Reference / Paper · 2017
Making IT Work — Proceedings (British Computer Society / TNMOC, May 2017)
Conference proceedings on keeping historic computers working: conservation versus operation, power-up, repair and documentation.
Reference / Paper · 2017
Conference proceedings on keeping historic computers working: conservation versus operation, power-up, repair and documentation.
Making IT Work 22-23 May 2017 Proceedings British Computer Society Southampton Street, London, UK The National Museum of Computing Bletchley Park, Milton Keynes, UK © 2017 British Computer Society and The National Museum of Computing Contents Martin Campbell-Kelly Introduction 5 Doron Swade The Historical Utility of Reconstruction and Restoration 7 Andrew Herbert, Chris Burton and David Hartley The EDSAC Replica Project 22 Christopher P. Burton Maintainability and Sustainability Issues in Restored and Replicated Computer Systems 35 John K. Chilvers Curatorial Lessons from Other Operational Preservationists 44 Robert Garner Restoring and Demonstrating 1960s Vintage Computers at the Computer History Museum 54 Robert Lusch, Nicholas Hekman, and Don Manning IBM System/360 Printer Revitalization at TechWorks! 68 Johannes Blobel and Jochen Viehoff The New Public Engagement in Computer Museums and in the History of Computing Machines 81 Benjamin J. Trethowan Capturing, Restoring, and Presenting the Independent Radar Investigation System (IRIS) 91 Peter F. Linington Constructing and Testing a Replica Store for the EDSAC 97 Nicholas Enticknap Report of the TNMoC Workshop, 23 May 2017 Proc. Making IT Work, 22-23 May 2017 107 3 Proc. Making IT Work, 22-23 May 2017 4 Introduction Martin Campbell-Kelly Editor The Making IT Work conference, 22-23 May 2017, was jointly organized by the Computer Conservation Society of the British Computer Society and The National Museum of Computing, the UK’s two leading institutions for computer conservation. The conference brought together the leaders of national and international computer conservation and restoration projects, including several recipients of the Google-sponsored Tony Sale Award for Computer Conservation. The conference enabled practitioners of computer conservation to share knowledge, and newcomers to learn about their motives and achievements. The conference took place over two days. The first day consisted of a formal lecture programme held at the headquarters of the British Computer Society, Southampton Street, London, while the second day consisted of a workshop showcasing projects at The National Museum of Computing at Bletchley Park. Following a welcome from the chair of the Computer Conservation Society, David Morriss, the first day opened with three of the UK’s pre-eminent computer conservationists addressing the issues of utility, management, and sustainability. Doron Swade, well known for the construction of Babbage’s Difference Engine Number 2, talked about the purpose and value of these activities — whether Conference speakers and attendees at The National Museum of Computing workshop Proc. Making IT Work, 22-23 May 2017 5 Introduction they be conservation, restoration, or replication. Andrew Herbert, formerly managing director of Microsoft Research in Europe, described issues of project management in building a replica of the EDSAC — easily the most ambitious recent project of its kind. Chris Burton, the doyen of British computer conservators, spoke from long experience about the chequered history of sustaining projects once they have been completed. The afternoon session began with a lecture from John Chilvers, a distinguished conservation architect and heritage consultant, who set computer conservation in the wider context of the preservation of cherished objects, from historic buildings to railway locomotives. The final three lectures of the day came from overseas speakers. Robert Garner described the restoration of a pair of IBM 1401 computers at the Computer History Museum, Mountain View, California — this project won the 2014 Sale Award. Next, Nicholas Hekman of the Vintage IBM Computing Center at TechWorks! in Binghamton, New York, described the restoration of an IBM 1403-N1 line printer originally supplied with IBM System/360 computers — the project was runner-up for the 2016 Sale Award. Lastly, Johannes Blobel and Jochen Viehoff described the building of an ENIAC accumulator for visitor interaction at the Heinz Nixdorf MuseumsForum — the winner of the 2016 Sale Award. For those able to attend, the day closed with a convivial dinner at Bella Pasta on the Strand. The morning of the workshop at The National Museum of Computing was taken up with a series of presentations from project leaders of five of the museum’s major restoration and replication projects. Following a welcome from Andrew Herbert, chair of the museum trustees, Delwyn Holroyd described the restoration of a 1980s-era ICL 2966, the museum’s largest exhibit and one of the most powerful British mainframes. Next, Ben Trethowan talked about the IRIS, until recently a key component of the UK’s air traffic control infrastructure. He was followed by Phil Hayes who described the rebuilding of the Colossus, the World War II electronic code-breaking computer. Next Peter Linington described the fabrication of a nickel delay-line memory for the EDSAC replica. Lastly, Kevin Murrell described the restoration of the Harwell Dekatron Computer, the world’s oldest functioning computer. (Trethowan and Linington contributed papers describing their work to this volume.) During the afternoon attendees were given demonstrations of these five projects. So that they could see systems close-up and ask questions, workshop participants were divided into three circulating groups of comfortable size. A detailed account of the workshop by the computer journalist Nicholas Enticknap appears as the last paper in this volume. Proc. Making IT Work, 22-23 May 2017 6 The Historical Utility of Reconstruction and Restoration Doron Swade Royal Holloway, University of London Abstract During the last several decades there has been an increasing number of initiatives to reconstruct historic computing devices and machines, and to restore original specimens to working order. These projects pose many questions. To what extent is it allowable to compromise the physical integrity of the original to achieve a working system? What status does a “restored original” have as an evidentiary source? In the case of reconstructions, to what category of object do these new-old machines belong — fictitious antiques, new primary sources, public monuments? How important is physical fidelity to historical authenticity? What is the social and historical value of such projects to their makers, to the scholarly community, and to visitors who view them? This paper uses as case studies major reconstruction and restoration projects from the modern era to address a raft of questions agitated by such projects — museological, historiographical and social. Introduction During the last several decades there has been an increasing number of initiatives to reconstruct historic computing devices and machines, and to restore original specimens to working order. The starting point here is to take such activity as a given and, without making assumptions about the original motivations or how they might have been justified at the time, seek to reverse engineer the implications and issues that relate to their historical, museological and social meanings. The process is based on a suite of projects drawn from the major flagship projects of the modern era. A word about “utility” in the title of this presentation. “utility” is not here a reference exclusively to “practical usefulness”. The concept is borrowed from economic theory where Utility is thought of as that property of a thing that makes it desirable to possess (Robinson, 1964: p. 48). So we might say that the utility of a knife is its ability to cut. One might say that the utility of art is its aesthetic appeal or its ability to affect one. Here utility is unrelated to practical usefulness. Entirely untethering utility from practical usefulness, we have the case of the architectural folly. We might say that a folly has cultural utility or socio-cultural utility as a demonstration of wealth or status, but it is not that a folly is practically useful that makes it desirable to possess. It is its very uselessness that confers meaning on the notion of its utility and that allows us to talk meaningfully of the utility of a folly, however paradoxical this might sound. So what is it about reconstructions or restorations that makes them desirable? What is their value to history, to their makers, to their host organisations where these exist, and to their public and private audiences? None of the individual projects considered here are treated in any detail. Rather each is used as a placeholder for a particular aspect of utility. Many of the individual features are not unique to specific projects but are shared by many. Case Studies The pool of machines and projects to be drawn from includes: Thomas Fowler’s Ternary Calculating Machine (c. 1840) Charles Babbage’s Calculating Engines: Difference Engine No. 2 (1847-9); Analytical Engine (1836-71) Konrad Zuse’s Z3 (1941) Proc. Making IT Work, 22-23 May 2017 7 Doron Swade Manchester “Baby” (1948) Colossus (1943) ENIAC (1946) Ferranti Pegasus (1959) Each is used here as a place-holder for particular features it exemplifies. Thomas Fowler’s Ternary Calculator Thomas Fowler was an impoverished self-tutored fell-maker who designed and built out of wood a digital three-state mechanical calculator demonstrated in the 1840s and witnessed by no less than Mr. Babbage and Augustus de Morgan amongst others. The poverty of sources here is extreme: not only did the physical original not survive but there are no plans or drawings. Fowler refused to release details of the invention, even for evaluation and endorsement by George Biddell Airy, the Royal Astronomer. Fowler was an outsider to the London scientific establishment and paranoid about his invention being stolen, and with good reason. He had earlier invented a thermosiphon for circulating hot water in heating systems and his unprotected invention had been ripped off by unscrupulous exploiters with no financial advantage to Fowler — something that aggrieved him till his death. The sources for the reconstruction by Mark Glusker (Figure 1) are partial contemporary (19th century) textual accounts, one by de Morgan, and a memorial stained glass window installed circa 1870 in a church in Devon. (Glusker et al., 2005). Reconstructing context, purpose and technical detail from such shards is as much archaeology as it is technical history. This speculative reconstruction is a placeholder for the role of sources in reconstruction, and issues of historical authenticity. Figure 1. Mark Glusker with the Fowler reconstruction (2005) Proc. Making IT Work, 22-23 May 2017 8 The Historical Utility of Reconstruction and Restoration Babbage’s Calculating Engines Babbage’s Difference Engine No. 2 (Figure 2), designed between 1847 and 1849, remained unbuilt until the modern era. No part of it was built before 1986 and the machine in its entirety (that is including the output apparatus) was completed in 2002 (Swade, 2005 and 2000). The sources are a set of 20 main drawings that are detailed enough to describe parts and their relationship to each other, but insufficiently detailed for manufacture. There is also almost no textual description of the design rationale or the logical function of the mechanisms. An understanding of the machine was essentially reverse engineered from the drawings, a process first embarked on by the late Allan Bromley. The built machine is routinely referred to as a replica or reconstruction. It is neither. There is no original machine of which this is a repeat. It is strictly a “construction” and, being the first physical realisation of the design; it is original in one of the definitional meanings of that word — that of being the first. Although 19th-century manufacturing techniques were not used, significant care was taken to ensure that the machine looked as though it had been made in the 19th century. Machining, rather than casting techniques, were used to produce the bright finish that was the hallmark of machining in Babbage’s time. Contemporary practice was used as a guideline for choice of materials, and composition analysis was made of bronze parts used by Babbage in the 1820s to establish appropriate the grades of bronze available at the time. We welded where Babbage would have forged, but the welds were dressed to appear as those produced by 19th-century processes. Tim Robinson in the U.S. has constructed in Meccano a difference engine based on Babbage’s original designs for Difference Engine No. 2 (Figure 3). He has also made some assemblies from the Analytical Engine designs (Robinson, 2008). It is clear that physical fidelity is not a sine qua non of meaningful realisation. What might we call such derivative constructions? Figure 2: Charles Babbage’s Difference Engine No. 2 (2002) Proc. Making IT Work, 22-23 May 2017 9 Doron Swade Figure 3. Tim Robinson’s Meccano Babbage Engines (2008). Top: Difference Engine No. 2. Insets: Analytical Engine. Tim Robinson Piers Plummer (2017) has constructed a difference engine based on Babbage’s Difference Engine No. 2. The calculating wheels are made by 3D printing using sintered nylon, false coloured to look like bronze and the construction includes laser-cut stainless steel, unavailable in the 19th century (Figure 4). The machine is driven by a small steam engine. It has fewer columns of differences and lower digit precision than shown in Babbage’s design and the machine is stretched vertically and flattened out front-to-back so that internal workings, obscured in the original, can be seen for pedagogic purposes. Again, what might we call such a construction? The Babbage Engine and its derivatives here serve first as placeholders for terminology — what might we properly call different categories of physical realisation; and second, as placeholders for issues of fidelity — visual fidelity, material fidelity, and fidelity of process. Proc. Making IT Work, 22-23 May 2017 10 The Historical Utility of Reconstruction and Restoration Figure 4. Piers Plummer Difference Engine (2017) Konrad Zuse’s Z3 (1941) The Z3 is a general-purpose programmable computing machine developed in Germany during War World II using electromechanical relay components. The original was destroyed in the wartime bombing of Berlin and a reconstruction was completed 1967 by Zuse himself. The reason for inclusion here is not to do with the 1960s rebuild, but with the reconstruction of the machine led by Raul Rojas (2005). The reconstruction looks nothing like the original (Figure 5). Components are laid out on large PCBs to convey the logical architecture of the design and it is operated interactively by the user. Relay activity is indicated by LEDs and data paths are illuminated to show signal routing. Rojas defines the overriding purpose of the physical reconstruction as pedagogic. In seeking to make the workings, operation and design of the machine intelligible to modern generations, he deliberately untethers the new machine from the traditional constraints of physical fidelity to the original. Rojas’s reconstruction explicitly challenges the assumption of visual and material fidelity being an indispensable feature of reconstruction and, in our suite of projects, is a placeholder for the place of fidelity in its various forms in the desiderata of reconstruction. Documentary records for the original Z3, as well as the 1960s rebuild, were incomplete and essential design and operational features not well understood. A remarkable feature of this project was the excavation, information recovery, capture and dissemination of technical and design information assisted by both Zuse himself and his son. The research, findings, history of the machine, Zuse’s papers, as well as simulations of its operations are captured and are publically accessible through the Konrad Zuse Internet Archive. Proc. Making IT Work, 22-23 May 2017 11 Doron Swade Figure 5. Reconstruction of Konrad Zuze’s Z3 (part). Front view of processor. Here physical fidelity is not regarded as essential to the purpose of its reconstruction. Also “reconstruction” here extends beyond the physical to include the reconstruction of lost data and knowledge of technical design and function. Manchester “Baby” (1948) The Small Scale-Electronic Machine (SSEM or the Baby) was the first computer to run an electronic internal stored program (June 1948). The Baby did not survive as a relic. It was a developmental machine that morphed through expansion into the Manchester University Mk. I, and was later cannibalised for spares. The machine was reconstructed by a team led by Chris Burton and the project offers us a master-class in systematic reconstruction with forensic attention to sources, and issues of authenticity (Burton, 2005). Researching the technicalities of the machine and its design was an example of unrepeatable timeliness in that surviving original engineers and practitioners played an invaluable role in authenticating technical and operational features. Recovered memory was a feature of the process. Engagement with the machine demonstrated the power of physical realisation to unlock forgotten and undocumented information in the memories of the original engineers and practitioners — technical, operational, developmental and narrative. The project serves here as a placeholder for uncompromising attention to fidelity in all its aspects. Figure 6. Manchester “Baby” (SSEM). Left: Chris Burton with reconstruction (1998). Right: original (c. 1948) Proc. Making IT Work, 22-23 May 2017 12 The Historical Utility of Reconstruction and Restoration Figure 7. Colossus, Bletchley Park. Left: Tony Sale with reconstruction (2008). Right: Original (1943) Colossus (1943) Colossus was code-breaking computer built at Bletchley Park during World War 2 using electronic vacuum-tube logic and punched paper tape. The reconstruction was led by the late Tony Sale (Sale, 2005) a now legendary driving force behind the saving of Bletchley Park and the founding of The National Museum of Computing. All but two of the several Colossus machines were deliberately destroyed after the war. None are thought to survive. Documentary sources were somewhere between incomplete and non-existent. A few original components survived in private hands. The operation of certain parts of the machine were known, or could be inferred, but with less than detailed technical documentation reconstruction posed new challenges to historical accuracy. A part, for example, was known to rotate. I remember Tony saying that without a specification for the original motor, a motor from the windscreen wiper of a car would do (this was possibly in the context of the Heath Robinson). In the case of the motor we have functional replication without attempt at physical fidelity. The thing does the same as the original but not by the same means. The project is here used as a place-holder for the notion of functional replication. ENIAC (1946) ENIAC, operational in November 1945, was a vast primarily vacuum tube computing machine built at the University of Pennsylvania to calculate artillery firing tables. If one was to select a single machine symbolising the start of the age of electronic computing it is surely this. It had some 18,000 vacuum tubes and its success established to the wider community the viability of the large-scale use of vacuum tubes in electronic computing systems. Various cabinets and panels from the original machine survive but none of the original hardware is operational. Documentary sources are fairly comprehensive. Various reconstructions have been attempted. One of these is ENIAC-on-a-Chip produced to commemorate, in 1996, ENIAC’s 50th anniversary. The inclusion of ENIAC here is as a placeholder for the reconstruction at Heinz-Nixdorf MuseumsForum (HNF) of an ENIAC accumulator by Jochen Viehof and Johannes Blobel. The HNF project won the 2016 Tony Sale award jointly with Robert Garner’s restoration of the IBM 1401 printer at the Computer History Museum in California. The HNF reconstruction allows members of the public to program an ENIAC accumulator in the same way as the original — i.e. using plug-in cables and switches. HNF has some original ENIAC hardware and this is displayed static and non-working alongside the reconstruction. The reconstruction itself is built from scratch and is driven by customised solid-state electronics and a microcontroller. Proc. Making IT Work, 22-23 May 2017 13 Doron Swade Figure 8. ENIAC. Left: Replica of ENIAC Accumulator, Heinz Nixdorf MuseumsForum (2016). Inset: ENIAC (1946) What is distinctive here is the explicit aim to create a visual facsimile of the physical hardware as part of the look and feel of operating the original. Extraordinary measures were taken to replicate the colour and texture of the panels and fittings. It is a reconstruction designed as a usable interactive visitor attraction. In this it shares features with the Zuse Z3 project. The HNF ENIAC project seeks to reconstruct the experience of operational practice through operational and visual fidelity of a very high order. Ferranti Pegasus (1959) The Ferranti Pegasus is a vacuum tube drum-storage machine that first came into service in 1956. Some forty were built between 1956 and 1962. This particular Pegasus dates from 1959 and is in the Science Museum’s collections. It was the flagship restoration project of the Computer Conservation Society when it was founded in 1989. It was successfully restored to working order and the restored machine ran successfully for over 25 years. It was retired a few years ago. Proc. Making IT Work, 22-23 May 2017 14 The Historical Utility of Reconstruction and Restoration Figure 10. Ferranti Pegasus. Left: Engineer’s Console. Right: Bit-level simulation by Chris Burton Its inclusion here is as a placeholder for computer simulations of historic machines. Computer simulations can be seen as a form of “logical replication”. In the case of Pegasus a bit-level simulation of the machine was created by Chris Burton. We can see from the screen images that little attempt is made to create a visual facsimile of the console controls beyond recognisable representations of console components and their layout. The functions that correspond to each console-control are scrupulously replicated. The issue here is not one of visual fidelity but of operational and behavioural exactness of the simulation. The virtual machine can be operated (programs loaded from paper tape, executed, and results punched or printed) by clicking on the controls depicted on screen and the responses are exact, even to the extent of the signal waveforms on the oscilloscope screens. The concern here is functional rather than visual realism. In simulations of this kind we find another example of the abandonment of physical realism in replication and reconstruction (Swade, 2003). Fidelity None of the original machines, apart from the Pegasus, was available to act as a physical datum on which to base a reconstruction. Some were deliberately destroyed, others accidently so. One was simply lost, and another did not exist in the first place. In the pool of examples there are reconstructions that look nothing like their originals (Z3, Meccano Babbage). There is at least one example of a machine that looks as exactly as possible like the original but works by other means (HNF ENIAC). There are restorations of original machines that are functionally correct but that do not use original or even contemporary hardware. Clearly, fidelity to the original has a wide interpretation in this spread of projects. If we take fidelity to mean “faithfulness to the original” we can identify four forms: Fidelity Functional Material Proc. Making IT Work, 22-23 May 2017 Visual Logical 15 Doron Swade In the case of functional identity the reconstruction does the same thing as the original but possibly by other means (Tony’s windscreen motor, modern electronics in the HNF ENIAC and Zuse’s Z3). Material fidelity applies where the reconstructed systems or subsystems are made of the same stuff as the original — original or contemporary components and materials (Manchester Baby). Visual fidelity entails facsimile appearance, as in the HNF ENIAC, which in this case also embraces tactile fidelity. Finally, a reconstruction is logically faithful if it implements the same rules as the original as in the case of the Ferranti Pegasus simulation (Swade, 2011). Some of the forms entail or presuppose others and projects are often faithful to the original in more than one respect. Terminology We use a variety of words to describe our agency in relation to artefacts. Some of these words are unproblematic. A fake is an object purporting to be an original but is not. We think we probably know what we mean by a simulation. What of “restoration”, “replication”, “reconstruction”, “reproduction”, “copy”, “model”, “rebuild” and so on? May we ask whether this vocabulary contains meaningful distinctions or whether there is a set of precise definitions that are internally consistent when describing these new-old artefacts? By restoration we convey the sense of returning an object to its original state. In the case of devices or machines, restoration may or may not include returning the object to working order. The livery, upholstery, and paintwork of a vintage car, for example, may be restored, but with the engine left non-operational. The essential point here is that restoration invariably presupposes the existence of the artefact or an identifiable part of it to begin with. Reconstruction Fake Restoration Model Realisation Reproduction Conservation Replication Simulation Copy Rebuild Facsimile Replicating a machine usually means creating a copy using an existing contemporary object as a reference datum. The etymology of the word (replicare — to fold back) implies a sense of “taking an impression”. When a replica-maker minimises the physical differences between the original and the copy, the replica is exact. But replicas can be deliberately inexact. Replicas can be scaled copies either larger or smaller than the original, and there are non-working replicas made without any intention of functional fidelity. The case of Guatelli’s half-scale non-working replicas of Babbage’s original surviving Difference Engine No. 1 is an example. Here again the process of replication usually implies the existence of an original as a source. So the use of “replica” as a descriptive can usefully be reserved for a copy for which there is an existing machine as a reference datum. We can then reserve “reconstruction” for the creation of a machine for which there is no surviving example. So the project described in the next chapter to realise the Cambridge EDSAC, of which there is no surviving original, would, by this suggested Proc. Making IT Work, 22-23 May 2017 16 The Historical Utility of Reconstruction and Restoration convention, be called a reconstruction rather than a replica. In cases where part of the original survives and is used, the recreated artefact is a hybrid of original and reconstruction (Nielsen, 2011). “Reproduction” usually refers to the act of making a copy, sometimes exact, but usually with the implication of imitation, especially in the context of furniture making. There is a pejorative implication that a reproduction is inferior in not being original. In the way we commonly use the word, the original of a reproduction may or may not exist. Simulation can be seen as “logical replication” i.e. a non-physical form of duplication in which all essential logical predicates are preserved. Where an original exists, simulation can be seen as logical replication; where it does not, as logical reconstruction. The use of the word “model” is an intriguing one. Of the many definitions and usages the one that is most relevant here is that of “a simplified representation for purposes of demonstration”. The essential feature here is that of a reduction of some kind — of scale (Guatelli’s half-scale replicas for example), capacity (Plummer’s Difference Engine), or reconfiguration (Plummer's DE; Robinson's Meccano realisations). The original Difference Engine No. 1, dating from 1832, was assembled for purposes of demonstration and, while full-scale, it is only one-seventh of the capacity of the whole machine as designed. So it could arguably be described as a model, as it possess at least two of the defining features if this attribution — demonstration and reduced capacity. There is a limit to how far one can pursue these distinctions and retain sanity. It is also the case that reality has an irksome tendency to defy exhaustive taxonomies. But thoughtful use of how we use these descriptors will help avoid historical misconception and also help us to be clear in our own minds about what aspect of originality we are being faithful to. Utility It is clear from our suite of projects that their utility is not coextensive with their value as historical evidence in traditional curatorial or conservationist terms. So how might we articulate other forms of utility exemplified by these projects? What makes them desirable to pursue? I would suggest that two categories are useful here: Historical utility and Social utility Historical Utility The following are aspects of the value to history of ventures exemplified by the suite of projects used as our case references. Draw one into a level of intimate detail with the machine that rarely occurs by other means Anyone who has been involved in reconstruction and restoration projects will immediately recognise what is meant. The attention to detail goes beyond general understanding of principles of operation. It is not even that the TTY is cabled to the main cabinet or even that it is a proprietary connector not a DB25 that terminates the cables. It goes down to the level of whether pin 5 at one end is connected to pin 7 at the other and what is the colour of the conductor. Involving oneself in the finest level of granular detail is unavoidable from the top all the way down — architecture, logical organisation, circuit and system functions, specific circuits, components, pin-outs and wiring. Contingent or unexpected findings not foreseen or foreseeable by analysis or theory There are countless instances of contingent findings in many of the projects. In the case of the Babbage Engine it was not clear ahead of the build how fast the Engine could be driven and what the limiting factors might be, nor was it practical to try to predict this by analysis. The limiting mechanism turned out to be an intermittent circular motion mechanism — an unexpected Proc. Making IT Work, 22-23 May 2017 17 Doron Swade finding. Theoretical prediction, even if attempted, would necessarily have been speculative and would not carry the conviction of an empirical finding. In another instance the function of two horns on the carry levers was not completely understood before building it and they remained a partial mystery. Adhering faithfully to the original 19th drawings the mechanism was built as drawn and we were dumbfounded at the subtlety of its operation and the ingenuity of its function in ensuring the integrity of calculation (Swade, 2005). Insights into contemporary ideas and practice The Difference Engine No. 2 construction completed in 2002 was the first realisation of a complete Babbage calculating engine. No one in the 19th century had the practical benefit of a complete Babbage engine, so there is no operational data on how practical it would have been or whether it was a going proposition at all. With a finished working difference engine to hand we are better placed to explore operational issues: how many people would it take to run; how long can a single operator drive the machine by hand before tiring; how many operators need to be on hand to act as relays for continuous operation; how many attendants are required to mix plaster to replace the stereotype trays in the output apparatus; whether it is possible to phase the mixing sequence so that plaster drying times ensure an uninterrupted supply of trays with exactly the right consistency of plaster; how reliable is the machine; is there any economic case for the machine in the event that it requires more than n attendants to run it, and so on. We can interrogate these objects about key features of contemporary practice in ways that would not otherwise extend beyond speculation, and we can measure contemporary expectations against actual performance. This is experimental history in a very pure form. Tacit Knowledge — we can learn what it might have been like To witness Pegasus clacking out square roots on a teletype conveys what it might have been like in ways not communicable by other means. If one cranks the handle of Babbage’s Difference Engine there are shock loads at several points in the cycle as locks go in and out. The feel of this is not communicable by anything other than the experience of doing it. Nor is the sound the machine makes fully describable, particularly the rhythmic slapping sound of the locks which is the aural signature of the machine working. The carry mechanism was well-understood but what was not foreseen was the dramatic spectacle of the wave-like rotation of the double helices at the rear of the engine. Sight, sound, and feel. A cocktail of experience that text cannot convey. Documentary Completeness — new material, recovered memory, e-archaeology Researching a machine in detail almost invariably involves gathering all known information in one place. Co-locating and collating known material has substantial archival value. There are instances of new material emerging attracted by the public visibility of the project. The current EDSAC reconstruction is a case in point. One of the EDSAC 2 engineering team had a cache of drawings he removed from a cupboard when the original EDSAC was scrapped. Hearing about the EDSAC project reminded him of their existence and the retrieved them. There are other examples of this — the Babbage family sent priceless material from the family archive when they read about the Babbage Engine construction in the international press. There are instances of recovering lost knowledge from pioneers or practitioners for whom re-engaging with the machines unlocked memories, information and operational practice they had forgotten they knew. The driver circuits for the Williams Tubes used in the Manchester Baby is one of many in which contemporary practitioners supplied essential information that had not been documented. There are instances of recreating lost knowledge. The hard disk on the Elliott 401 is an instance. There was no data or metadata about the disk’s contents. Tony Sale and Chris Burton designed and made the instrumentation to read the bit-stream from the disk, capture this on a Proc. Making IT Work, 22-23 May 2017 18 The Historical Utility of Reconstruction and Restoration PC, analyse and recover the content and structure. The data was intact after 50 years. 35 mm magnetic tapes for the Elliott 803, stored in cans in a garage whence they were recovered, were found to be readable 40 years later. Both these findings have implications for digital curation, especially at a time when the impermanence of magnetic media and its longevity remain pressing issues. These two examples are as much part of electronic archaeology as they are computer history (Link, 2016, pp. 79-112). In the case of restorations there is nothing like working from a circuit diagram to expose discrepancies between the object and its technical description. It is not uncommon for the documentation of the last hardware version of a machine or system not to have been fully updated. There are instances of modifications made that are not recorded, especially in the final developmental stages, as well as local variants of machines that do not conform to the generic descriptions. Restoration to working order and subsequent maintenance will more than often involve a forensic examination of technical description and an assessment of their correspondence with the physical machine. Physical Completeness When large and medium-sized computers are acquired they are almost invariably uncabled and dismantled into manageable units for transport. The process of acquisition — the taking of a machine into protective custody is, paradoxically, sometimes the most traumatically destructive process in the life of this supposedly prized machine. Physical reassembly is the first stage of restoration, which in itself creates a physical datum of greater historical probity than leaving the machine in store in pieces. Getting the machine operational tests and verifies electrical completeness. Ultimately, when switched off, one has an exact knowledge of the machine’s electrical and physical state of completeness as a direct outcome of restoration activity. Benchmark for Simulation It is likely, and probably inevitable, that ultimately, on archaeological time scales, none of the restored or reconstructed machines will be working at some distant or not so distant future time. Bit-level simulations of these historic systems and migrating these from platform to platform hold the prospect of a kind of logical and operational immortality for computing machines. Restoring a machine to working order allows the simulation to be bench-tested against the live original, especially in relation to execution speeds. So “flaring” a machine into working order for even a relatively short time offers the benefit of verifying the conformity of a simulation to the behaviour of a physical original or a restored or reconstructed physical machine. Social Utility There is another dimension to utility: the value to those who make up the project teams, to visitors who view and experience the outcomes, and to museums that host the machines and sometimes the activities. Meaningful Work Reconstruction and restoration projects provide social and organisational context for veteran experts to share their expertise, exercise their craft, and extend their professional activities in historically and educationally meaningful ways. Pedagogic Value For participants in such reconstructions and restorations, as well as for the users and visitors, such projects are the source of both explicit and tacit knowledge as exemplified by the Z3 and HNF ENIAC reconstructions. Proc. Making IT Work, 22-23 May 2017 19 Doron Swade Nucleus — focus of visitor attention Reconstructions and restored machines on display act as a focus for visitor attention and a nucleus for story-telling. Babbage, Pegasus, Colossus, Bombe, EDSAC are large machines, sculptural spectacles in their own right that monumentalise and memorialise episodes in history of which they are an expression. Babbage’s Difference Engine No. 2 was built in public view. A commonly asked question was “how do you make this or that?” This was unexpected. The public build prompted conversations about manufacturing, for example, as well as about what the machine was and how it worked. Cultural objects Large systems, publicly displayed, memorialise important and often extraordinary episodes and practices for which they act as both placeholders and monuments. They serve as generational bridges to an otherwise lost past. Conclusion In articulating the historical and social utility of restoration and reconstruction projects, it is clear that they are desirable to different constituencies for a variety of different reasons. They are of value to museums, curators, archivists and historians concerned to preserve physical artefacts, documentation, narrative histories and operational knowhow. To visitors and users who experience them, they are formidable public attractions that memorialise major events and developments and serve