In this case study, Gil gives a short description of the design philosophies behind Ed, a system for producing online digital editions. These design philosophies focus on the concept of minimal computing, which includes a holistic analysis of overall system costs in creating and, as importantly, maintaining online resources. The minimal computing approach analyzes these overall costs in the context of historical and current global inequalities in access to resources, including technologies, and suggests a way forward that increases local control while decreasing long-term maintenance costs.
Alex Gil, Digital Scholarship Librarian, Columbia University Libraries
The inevitable false god of diversity
Diversity is pure Public Relations1. It has less to do with justice, equality and change than it does with glossy pamphlets. We fold it in three and throw it in the trash. Diversity is also mostly an anglophone, North Atlantic, intra-institutional affair. The world is already a multiplicity, and white folks are a minority in it. Most Professionals of Color working in predominantly white institutions in the United States know this, and still, we play. Some of us have families, most of us have debts to pay. Participating in this diluted, reduced language, knowing that there will be very little actual understanding of the more honest and rich language that motivates our participation, is par for the course. And yet, all is for naught if we do not build into our work careful mechanisms of rebellion—small axes compared to the rockets of those who secure dominance rather than persuasion—but decent documentation of our struggle and hope.
The Design for Diversity of non-white folks is not the pamphlet. Its consciousness goes beyond the janus-being of our institutional selves, adept in the orthopedic white voices needed to survive or thrive in close quarters. As a result, most non-white folk—at least those who actively work for equal integration—are aware that building in relation to an imagined Global South is not the same as a D&I initiative meant to bring American institutions in line with the demographics of an aspirational pluralist nation. At the same time, these two goals—to arrive at a solidarious relation to the Global South, and to redress inequality in access to resources and control of agency in integrated institutions—are not mutually exclusive either. One can think both together, and more.
In many years of listening to marginalized peers listing obstacles to participation in the popular genres of digital humanities, several common themes resurfaced: lack of resources, steep learning curves, and limited time. Sometimes these obstacles became fertile ground for timely critiques, but the obstacles remained. Many of my PoC peers from the USA and colleagues from the tri-continental area (Latin America, Africa and Asia) straddled the doors to two insider clubs—one was a social projection, a construct, the other had a different, less-forgiving reality grounded in the lack of familiarity with certain fundamentals of computing.
From my point of view, as someone who was lucky enough to learn the fundamentals of computing from humanists at the University of Virginia, I could see how the difference between using off-the-shelf tools out of necessity or choice was a central part of the problem. Most students in the United States—regardless of identity—are being introduced to digital humanities through a mere handful of tools: Omeka, TEI, Voyant, Scalar, Carto, etc. These “easy” tools have something in common: students learn very little about computation, systems, networks, and by extension the social implications of the modes of production that they enable. Even more importantly, use of these tools hides the vectors of control, governance and ownership over our cultural artifacts.
The design or project I will be talking about today, Ed, is the first prototype I built under the banner of a burgeoning conversation on minimal computing. I built it because of what we were hearing in the United States and around the world about access to and control of production, especially for peer, non-white aspiring producers, and because I had a rudimentary grasp of some computing fundamentals. It is an example of the radial thinking that seeks equality in an integrative national context, while remaining solidarious in an international one, and more.
What is Ed?
Ed is first and foremost a framework for producing digital editions of literary or historical texts. A digital edition is as the name implies, a reproduction or original production of a document or series of documents for reading on an electronic device—a laptop, phone, tablet, etc.2 Ed is said to produce “minimal” editions because it reduces the final product to a few main functions: a clean reading text, scholarly footnotes, searching, and public annotation. To learn more about what it is, and how it works, you can visit the Ed sample site
Many other frameworks for producing digital editions are available to editors and technologists today. You may have heard of editions “using TEI.” TEI, or the Text Encoding Initiative, is an international standard for describing texts that is the baseline for many modern digital scholarly editions, using various frameworks that take these descriptions and convert them into readable text. In many cases, texts which can be interacted with extensively, and even through non-textual derivatives like visualizations. You probably have also already encountered ePub file types and Kindle devices, favorites of many commercial presses, and further mainstays of the ecosystem of digital editions. These are digital editions too. No matter how or where you read, the production of these editions is beholden to a framework of some sort, and that framework has consequences at scale that go beyond the edition itself. Ed intervenes in these collaterals through both the process of production it asks for and the final product itself.
Ed is understood in another sense, more mundanely, as a distinct Jekyll theme. Jekyll is a static website generator. A static site differs from a dynamic one in the way the documents are produced, not in whether they are interactive or not on the front-end, a common misconception. A static site is produced long before the reader asks for one of its documents on the browser. A dynamic site is produced on the fly, at the moment the reader asks for a document on the browser. In most cases, a static site implies the absence of a database. This has enormous implications for the production of literary documents that have hopes of becoming part of a record. Having your documents recreated every single time you need them based on an algorithm is probably not a good bet for the sake of longevity, not to mention the cost of using electricity to re-create the document every time, when once or twice probably suffices.
Though still the most urgent one, producing reading texts is by far not the only purpose of scholarly digital editions. Second to this I find is the addition of commentary. Ed makes it relatively easy to do so. I also added search for editions with many separate documents. That’s it, though. Many other frameworks can do this and much more. How Ed goes about it, though, is where the subtle differences make all the difference.
In my childhood, mechanical sound-system equalizers required you to move one knob at a time. If you moved the bass, you would not affect the midtones or the treble unless you moved them too. Sometime in the 1990s we started using digital equalizers. When you changed a range in these ones, the other knobs adapted in tandem to keep the curve smooth depending on the setting: concert hall, rock, etc. I tend to think about design in general as participating in the dynamics of mechanical and digital equalizers. Instead of sound waves, our design choices move knobs controlling cost.
When you design a feature in a digital project to decrease a cost, another may increase or it may not depending on the context and the choice itself. For example, to reduce the cost of data curation, you may inadvertently increase the cost of systems administration; some choices in data curation, though, can actually leave sysadmin cost the same or even decrease it.
Considering as many costs as possible in the production of digital artifacts, a minimal computing design attempts to bring the overall curve of cost down while remaining true to the question “what do you really need?” The number of possible costs is substantial, and we should weigh the most important individually before we can understand how to assess the overall cost. While each of these factors is a “knob” that can be adjusted up or down, it affects the other knobs in tandem.
The most relevant consideration for questions of equality and agency. A high cost is equal here to a complete loss of control over all aspects of a project: ownership of the data, design of the presentation layer, decision-making on the content, etc. Minimal computing, and Ed as an example of it, is a corrective to historical inequities in ownership of sources and means of production. Control, naturally, comes to the fore in the burgeoning international collaborations over cultural heritage, in collaborations between first nation communities and modern universities, the fate of an African American past, and other similar situations. Needless to say, its ramifications go beyond the sensitive work of cross-cultural encounters and the need to reconfigure the conditions of our material pasts.
Ed allows you to retain multiple copies of your project in plain text in your computer, hard drives, USB sticks, etc. Because of the nature of static site generation, which Ed relies on, you have an exact copy of the project as it appears on the Internet, without the need for access to the cloud to read your texts.
That takes care of the material control over your work and data. The rest is up to you.
All scholars belong to a chain of production of specific material artifacts (books, ebooks, articles, digital editions, maps, etc) and therefore are implicated in that molecular rearrangement of the planet we call today the Anthropocene—a geological era defined by our mostly toxic impact on the planet. Although our scholarly use of PC or server processing time is small compared to the blockchain rush or the heavy use of social media, we can certainly lead by example where we can.
Ed requires three tools that used properly do not tax your own system while producing an edition: a terminal, a plain text editor and a browser. Because Ed eliminates the need for a database or AJAX, and your website is static, there is almost no processing time on the servers you use either. In fact, an Ed edition could be served from storage — a medium like a USB drive or compact disc that stores files, requiring minimal computing power to view. This connects back to the issue of control of course, and other issues listed below, because you can, for all intents and purposes, create editions that can circulate in USB sneakernets, evading surveillance and obviating constant need for the Internet.
3. Bandwidth and data
While many countries in the North Atlantic world now have broad access to unlimited data plans, 4G networks and high-speed internet, many within that North Atlantic world and broadly across the tri-continental area—Africa, Asia and Latin America—remain beholden to limited data plans and slower networks. When we are speaking of producing scholarship from, for or with these areas, bandwidth and data transfers determine whether a project gets used or not. The more bandwidth and data cost, the less likely the project will reach or attract our peers.
Ed is extremely lightweight, and does not depend on constant exchanges with the server. This makes Ed a good choice for sharing editions of texts when considering access as a material reality tied to bandwidth and data.
4. The Internet
Related to bandwidth and data, but not exactly the same, limited internet access can be a major issue in many scholarly communities around the world, but especially in the tri-continental area and the marginalized or oppressed North. In some cases we are dealing with no access at all. This is the case of prisons, for example. In other cases, the internet is simply not the best answer. Take for example environments with heavy government censorship.
Because a whole edition using Ed can be downloaded once and made functional thereafter this increases its usefulness in limited internet scenarios. Because an Ed edition can be made functional in a USB or other storage media, this allows us to still be able to share our scholarly work in no-internet or censored environments.
5. The maintainers
That digital projects generate indirect costs should not come as any surprise to anyone with years of experience in our domain. The rise of Content Management Systems and other database-driven solutions led to specific relationships between Libraries and Schools. In many institutions faculty and students wanted to contribute to a digital scholarly record by using these CMS tools because of the ease of data entry, and the lulling comforts of a Graphic User Interface. Libraries began hiring folks like me or reshaping job descriptions to help meet that demand, and LIS schools re-shaped programs around the CMS to respond. This phenomenon led to the infamous “library server problem”: not many libraries, including mine, could provide across-the-board, flexible server solutions to host an unlimited amount of projects created with databases, which demand a continuous regeneration of documents — it creates exponentially rising costs for maintenance and preservation. Ed and other systems that serve flat HTML are far easier to maintain.
The reason for “the problem” is the often unseen indirect cost of database-driven infrastructure on the technical support systems and personnel within libraries. Database systems require maintenance and security in a way that flat HTML files do not. Database systems require more energy (more processing, more memory, more servers, more experienced computer administrators to keep them running) than flat HTML files. Despite not always being able to recoup costs, libraries have tended to prioritize grant-funded projects which themselves tend to rely on large-scale database-driven platforms. First, this creates intra-institutional inequality of access, considering that most senior or influential faculty (more likely to be awarded funding) skew disproportionately white according to the national data. Second, this creates situations where libraries are the long-term stewards of a database-driven system where indirect costs are typically vastly underestimated.
TEI and Ed projects both can be produced without a database. Ed is different only in that the system of transduction (or transformation)—Jekyll in this case—remains immutable. In other words, you don’t create multiple transformation scenarios for different editions, as you would when transforming TEI for the web browser. What we lose in the added functionalities of TEI-based editions, we gain in even less cost to the maintainers, and those in charge of transformation scenarios. The absence of a database also implies that Ed editions are far more secure, close to attack-proof even, than any edition that requires a database.
6. The readers
An often unconsidered cost is the cost of poor design on readers. An elaborate digital scholarly edition, rife with functionality (comparison modes, image/text interaction, multiple relations between n-grams, etc), often fail at the most simple and common need: readability.
Using insights from book history and design, Ed was designed on the front end to make texts pleasant to read on any device. Using web fonts available on most devices on the planet, readers will have similar experiences across devices. It prints nice too. If emulating the clean contours of a page in a well-designed book is minimalism, then Ed is guilty of it. Otherwise, Ed would like to disconnect ties to the modernist trends often referred to as minimalism.
Though some basic accessibility features have already been implemented, the next round of development will address national guidelines on accessibility and community recommendations for those who use screen readers.
7. The learners
One clear cost of using Ed compared to a CMS is the learning curve. Comparing to TEI is more difficult as you will soon see. Ed producers who are not familiar with the command line may have a difficult time installing and producing an edition in the beginning if they have to depend solely on the documentation. Although I tried to make that documentation as accessible as possible, in my professional experience, the obstacles to divest from the metaphors of graphical user interfaces — to begin to understand the machine as a symbolic device — are substantial. Symbolic device here points to the fact that most of what you see on your screen, or hear on your speakers, is determined behind the scenes by the manipulation of discrete symbols, encoded at a low level in bytes—what some vaguely call code, and others computation.
If on the other hand, I have the opportunity to teach it, this cost is converted into a gain for a very simple reason: Ed allows me to teach a complete workflow from conception to production and maintenance of a digital edition that includes but is not limited to key fundamentals of computing: plain text, the terminal, markup, data structures, networks, versioning, preservation, web design and much more. In other words, as opposed to the average introduction to Digital Humanities course or a TEI-centered learning environment, Ed allows me to teach a wider range of fundamental and proto-universal literacies in one semester—enough to enable students to be independent producers of mature digital editions they control, and a solid foundation for other digital humanities work. Outside the classroom, I’ve been able to teach the necessary fundamentals in four sessions of 2 hours each, with roughly 16 hours of homework.
If you would like to see an example of a digital edition produced by undergraduates during the course of a single semester, please visit Mini Lazarillo. It bears mentioning that we also had time to hold seminar readings and discussion, several modules on textual scholarship, and advanced Spanish instruction for non-native speakers.
8. The editors
The technical work of Editors can be seen as a cost or not, depending on whether we are comparing to a CMS edition or to a TEI-based edition. The technical requirement for Ed can be reduced for editors to a knowledge of markdown. This depends on whether the editor herself is the producer of the edition. If the editor is being assisted by someone else who is producing the website, then the editor only needs to worry about learning markdown. This is of course something that takes about two or three hours to learn, and with a cheat sheet on hand, not much afterthought.
Because Ed focuses on bibliographic markup, as opposed to semantic markup, Ed makes all commentary overt in its endnotes. Layout and font are driven, later in the process, by design, rather than meaning.
Ed is free to use, modify and share. It can be published freely using your university’s UNIX servers or GitHub—to name a few options. If you pay for hosting, you really need nothing more than the most basic server. While there may be some financial cost in paying a skilled person to teach the process of creating an Ed edition, its financial maintenance costs elsewhere in the equalizer lower greatly.
Jekyll, the engine behind Ed, generates static HTML pages from a series of files, mostly written in Markdown and YAML, some of the simplest forms of data structuring standards available. A complete edition consists of nothing but these plain text files. The innovation of Ed, if the word is appropriate, is to look at the past, rather than the future, to make its preservation gambit. We find that simple HTML pages and plain text have historically survived longer than other solutions. Instead of guessing a futurity that eats its own neck, Ed places the bet on the stable component.
Ed‘s generation scenario allows producers to keep multiple exact copies of the edition, in all its functionality, in many storage devices. This increases the chances of an Ed edition surviving even the death of the internet.
The memory of the text is what’s at stake, though. No edition should be the last. Because of the simplicity of the markup and the ease of access to the data, an Ed edition can become part of that continuous chain that we expect of editions, that the reverberation of the past may continue to transform long into the future.
All together now
In creating an Ed edition, each of these costs — each knob in the equalizer — could have raised the cost of the other if attention was not paid to the whole. It mostly didn’t, and where it did, it displaced it to those who should bear the cost, or was reabsorbed into a pedagogical opportunity. The lesson, I hope, is that a consideration of cost from the point of view of the minority, diaspora, or marginalized group which can escape the pull of the center will tend to be, by sheer centrifugal escape velocity, more capacious in scope. The benefit for the very communities from whom the design arises is access and control to the production of literary editions. Though Ed is fully functional and reliable, the point of it is not to become a software solution for universal problems, but to serve as a prototype for thinking deeper about the asymmetric and complex contexts we operate in. I am leery of software solutions in that they gather moss and scope creep, and re-institute the division of labor between maintainers, architects, and pseudo-producers working on GUIs.
You will not see many uses of Ed. What I did not address here, and Ed can only address indirectly, is the societal reward structures for literary editions. To produce these documents one needs a certain amount of privilege (tenure, time, etc) or enormous sweat equity. Predominantly white institutions force us to do this work outside reward structures, and folks need to pay bills. If you attend the Society for Textual Scholarship, for example, you will notice the demographics are concerning. What in the 1990s was a surge of editorial work on the web from PoC, subsided because of these hard pressures that escape the design of Ed. Though Ed exposes metadata for aggregators, that will never be enough. In that sense Ed is also a speculative exercise that imagines a future where we can return to the careful curation of our literary past, for those who are invested in decolonizing the written word and the record. Useless to those who will burn down the episteme completely, of course. Ed is also just the first prototype of many that have followed and many more to come, all reimagining and addressing the different genres of humanistic production, and even speculating on some new ones on the minimal computing key.
In the practice of looking at many more costs than a regular design analysis, for the sake of restorative and justice work—we see what remains unseen in a double sense, and connect what was jammed away perversely into separate compartments.
but can one kill Remorse,
a perfect stupefied face of an English lady discovering a Hottentot skull in her soup tureen?
These lines from Aimé Césaire, whose poetry I’m editing in Ed, always gave me the answer to surrealism and anti-colonialism in one poetic shot to the vein. What was supposed to remain separate—the comfortable life of the English lady and the brutal realities of dominance—is brought together again as horror. Surrealism, like minimal computing, participates in a form of reunion between that which is not seen and what is seen. Out of our new horrors, a new form of complex praxis is born: the umbellate virtuosity of necessity.
- For an in-depth account of how “Diversity” is used in contemporary anglophone institutional environments, and how un-effective it is, see Ahmed, Sara. *On Being Included Racism and Diversity in Institutional Life.* Durham; London: Duke University Press, 2012. https://doi.org/10.1215/9780822395324
- We reproduce literary texts online for many reasons. Literary editions are one of the many scholarly products of the humanities, and perhaps the main mechanism by which texts are kept alive, active in our classrooms and societies. In the 1990s the work of many women and minorities by-passed the trends and obstacles of the presses to produce many open editions of literary and historical documents that were not easily available otherwise. The need for that restorative work continues today. For an introduction to these themes, see Earhart, Amy E. Traces of the Old, Uses of the New: The Emergence of the Digital Literary Studies University of Michigan Press, 2015.