of the earliest independent companies to develop and sell proprietary software.

The Simula connection

The most consequential thing SIMSCRIPT did in its first decade happened by accident. In the summer of 1963 Bernard Hausner was employed by the Norwegian Computing Center, where Kristen Nygaard and Ole-Johan Dahl were working on Simula. Their 1978 history of the Simula languages records what came of it:

“Hausner was one of the fathers of SIMSCRIPT (Markowitz, Hausner, Karr 1963), and through him we got to know the SIMSCRIPT language and its implementation rather well. This was our first encounter with the pointer concept in a high level language. To some extent, however, our reaction to SIMSCRIPT can best be described as that of a ‘cultural clash’. It helped to make an important design goal conscious and explicit.”

The clash was over safety. SIMSCRIPT’s pointers were raw machine addresses in a language with no type discipline over them; Nygaard and Dahl decided their language had to offer the “security” of ALGOL 60, where an erroneous program is caught by the compiler or at run time rather than quietly corrupting memory. That decision - typed, checked references instead of untyped pointers - is one of the things that made Simula’s classes into the ancestor of object-oriented programming rather than a curiosity. SIMSCRIPT’s influence on Simula is thus partly by imitation and partly by reaction.

Elsewhere the design was adapted directly: F. M. Tonge and co-authors published QUIKSCRIPT, “A Simscript-like Language for the G-20”, in Communications of the ACM in June 1965.

SIMSCRIPT II and the level structure

The 1968 redesign, published by RAND as R-460-PR under the names of Philip J. Kiviat, Richard Villanueva and Markowitz, is the version that established SIMSCRIPT’s characteristic look. Gone were the punched definition forms of SIMSCRIPT I; in came free-form, deliberately English-like source text in which let X = 10 can be written X = 10, incrementing is add 1 to X, and an if block is terminated not by end if or fi but by the word always:

define J as an integer variable

read J

if 1 <= J <= dim.f(Y)
   write Y(J) as "The value is ", d(7,2), /
else ''invalid entry
   write as "The index is out of bounds!", /
always

Two details in that fragment are pure SIMSCRIPT. The chained comparison 1 <= J <= dim.f(Y) is an abbreviation for the conjunction; the reference grammar allows a relational operator to be repeated across a chain of expressions. And ''invalid entry is a comment: per the reference syntax, comments are enclosed in quotation marks “formed by two consecutive apostrophes”, and the closing pair may be omitted if the comment is the last item on the line. A variable name is any sequence of letters, digits and periods that does not form a number, and - remarkably - SIMSCRIPT II.5 has no reserved words at all: CACI’s manual notes that the keywords “are not reserved words and thus may be used as variable names”, while advising against it because it produces unreadable programs. Combined with the absence of an explicit statement delimiter, that is also why the compiler’s error recovery can cascade.

The other structural idea of SIMSCRIPT II was that the language was organised in ascending levels, so that it could be taught and adopted incrementally:

A programmer could use levels 1 to 3 and have an ordinary if idiosyncratic procedural language; the simulation content only arrived at levels 4 and 5. A version of SIMSCRIPT written in PL/I, based on the public-domain RAND release, was developed around 1968-1969 and is recorded in the Encyclopedia of Computer Science and Technology.

SIMSCRIPT II.5, introduced by CACI in 1971, is the version that mattered commercially and the one practitioners generally mean when they say “SIMSCRIPT”. Its main addition was the process interaction world view layered on top of the event world view: a process is a collection of related events separated in time, written as straight-line code that can work for a duration, request and relinquish resources, and be suspended and resumed by the scheduler. The classic textbook example, from Edward Russell’s CACI manual Building Simulation Models with SIMSCRIPT II.5, is an aircraft:

process AIRPLANE
     call TOWER giving GATE yielding RUNWAY
     work TAXI.TIME (GATE, RUNWAY) minutes
     request 1 RUNWAY
     work TAKEOFF.TIME (AIRPLANE) minutes
     relinquish 1 RUNWAY
end '' process AIRPLANE

Over the following decades CACI ported SIMSCRIPT II.5 from mainframes down through minicomputers to the PC, added the SIMGRAPHICS graphical subsystem (presented at the 1988 Winter Simulation Conference), a database interface, and an integrated development environment - and built packaged simulators on top of it, including SIMFACTORY II.5 for manufacturing, NETWORK II.5 for computer systems and COMNET II.5 for communications networks.

SIMSCRIPT III

By the early 2000s the problem was that SIMSCRIPT’s world view had been overtaken by the vocabulary it had helped create. Markowitz recalled telling Ana Marjanski, who headed the SIMSCRIPT III project at CACI, that SIMSCRIPT already had entities, attributes and sets - and being told that the clients wanted objects.

SIMSCRIPT III, presented at the 2004 and 2005 Winter Simulation Conferences by Stephen V. Rice, Ana Marjanski, Markowitz and Stephen M. Bailey, is a strict superset of SIMSCRIPT II.5 that adds:

• Classes and objects, with multiple inheritance, and methods including process methods - object activities that occupy simulated time
• Modularity: modules and packages that are imported by name, with public interfaces separated from private implementations
• Stronger typing: type checking across module boundaries and far more extensive compile-time error reporting, in place of II.5’s generic untyped pointer mode
• 64-bit support, giving the large address spaces that high-fidelity theater-level scenarios need

The authors were explicit that the object-oriented features “were influenced by C++, Eiffel, Java, MODSIM, and SIMULA” - which closes the loop rather neatly, given where Simula’s designers first saw a pointer. Implementation-wise, SIMSCRIPT III is a translator to C: the JTLS documentation refers to it plainly as the “SIMSCRIPT III (SIMSCRIPT to C) translator/compiler”, which is why CACI’s Windows editions listed a specific Microsoft Visual Studio version as a prerequisite.

CACI dated the release of SIMSCRIPT III to 2007 on its own history page. Its archived platform tables for the Release 4.0 era listed 32-bit Windows XP/Vista/7 with Visual Studio 2005/2008/2010, 32-bit Sun SPARC under Solaris 8 or later, 32-bit Linux on Red Hat Enterprise 4, 64-bit Linux on Red Hat Enterprise 4 through 5.5 for AMD64, and 64-bit Windows 7; by the Release 5 era the offering had narrowed to 64-bit Windows 7/10 with Visual Studio 2015-2017 and 64-bit Red Hat Enterprise Linux 7.4 or higher. The final publicly announced version, Release 5.1, added SimEclipse, an Eclipse-based development environment with on-the-fly syntax and semantic error highlighting, content assist, class-hierarchy and call-hierarchy views and integrated graphics editors, together with a business-rules interpreter and a checkpoint/restart facility aimed squarely at live training exercises.

Current status

SIMSCRIPT’s status is genuinely split, and it is worth being precise about it.

The vendor is gone from the web. CACI’s SIMSCRIPT product site was still serving product and download pages when its mirror at simscript.net was archived in April 2021; simscript.com itself was last archived serving pages in early 2020, resolves today to a parking address that does not answer, and simscript.net now redirects to a domain-parking lander. caciasl.com, the site of CACI’s Advanced Simulation Lab in San Diego, was erroring on 14 April 2021 and had been re-registered by an unrelated party as a Norwegian WordPress site by June 2021, running consumer-lending content from that August. The pages carried CACI copyright footers that were never brought past 2019, and some sections were still stamped “Copyright 1995-2011” as late as 2017. No release after 5.1 has been publicly announced, and there is no public issue tracker, package repository, mailing list or open-source implementation to look at instead. On the ordinary measures of language liveness, SIMSCRIPT is dormant.

The code is still running. JTLS-GO, the theater-level wargame used by U.S. and allied commands, is SIMSCRIPT: version 6.4.0.0 was delivered on 30 June 2025, its August 2025 Version Description Document lists the SIMSCRIPT III translator among the tools a government organisation needs in order to recompile from source, and version 7.0 was planned for August 2026. Its September 2025 Executive Overview still contains a section headed “Why SIMSCRIPT?”, giving two reasons for choosing an event-driven simulation: that “the key processes of theater-level, air-sea-land battle are most easily visualized as collections of discrete events”, and that “event-driven simulations are relatively fast which makes efficient use of computing resources” - a comparative claim the document makes without benchmark figures. That is not a legacy footnote; it is a design rationale being restated to new users sixty-three years after RM-3310.

This is an unusual failure mode for a language: the compiler is effectively unobtainable to the general public while the models written with it are actively maintained, because the population that needs the compiler - U.S. government organisations holding the source - is small, well-defined, and served by other channels.

Why it matters

Three things.

First, SIMSCRIPT invented the vocabulary that discrete-event simulation still uses. Entities with attributes, sets that entities own and belong to, event notices scheduled into a future events list, simulated time that advances event to event rather than by fixed increments - all of that is standard equipment now - in SimPy, in Arena, in AnyLogic, and in every hand-rolled event loop in a simulation textbook. It was not standard equipment in 1962.

Second, it is a load-bearing part of the prehistory of object-oriented programming. Simula’s designers learned about heap-allocated, dynamically created objects reached by pointers from Hausner and SIMSCRIPT, and defined their own language partly in opposition to what they found there. Markowitz’s account of being told that “the clients want objects” fifty years later is the same story from the other end.

Third, it is the best available evidence that a well-chosen domain abstraction outlives almost everything else about a language. SIMSCRIPT’s syntax is dated, its tooling is a decade behind, its vendor has vanished from the internet, and it has no community in the modern sense. What it has is a world view that fits the problem so exactly that the organisations who depend on it have found it cheaper to keep a 1971-vintage language alive than to re-express their models in anything newer.

Markowitz won a Nobel Memorial Prize for portfolio theory, but the John von Neumann Theory Prize he received the year before, in 1989, cited three bodies of work: portfolio theory, sparse matrix methods, and simulation language programming. SIMSCRIPT is the third one.