Alex Chassanoff is a CLIR/DLF Postdoctoral Fellow in the Program on Information Science and continues a series of posts on software curation.
As I described in my first post, an initial challenge at MIT Libraries was to align our research questions with the long-term collecting goals of the institution. As it happens, MIT Libraries had spent the last year working on a task force report to begin to formulate answers to just these sorts of questions. In short, the task force envisions MIT Libraries as a global platform for scholarly knowledge discovery, acquisition, and use. Such goals may at first appear lofty. However, the acquisition of knowledge through public access to resources has been a central organizing principle of libraries since their inception. In his opening statement at the first national conference of librarians in 1853, Charles Coffin Jewett proclaimed, “We meet to provide for the diffusion of a knowledge of good books and for enlarging the means of public access to them. 
Archivists and professionals working in special collections have long been focused on providing access to, and preservation of, local resources at their institutions. What is perhaps most unique about the past decade is the broadened institutional focus on locally-created content. This shift in perspective towards looking inwards is a trend noted by Lorcan Dempsey, who describes it thusly:
In the inside-out model, by contrast, the university, and the library, supports resources which may be unique to an institution, and the audience is both local and external. The institution’s unique intellectual products include archives and special collections, or newly generated research and learning materials (e-prints, research data, courseware, digital scholarly resources, etc.), or such things as expertise or researcher profiles. Often, the goal is to share these materials with potential users outside the institution. 
In our quest to understand how software may be used (and made useful) as an institutional asset, we engaged in a two-pronged approach. First, we aimed to identify the types of software that MIT may consider providing access to What are the different functions and purposes that software at MIT is created used, and reused for? Second, we aimed to understand more about the active practices of researchers creating, using, and/or reusing software. We anticipated that this combined approach might help us develop a robust understanding of existing practices and potential user needs. At the same time, we recognized that identifying and exposing potential pain points could potentially guide and inform future curation strategies. After an initial period of exploratory work, we identified representative software cases found in various pockets across the MIT campus.
Collection #1: The JCR Licklider Papers and the GRAPPLE software
Materials in the collection were first acquired by the Institute for Special Archives and Collections in 1996. Licklider was a psychologist and renowned computer scientist who came to MIT in 1950. He is widely hailed as an influential figure for his visionary ideas around personal computing and human-computer interaction.
In my exploration of archival materials, I looked specifically at boxes 13-18 in the collection, which contained documentation about GRAPPLE, a dynamic graphical programming system developed while Licklider was at the MIT Laboratory for Computer Science. According to the user manual, the focus of GRAPPLE on “the development of a graphical form of a language that already exists as a symbolic programming language.”  Programs could be written using computer-generated icons and then monitored by an interpreter.
Institute Archives and Special Collections, MIT Libraries, Cambridge, Massachusetts.
Materials in the collection related to GRAPPLE include:
Printouts of GRAPPLE source code
GRAPPLE program description
GRAPPLE interim user manual
GRAPPLE user manual
GRAPPLE final technical report
Undated and unidentified computer tapes
Assorted correspondence between Licklider and the Department of Defense
Each of the documents has multiple versions included in the collection, typically distinguished by date and filename (where visible). The printouts of GRAPPLE source code totaled around forty pages. The computer tapes have not yet been formatted for access.
While the software may be cumbersome to access on existing media, the materials in the collection contain substantial amounts of useful information about the function and nature of software in the early 1980s. Considering the documentation related to GRAPPLE in different social contexts helped to illuminate the value of the collection in relationship to the history of early personal computing.
Historians of programming languages would likely be interested in studying the evolution of the coding syntax contained in the collection. The GRAPPLE team used the now-defunct programming language MDL (which stands for “More Datatypes than Lisp”); the extensive documentation provides examples of MDL “in action” through printouts of code packages.
Figure 2. Computer file printout, “eraser.mud.1”, 31 May 1983, box 14, J.C.R. Licklider Papers, 1938-1995 (MC 499), Institute Archives and Special Collections, MIT Libraries, Cambridge, Massachusetts.
The challenges facing the GRAPPLE team at the time of coding and development would be be interesting to revisit today. One obstacle to successful implementation that the team notes were the existing limitations of graphical display environments. In their final technical report on the project from 1984, the GRAPPLE team note the potential of desktop icons for identifying objects and their representational qualities.
Our conclusion is that icons have very significant potential advantages over symbols but that a large investment in learning is required of each person who would try to exploit the advantages fully. As a practical matter, symbols that people already know are going to win out in the short term over icons that people have to learn in applications that require more than a few hundred identifiers. Eventually, new generations of users will come along and learn iconic languages instead of or in addition to symbolic languages, and the intrinsic advantages of icons as identifiers (including even dynamic or kinematic icons) will be exploited. 
Despite technological advancement, some fundamental dynamics in human-computer interaction remain relatively unchanged; namely, the powerful relationship between representational symbols and the production of knowledge/knowledge structures. What might it look like to bring to life today software that was conceived in the early days of personal computing? Such aspirations are certainly possible. Consider the journey of the Apollo 11 source code, which was transcribed from digitized code printouts and then put onto Github. One can even simulate the Apollo missions using a virtual Apollo Guidance Control (AGC).
Other collection materials also offer interesting documentation of early conceptions of personal computing while also providing clear evidence that computer scientists such as Licklider regarded abstraction as an essential part of successful computer design. A pamphlet entitled “User Friendliness–And All That”notes the “problem” of mediating between “immediate end users” and “professional computer people” to successfully aid in a “reductionist understanding of computers.”
Figure 3. Pamphlet, “User friendliness-And All That”, undated, box 16, J.C.R. Licklider Papers, 1938-1995 (MC 499), Institute Archives and Special Collections, MIT Libraries, Cambridge, Massachusetts.
These descriptions are useful for illuminating how software was conceived and designed to be a functional abstraction. Such revelations may be particularly relevant in the current climate – where debates over algorithmic decision making are rampant. As the new media scholar Wendy Chun asks, “What is software if not the very effort of making something intangible visible, while at the same rendering the visible (such as the machine) invisible?” 
Building capacity for collecting software as an institutional asset is difficult work. Expanding collecting strategies presents conceptual, social, and technical challenges that crystallize once scenarios for access and use are envisioned. For example, when is software considered an artifact ready to be “archived and made preservable”? What about research software developed and continually modified over the years in the course of ongoing departmental work? What about printouts of source code – is that software? How do code repositories like github fit into the picture? Should software only be considered as such its active state of execution? Interesting ontological questions surface when we consider the boundaries of software as a collection object.
Archivists and research libraries are poised to meet the challenges of collecting software. By exploring what makes software useful and meaningful in different contexts, we can more fully envision potential future access and use scenarios. Effectively characterizing software in its dual role as both artifact and active producer of artifacts remains an essential piece of understanding its complex value.
 “Opening Address of the President.” Norton’s Literary Register And Book Buyers Almanac, Volume 2. New York: Charles B. Norton, 1854.
 Dempsey, Lorcan. “Library Collections in the Life of the User: Two Directions.” LIBER Quarterly 26, no. 4 (2016): 338–359. doi:http://doi.org/10.18352/lq.10170.
 GRAPPLE Interim User Manual, 11 October 1981, box 14, J.C.R. Licklider Papers, 1938-1995 (MC 499), Institute Archives and Special Collections, MIT Libraries, Cambridge, Massachusetts.