Abstract – The aim of this collaboration was to create a backup or safety copy of the media artwork Three Windows - Hommage à Robert Lax by Nicolas Humbert and Werner Penzel. This involved creating a disc image of the two SCSI (Small Computer System Interface) hard drives installed in a Celsius Siemens 400 computer. To complement the in-house skills, the media conservation department of the Kunsthaus Zürich called upon the expertise of Tom Ensom and Robin François to benefit from their experience in disk imaging, legacy computing and Linux applications. After testing a combination of approaches and tools to image the drives, we used the Aaru Remote tool to remotely image the computer over a network connection. After backing up the hard drives, the original setup was tested and the computer was found to be functional. For the first time in 20 years this original hardware could be documented and preparations made for ensuring the longevity of the artwork. Through this interdisciplinary collaboration, we have demonstrated the use of remote disk imaging, an approach that will have wider applications in digital preservation — particularly where physical access to storage media is difficult or obsolete interfaces are present.
Keywords – disk imaging, time-based media art, open-source, legacy hardware, collaboration.
This paper was submitted for the iPRES2024 conference on March 17, 2024 and reviewed by Dr. Panagiotis Papageorgiou and 3 anonymous reviewers. The paper was accepted with reviewer suggestions on May 6, 2024 by co-chairs Heather Moulaison-Sandy (University of Missouri), Jean-Yves Le Meur (CERN) and Julie M. Birkholz (Ghent University & KBR) on behalf of the iPRES2024 Program Committee.
Three Windows - Hommage à Robert Lax, 1999/2000, is a three-channel video installation, or, as described by the artists Nicolas Humbert and Werner Penzel, a film triptych (see Fig. 1). The 45-minute endless loop is a meditative exploration of the life and poetry of the American poet Robert Lax during his time on the Greek island of Patmos. It is soundtracked with minimal poems read by the poet himself. As described by Guido Magnaguagno: "The simultaneous ‘three window’ views tell of the simultaneity of day and night, of inside and outside, of island and space, shepherd and monk, cosmos and word. [...], the flow of images draws the viewer into the spell of a completely different, archetypal world of longing for fulfillment, wisdom and happiness that lies dormant in everyone." [1].
Three Windows - Hommage à Robert Lax was exhibited at the Kunsthaus Zürich from 26 November 1999 to 6 February 2000 and acquired by the museum as an Edition of 1 out of 5 shortly thereafter. In a letter sent to the curator Tobia Bezzola in 2001, the artists mention three forms of presentation. Depending on the size of the exhibition, the artwork can be either projected on white 16:9 projection screens, or presented on three flat-screen monitors. Alternatively, in a larger space, there is also the option to present the work as a rear-projection on parchment paper. The artwork was sold with a Siemens-Celsius workstation computer with keyboard and mouse, including special software, as dedicated playback equipment (see Fig. 2).
The computer is not visible in the exhibition setup and is described by the artists as a plug and play device: "[...] the computer simply needs to be switched on with the power switch at the back of the device and then put into operation with the start button on the front - the computer then boots up the program automatically and starts playing the program after about 30 seconds." [2]. The digitized film/video channels are stored as MPEG-2 files on the computer hard drives and video is output by three built-in video cards. The stereo sound is output by the middle video card. There is external genlock hardware used for synchronizing the three image streams. Additionally, three Digital Betacam tapes were delivered in 2005 as archival master tapes.
Given that the work relies on sensitive and outdated hardware, it was urgently necessary to deal with its preservation. The artwork has not been put into operation since it was last exhibited 23 years ago. There was no backup of the computer made, which contains the only version of the video content beyond the archival tapes. So far, these two versions of the media have not been compared with each other. Furthermore, there was no documentation of the behavior of the computer as a "plug and play" playback device. For the future presentation of the work, understanding the role of the computer as both a playback device and a storage device was essential. However, due to equipment obsolescence, there was a certain risk in starting up the computer without first extracting the files from the hard drives. For this reason, the decision was made to disk image the computer before proceeding to testing it.
Disk imaging is the process of using suitable hardware and software to create a disk image from a physical storage medium (e.g. a hard disk drive or CD-ROM). A disk image file encapsulates the contents of the physical storage medium from which it was created and can be used in its place. Disk imaging is used in the preservation of digital media as a means of duplicating data while ensuring its accuracy, authenticity and integrity [3]. In time-based media conservation it is used as a strategy for preserving, examining and exhibiting software-based artworks and other digital objects [4] [5].
Disk imaging of the storage media inside the Three Windows computer was identified as a priority. This would create a bit-level backup of the storage media contained, which could then be archived and preserved independently of the physical hardware. By combining imaging with the digital forensics techniques of write-blocking and checksum validation, we aimed to ensure that the disk images would be safely and accurately acquired from source media. An image could then be used as a basis for further examination and analysis work, including extraction and assessment of the video content and playback system.
In preparation for disk imaging, we determined the tools which would be needed for this process by identifying the interfaces used by the two internal drives. Initial examination of the internal components of the Siemens Celsius 400 PC found that it contained two 3.5" hard disk drives, with High Density 68-pin connectors, manufactured by IBM and Fujistsu respectively in 1999 (see Fig. 3).
From here, we needed to identify a means of connecting these SCSI drives, ideally with write-blocking active, to a modern computer for imaging. This first involved us trying to better understand the unfamiliar SCSI standards so that we could determine how we might connect them to a modern PC for imaging.
Small Computer System Interface (SCSI) is a set of standards for connecting computers with storage and peripheral devices. First standardized in the 1980s [6], it has a long history involving many versions and a variety of physical connector types. While there are SCSI-based standards still in use today, it is the first version of the standard which is used by the two hard drives we were aiming to image: Parallel SCSI. This standard has numerous variants, providing differing data throughputs, configuration options and associated connection protocols.
Parallel SCSI was widely used as an interface for attaching internal hard disk drives to desktop computer motherboards, primarily in professional and server systems rather than home computers. However, its use declined after it was superseded by Serial Attached SCSI (SAS) in the early 2000s and modern equipment is difficult to source. We are not aware of any published literature on the disk imaging of Parallel SCSI hard disk drives in contemporary computing environments, nor any modern SCSI write-blocking and adaptor hardware, factors which impede efforts to apply disk imaging effectively. We would therefore have to take an experimental approach to imaging. After identifying the presence of Parallel SCSI hard disk drives inside the Three Windows computer we carried out some initial research into connection methodologies to better understand how imaging might proceed.
From online resources, [7] [8] we were able to gain some insight into Parallel SCSI standards and how individual devices are connected and configured. SCSI devices are arranged in chains, consisting of multiple devices attached to a host on a single ribbon cable. Devices in a chain are each given an identifier number. Electrical termination of the cabling is required and used to indicate the end of the chain. Termination can be built into a cable, added to a cable or integrated into a SCSI device. SCSI devices (including adapters) can use three different electrical interfaces: Single-Ended (SE), High Voltage Differential (HVD), or Low Voltage Differential (LVD). These interfaces are not compatible with each other and all devices within a chain must support the same interface. Some devices support switching between SE and LVD. Using this information, we determined an initial set of equipment to use for imaging (detailed in Section 2.1).
The imaging approaches we used can be characterized as falling under two distinct categories, for which we have developed new terminology of ex-situ and in-situ. Ex-situ methods are those which involve detaching the target hard disk drive from the source computer and connecting to a host computer for imaging using suitable write-blocking and/or host adapter hardware. An example of an ex-situ approach would be opening the case of the source computer, removing the hard disk drive and attaching it to a modern computer via a hardware write-blocker for imaging with software tools.
In-situ methods are those which involve leaving the hard disk drive attached to the original host computer and accessing them for imaging via the original hardware. An example of in-situ imaging would be booting the source computer using an operating system stored on a bootable USB stick and using software tools to image the target hard drive via that operating system.
We decided that we would try using ex-situ methods first, as these presented a lower risk of altering the contents of the drives. This is because write-blocking tools can be used to prevent write operations reaching the drive and, by not booting the original machine, there is no risk of startup processes altering the drive contents. In-situ methods provided an alternative option when ex-situ methods failed. In-situ methods are useful because they do not require direct access to the physical drive or any use of an adapter to connect the drive to contemporary computer interfaces. However, they present a higher risk of modification of drive contents during booting.
To apply this approach, we removed the two 3.5” SCSI hard disk drives from the source computer and attempted to image them from an HP Z2 workstation computer running Windows 10 (and Linux Ubuntu 18.04 via VirtualBox). With our limited knowledge of the SCSI standards, and little information available online to fill in the gaps, we struggled to understand and apply the technical requirements for accessing a SCSI drive on a modern computer system. While we were able to get the drive to spin-up, and so eliminate the possibility of mechanical failure, the drive was ultimately not accessible through the operating systems we tested (Ubuntu 18.04 and Windows 10). After reaching the limits of what we could achieve using ex-situ methods with the time available, we decided to explore in-situ options (see Section 4). A brief account of the ex-situ approach taken and tools used are summarized below, with the aim of supporting future investigation into SCSI imaging (see Fig. 4 and Table 1).
SCSI write-blockers are no longer manufactured, and we were not able to source second-hand hardware to achieve this. Instead, we experimented with a range of SCSI equipment and configurations to make a SCSI chain and achieve a connection between the SCSI hard drives and the HP Z2 workstation from which we would image them (see Fig. 5). The equipment items, their purpose and limitations encountered are summarized in Table 1.
Equipment specifications | Purpose within chain | Limitations |
---|---|---|
SCSI 50 pin (High Density connector type, female) to SCSI 68 pin (High Density connector type, male) adapter, labeled “Differential” | Adapter between the SCSI 68 pin drives and the SCSI 50 pin to USB adapter. | Unknown if using the HVD or LVD electrical interface. |
Castlewood Orb USB Smart Cable (product number: 88205-001) providing SCSI 50 pin (High Density connector type, male) to USB (type A connector) interface, lent by the Cinémathèque suisse via Robin François | Connecting the SCSI drives to the workstation computer. | Was “plug and play” in Linux but not in Windows, where the most recent driver version is for Windows XP. Unknown if using SE, HVD or LVD electrical interface. |
Tableau Forensic USB 3.0 Bridge T8u write-blocker | Protecting the contents of the original drives from accidental write operations. | Unknown impact on SCSI communication with host (e.g. may have blocked access). |
LSI Logic LSI20320IE PCI-Express Single Channel Ultra320 SCSI Controller Card with SCSI 68-pin (1x Ultra-High Density and 1x High Density connector types, female) | Connecting the SCSI drives to the workstation computer. | Not possible to include a hardware write-blocker in the chain when using it. Most recent Windows drivers available are for Windows Vista. |
SCSI 68 pin (High Density connector type, male) SCSI ribbon cable | Forming the SCSI device chain. | None. |
SCSI 68 Ultra 320M Terminator with the specifications “LVD+SE ACT NEG + HVD ISO” | Terminating the SCSI device chain. | None. |
230V mains to 5/12V 4 pin Molex power supply | Power supply for hard drives. | None. |
Table 1. SCSI equipment used during our experiments with an ex-situ SCSI imaging approach with description of purpose and limitations encountered.
The source of these problems may have been incompatibility between components in our SCSI chain. It is not clear from their labeling whether the Castlewood Orb adapter and SCSI 50-pin to 68-pin adapter conform to HVD or LVD. Ultimately, we concluded that either elements of the equipment chain were incompatible or configured incorrectly. Without time for further experimentation, we moved on to exploring an in-situ approach.
In order to image the hard drive using an in-situ approach, we would return the two drives to the original computer case, reconnect them and use an appropriate software tool to image them. We would need to carefully choose this software, as it would need to run on either the original obsolete operating system Windows NT 4.0 or an alternative operating system that could be run on the original computer hardware, while providing an accurate and verifiable image.
Institutions and digital archivists usually rely on a variety of imaging software that were designed for either backups, digital forensics or even piracy, such as ddrescue, Guymager or IsoBuster. Communities focusing on the preservation of computer and video game history have been very fruitful in testing existing tools and developing new tools and methodologies to improve media imaging in their own context, rich with media diversity and copy protection mechanisms. From such communities have emerged new tools, such as the Aaru Data Preservation Suite (Aaru-DPS).
Aaru Data Preservation Suite [9] development started in 2014. It was initially named DiskImageChef (DIC) and mainly focused on extracting files from outdated file systems from a variety of disk image formats. Functionality has been progressively added to allow imaging of media to existing image formats. Aaru’s own open-source image format was also introduced to store all the data and metadata that could be captured during the imaging process of a media. Aaru is also capable of exporting this metadata to a structured XML/JSON file. Aaru’s main appeal for digital preservation is the consistent metadata across data carriers or filesystems and providing a unified workflow within one tool.
The Aaru-DPS also includes Aaru Remote, a “slim miniature application designed to receive device commands from a remote Aaru instance, sends it to a local device, and returns the data to the instance” [10]. Aaru Remote allows, through a network, to remotely image from a machine that cannot run the full Aaru software. Typically, older machines that cannot run modern operating systems cannot run Aaru, but could run Aaru Remote and have network connectivity that would allow remote imaging.
In the context of this project, Aaru Remote was selected as an alternative in-situ imaging method, as we could not run a modern operating system using Aaru. The idea was to execute Aaru Remote on the artwork computer, connected to a network. A modern machine running Aaru on the same network would then connect to the Aaru Remote instance to send commands and receive the data from the imaging process. To run Aaru Remote, we had to run an alternative operating system, since using the existing operating system was out of the question, because we did not want to boot from the original drives before imaging them and risk altering the data contained.
Booting the machine without the original SCSI drives, while running an alternative operating system, proved to be rather challenging. To find out how to get into the BIOS, we searched for the data sheet of the motherboard. Thanks to a forum entry from 2008 and the Internet Archive, it was possible to retrieve the correct data sheet for this Siemens W26361 D1107 motherboard. Pressing down the F2 key at boot gave us access to the BIOS, we could change the boot setup configuration to have CD as the first boot option. Booting from the CD drive was the only option, and booting from a USB or a floppy disk was not possible.
Unfortunately, the internal SCSI CD-ROM drive was not connected to the power supply or to the motherboard. We tested connecting the CD-ROM drive to the motherboard with the original SCSI cables used for the hard drives, however, this did not work. To remedy this, we connected a 40XMAX CD-ROM drive removed from another machine (dated March 2000) to the motherboard with an IDE connector. Using the IDE CD-ROM drive would allow us to boot the machine from a Linux LiveCD. A LiveCD is a bootable operating system stored on and booted from removable optical media (e.g. a CD-ROM).
We decided to use the Knoppix operating system, as it is known to be one of the more robust and hardware compatible Linux distributions. Initially, we booted the machine with Knoppix v3.1 (released in 2002), finding that hardware was properly detected and network access could be configured. However, compiling Aaru Remote for Knoppix v.3.1 proved to be too challenging as necessary packages could not be downloaded since package servers for Knoppix v3.1 have been decommissioned. Knoppix v9.1 (released in 2021) was then tested, but the modern kernel and drivers were not compatible with the network card. Additionally, the whole system was very slow due to the age of the hardware and therefore not practical to use.
Therefore, we had to find a version of Knoppix that would be compatible and usable on the machine, but for which we could also compile a functional version of Aaru Remote. We picked Knoppix 7.2 (released in 2013) and we were pleasantly surprised that this version was still compatible with the hardware. Compiling Aaru Remote after booting the machine on Knoppix proved very challenging due to missing dependencies. Instead, we changed our approach and decided to prepare a LiveCD with a built-in working Aaru Remote.
Using a QEMU virtual machine, we followed the “Knoppix Remastering Howto” [11] documentation and prepared a LiveCD with built-in Aaru Remote. The LiveCD has been shared with the Aaru project team and is available on Archive.org [12]. With this custom LiveCD, we successfully executed Aaru Remote on the machine and managed to get another machine running Aaru to connect to it through the network (see Fig. 6). Finally, we were able to start imaging the hard drives in good conditions and with appropriate software.
Booting from the Knoppix 7.2 CD-ROM was successful, and it was possible to connect the original machine to the local network, from which we could then run “aaruremote” from the command line (see Fig. 7). We then launched the imaging process through the Aaru software on the Windows 10 OS of the HP Z2 station. First, the two drives were identified as /dev/sda (IBM drive) and /dev/sdb (Fujitsu drive) using the “aaru dev list” command. Then, the IBM drive was successfully imaged using the “aaru media dump” command, before the Fujitsu drive was imaged using the same command.
Aaru does CRC (Cyclic Redundancy Check) checks during imaging, ensuring there are no modifications of the data during the writing process. To further verify the integrity of the process, both drives were imaged again, in order to make sure that there were no accidental reading errors during the imaging process. The two series of images were then compared to each other using the “aaru image compare” command and no difference was found.
Although it is likely to be added as a new feature in future releases of the software, it is currently not possible to extract data from NTFS, HFS and HFS+ file systems directly with Aaru. To increase compatibility, the Aaru Format (.aaruf extension) disk images were converted to the more widely supported RAW format (.img extension). With the FTK Imager software it was possible to mount the two disk images as physical & logical drives, with the mount method “File system / Read Only”, and extract the file contents.
One of the drives turned out to be additionally partitioned. On the first drive (IBM) there are two partitions: 2047 MB partition with the Windows NT operating system and 6698 MB partition with files. On the second drive (Fujitsu) there is one 8699 MB partition with files. The disk images were secured in the Kunsthaus Zurich digital repository and the MPEG video files were compared to the 10-bit uncompressed video files created from the Digital Betacam tapes by the Atelier für Videokonservierung Bern.
When the computer was inspected to assess internal hard drive interfaces, other internal components were also identified and documented. As part of this process, all peripheral devices, cables and video technology belonging to the work were photographed. From the outside, the ports of the three built-in graphics cards are already visible. Each graphics card was labeled with a letter: R, M and L, presumably right, center, left (in German: rechts, mitte, links) (see Fig. 8). Further details about the manufacturer and model could not be determined, even after opening the case. The Black Burst / Color Bar Generator SG-6003B from Kramer is used to synchronize the 3 video channels.
In order to document the intended functionality and behavior of the artwork, the original machine was started with the original operating system, once the imaging process had been successful. The testing setup involved the original Celsius Siemens tower, including the two hard drives that were previously imaged, as well as the original mouse and keyboard. To monitor the output, three different screens were connected with the computer, none of them original equipment. The LCD computer monitor was connected directly with the tower, while the 4:3 Sony CRT SD and the 4:3 Wallimex LCD HD control monitors were both connected to one of the video cards via BNC cable each. Additionally, a Neumann speaker, also not original, was connected to the middle video card via a dedicated DIN to XLR cable.
Due to the age of the BIOS battery, it was replaced as a preventive measure. Booting the computer was a smooth operation, despite the error message regarding the internal clock of the computer due to replacement of the BIOS battery. Once the computer interface was available, we observed that software named Super MDC Network Active 3.0.216 launched automatically (see Fig. 9). As soon as the operating system booted, the HD control monitor showed a pink image, confirming that there was some kind of signal transmitted, however, no moving image was visible. Rebooting the computer after resetting the internal clock to the current time and date, unfortunately, this did not improve the signal output.
A closer study of the Super MDC software made it obvious that while the software synchronizes and loops the video content, the genlock also plays an essential role within the setup. Indeed, once the genlock was connected to the video cards and the computer was rebooted, the video signal was finally displayed on the two control monitors. The playback of the video content is smooth and trouble-free, even when the computer is busy with the shutdown process.
One interesting thing to note is that the software changes the start position of the video files, in order to skip the color bars and countdown sections that are present in the first minute of the videos. Furthermore, the files found on the drives are all in a 4:3 ratio and are also being played back by the video cards in 4:3. However, the projection screens that were purchased with the artwork are in 16:9 ratio and the only reference recording we have for this work also shows use of a 16:9 presentation. The video files are anamorphic material; stored at a 4:3 aspect ratio and then scaled back to 16:9 on the projectors or monitors displaying the artwork. The content of the Digital Betacam tapes was found to be identical to the MPEG-2 files and anamorphic as well.
The Digital Betacam cassettes showed a slightly better resolution and fewer compression artifacts than the MPEG-2 videos. The time code of the MPEG-2 videos is set one full frame later. The video copies show the same technical opening credits with color bars and countdown. Only the audio track assignment is different.
The dominant disk imaging workflows in digital preservation and time-based media conservation use ex-situ methods. Ex-situ methods involve the removal of a storage medium from its host computer and the use of hardware write-blockers to ensure data integrity. There are situations where removal of a target drive is not possible, or a suitable interface for connecting to the drive cannot be sourced, and applying an ex-situ approach may not be practical. We have demonstrated the use of in-situ methods that circumvent the need to physically remove drives. Instead, an in-situ approach uses software to interact with the original hardware environment.
More specifically, we have demonstrated how the Aaru and Aaru Remote software can be used to image storage devices where it is not possible to access them via modern hardware interfaces (e.g. hardware write-blockers). As digital preservationists dive into older systems, they require modern tools and functionalities to perform their preservation tasks up to current standards. Aaru Remote is a tool bridging older systems with the requirements of modern media imaging, enabling the in-situ approach. The LiveCD created for this project (Knoppix 7.2 with Aaru Remote included) has been made available [12] for reuse to facilitate the replication of our methodology in similar scenarios. This approach may be particularly relevant for small form factor computers where drives can be difficult to access or remove (e.g. Apple Mac Mini) and proprietary storage media found in commercial video game consoles (e.g. the UMD disc format for PSP).
There are some limitations to this approach. It presents a higher level of risk of accidental damage to the target drive if not applied with care. Ex-situ approaches have been adapted from the digital forensics field, where stringent methods for ensuring data integrity are used during imaging so that material could be used as evidence during criminal investigation. Many of the imaging tools used in a digital preservation context today carry this legacy, not least of all in write-blocking hardware which is produced by companies supplying the digital forensics industry. In-situ imaging exposes the target drive to more risk through the lack of hardware write-blocking and the need to power on the original hardware. Care must be taken to ensure that the computer hardware is electrically sound before commencing in-situ imaging and that authenticity is verified during imaging (e.g. using checksums). Inclusion of reliable write-blocking software with the Aaru Remote LiveCD would be a worthwhile goal for future work.
Using Aaru Remote means that imaging is carried out using the full Aaru software, which has advantages over other imaging tools in a digital preservation context. This includes extensive metadata created during imaging, built in media analysis tools and file extraction from images. The NTFS file system is not yet supported by Aaru, and we had to rely on additional software for file extraction. We understand that the Aaru project would benefit from developers and attention to get additional functionalities implemented.
The in-situ approach trades the hardware challenges of the ex-situ approach with software challenges. Even though Aaru Remote has explicitly been designed to run on older systems, we have seen that the compilation of Aaru Remote proved to be difficult. We were fortunate to find a version of Knoppix that would allow the compilation and that would still run on the original hardware of the art installation. Thanks to the cumulative work of the open source communities that work on retro computing and operating systems of older environments, we were able to get a compiling and executing environment for Aaru Remote.
Aaru Remote dependency to CMake for the compilation process makes it less portable. Additionally, repositories for older versions of Knoppix were not accessible anymore which meant that we would have to manually find and install this dependency. Access to the source code for a software does not translate to an easy compilation due to dependencies. These challenges, limited in our case, are much more problematic in the context of rebuilding more complex software, such as video games. The preservation of the compilation/building process is then as important as the preservation of the source code.
There are also practical limitations to the wider application of this approach, which requires relatively specialized knowledge of boot configuration, bootable media creation and Linux operating systems. Not all those carrying out disk imaging work will hold this knowledge, which may present a barrier to the adoption of in-situ methods. We have demonstrated that collaborative partnerships (in this case between collection care staff, external specialists and open-source tool developers) can support this kind of work and provide opportunities to overcome challenges associated with unfamiliar media. Despite great progress in the development of imaging workflows in the digital preservation and time-based media conservation fields, there is still much to learn, and exciting opportunities to support the development of tools which serve our needs better than those we have adopted from digital forensics.
The preservation approach explored in this article turned out to be very successful for the case study Three Windows - Hommage à Robert Lax by Nicolas Humbert and Werner Penzel, despite the technical challenges and limitations of the project. With innovative solutions and the help of an interdisciplinary team, it was possible to disk image the old hard drives and extract the original files. We used the Aaru Remote tool, running on a LiveCD in the original hardware environment, to image the drives from another computer over a network connection. We adapted workflows and tools to develop a novel approach to disk imaging in the conservation of time-based media art. This approach not only provided a backup of the work, but was also the first time in 20 years that the hardware could be tested for functionality. This was an essential step for the accessibility, but also the longevity of this artwork. The approach developed will have applications in disk imaging work for digital preservation, particularly where physical access to media is difficult or obsolete interfaces are present.
At this stage, the exhibition of the work is technically possible, although the preferred form of presentation should still be clarified in dialogue with the artists. An interview with the artists is scheduled to take place in Spring 2024. This will be an opportunity to further deepen our understanding of the artwork. As the display format and playback devices play an essential role in the experience of the artwork, it is clear that the preservation of the artwork goes beyond the extraction of the video files. It is fundamental to understand the artistic intention and production context in order to grasp the full impact of the original installation and make decisions about future presentations of the artwork.
This case study was a particularly successful collaboration between the Media Conservation department of the Kunsthaus Zürich (Eléonore Bernard and Tony Kranz, Switzerland), the Software Platform Department of the Cinémathèque suisse (Robin François, Switzerland) as well as freelance digital conservator specialized in the conservation of software-based art (Tom Ensom, UK). The sharing of resources and the combination of expertise made it possible to successfully secure the historical computer in a very short timeframe of less than a week. When it comes to the preservation of audiovisual media and modern art, the diversity of technologies requires a diversity of expertise and the implementation of a local and international network of professionals and peers.
The preservation work as well as the research regarding this case study was supported by Kunsthaus Zürich, Memoriav (Association for the Preservation of the Audiovisual Heritage of Switzerland) and Cinémathèque suisse.
We would like to thank the artists Nicolas Humbert and Werner Penzel for kindly answering our questions. Additionally, we would like to thank the Aaru Data Preservation Suite open-source project, and more specifically Natalia Portillo, lead architect and developer, for developing and open-sourcing Aaru and Aaru Remote.
Thank you also to those who reviewed earlier drafts of this paper: Brian Castriota, Chris King, Aurore Lüscher, Kerstin Mürer, Luca Rey, Rebecca Rochat, Magalie Vetter.
Work by Robin François was partially supported by docuteam SA.