A look at how the library infrastructures of New York City work, and the “middleware” that holds them together.
As a conveyer belt whirs past, workers pull objects from plastic crates and place each on an empty panel, barcode-side up. The objects pass beneath a laser scanner linked to a central database, which determines how each will be dispatched. After rounding a corner in the football-field-sized room, they drop into bins, which workers route to the room's perimeter once full. There the objects wait to be shipped out.
This isn't a FedEx distribution center. It's the shared technical services center of the New York and Brooklyn Public Library systems, home to the world's second largest automated sorter of library materials. BookOps, as the center is known, is located in Long Island City, Queens (somewhat paradoxically, since Queens is the only borough neither system serves)—near the geographic center of New York City and proximate to the Brooklyn-Queens Expressway and Queens Boulevard.
Ten delivery vehicles, including one refrigerated, climate-controlled van for special materials, are parked on the roof. From here, drivers transport books, DVDs, archival materials, and other library resources to and from 151 branch libraries in Manhattan, the Bronx, Brooklyn, and Staten Island. Along their routes—determined by volume, proximity of stops, and the timing constraints of each location—the trucks also deliver newly processed materials to the systems' central and research libraries (and occasionally to schools and senior centers), reroute errant materials from other systems, and pick up discards for donation.
If you've ever reserved a library book online in New York City, it likely passed through BookOps before delivery to your local branch. If you notice new materials arriving or shelf-sitters departing your branch's collection throughout the week, this too is probably attributable to the behind-the-scenes action at the center. And if you've ever downloaded an e-book or accessed an electronic database through the library's website, you can once again thank the BookOps staff. Since 2013 they've handled the acquisition, cataloging, processing, and delivery of new items for the circulating collections of the New York and Brooklyn library systems and NYPL's research collections while also managing periodical subscriptions and e-resource licenses for the systems' branch libraries.
"Library systems" typically call to mind a public, architectural geography composed of central libraries and branches. Rarely do we give thought to the networks linking these individual nodes, let alone the über-system that extends far beyond the libraries' built environments, catalogs, and databases. But these networks are crucial: the wealth of our libraries' resources and services could never be contained within their walls. If our branch libraries were reconstructed to accommodate the entire material and digital collection its patrons can access, each local Carnegie library would rival the size of the massive Starrett-Lehigh building. Even branches have their own "branch operations," in the form of distributed, off-site storage, servers, and managerial operations.
BookOps is but one example of the integral, if often invisible, components that make up a complex inter-network of library logistics—its built environment, back-stage spaces of labor and resource allocation, and a widely distributed digital terrain—and enable the movement and sharing of library resources and services.
In 2004, Lorcan Dempsey, now a vice president at global library cooperative OCLC, proposed that librarians think about their work, particularly the pathway from resource "discovery to delivery," in logistical terms. The manufacturing and shipping worlds have, of course, long labored over the perfection of logistical systems, and Dempsey argued that libraries' expanding collaborative efforts—among them, the management of e-resources and shared print collections in high-density storage facilities—pose similar supply-chain management challenges.
Book stacks, delivery trucks, call numbers, catalogues, collection management software, and interlibrary loan systems all might be counted among libraries' wide range of logistical media. As I toured BookOps and marveled at its conveyor belts and computer monitors displaying a panoply of enterprise management software, it became apparent just how critical interoperability among these media, their alignments and misalignments, are to the functioning of library systems' many interlocking networks.
Simply put, libraries need software, termed "middleware," to mediate between the countless software platforms they use to operate. But the definition of libraries' middleware might also be extended to include hardware, as the material operations within library buildings and the physical connections between those buildings are profoundly dependent on layers upon layers of interconnected software, just as their digital resources rely on physical, place-based materials and labor. In nodes like BookOps and across library systems, middleware ties the physical and the virtual together.
It's hard to wrap one's head around the breadth of these distributed systems—all the far-flung truck routes, database subscriptions, interlibrary loans, and protocols. But acknowledging this complicated logistical network makes visible the labor, equipment, and expertise required to build and maintain our libraries, one of our society's few remaining intellectual and cultural commons. Grappling with the means by which our technical and intellectual resources are inter-networked also helps us recognize that logistical consolidation can be put toward myriad ends: we can produce a global panoptical Google, or cultivate a universal public resource. We choose whether to advocate for corporate monopolies or cultural commons. Better understanding and investing in the logistical systems that route our power, packages, and—particularly with libraries—knowledge means greater potential to shape not only our built landscapes, but our political and intellectual ones as well.***
The New York Public Library "floats" its collection, which means that collection materials are not housed at any particular branch, but instead move between them based on returns and librarian and patron requests. Before 2010, when these requests began to be routed through BookOps, the library's technical services were cramped into an insufficient Manhattan facility. The 145,000-square-foot BookOps facility had the capacity to bring in additional partners; in an era of decreasing library budgets and amid calls for the city's three library systems to find new ways to collaborate, the Brooklyn Public Library soon joined the NYPL at the facility in 2013. The Queens Library never did; BookOps administrators suggested the decision stemmed from Queens having the furthest to come to align its systems and operations with those of the other libraries.
The collaboration allowed the libraries to bring work in-house that previously had been inefficiently outsourced. Brooklyn, for instance, had contracted with UPS to deliver materials between its branches. Since BookOps came online, the delivery error rate across the two systems has dropped from twelve to one percent. BookOps Director for Logistics Sal Magaddino also estimates that centralized sorting saves the branches two hours of staff time daily. That's time that can now be used to curate and manage local collections and to organize public programs, according to Acting Director of BookOps Charlene Rue.
What's more, the Center for an Urban Future proposes, "the tens of thousands of square feet that were formally dedicated to collection management activities in the branches can now be turned into programmable space for patrons." BookOps also offered new programmatic possibilities. Brooklyn had closed the bindery and preservation lab once housed in its Central Library due to space concerns; in the Long Island City facility they have access to a preservation lab, digital imaging lab, and archives and manuscripts processing area.
The big tent of BookOps further permits staff a macro-scale view of library materials for a more holistic and data-driven approach to collection management. Rue says centralization does not, however, mean homogenization or de-specialization. Acquisitions librarians are still specialists in their fields, and staff at the individual libraries still submit requests driven by their own patrons' needs. In these cases, only the billing and order-placement are centralized—and the system's economies of scale allow the libraries to negotiate better deals with publishers and e-resource vendors. The large staff also includes multilingual employees who can catalog materials in a wide variety of languages, enabling more patrons to discover them. These examples demonstrate the potential benefits of calibrating the division of responsibility by deciding what can be efficiently handled in a centralized fashion and what needs to be managed at the branch level. And these managerial and logistical decisions have political implications: they determine how to address community interests, empower staff, and account for and accommodate patrons' intellectual and social needs.
Rue and Magaddino estimate that the collaboration saves the two library systems, in total, $3.5 million per year. Still, the merger hasn't been seamless. BookOps, Magaddino says, still handles material using a mélange of classification systems: Billings, Library of Congress, Dewey, fixed-order. Materials still filter through the building on three separate tracks for different circulating and research collections. And the most glaring redundancy–or omission–is the Queens Library's absence.
"Floating" has also proven less fluid than the term might suggest. While patrons can request delivery of resources to their local branch online, there's often a mismatch between the discovery afforded through an online hold and that found through in-person browsing. Floating continually refreshes individual branches' on-site collections, but that refresh rate can be inconsistent across branches. Locations near transit hubs and those with longer hours attract the majority of returns, yielding more materials on the shelves and obligating staff there to spend a disproportionate amount of time managing that collection. More off-the-beaten-path branches are left under-supplied. In other words, patrons' commuting logistics complicate the management of book traffic. Librarians at less logistically central branches sometimes have to "shop" for books by scouring online listings for overstocked locations to request their delivery to understocked branches. Again, the political consequences of logistics are visible: without proper management, particular nodes in the network and the people who rely on them are left under-resourced.
As Magaddino acknowledges, "the causes of floating [collection imbalance] will always be there." But an optimized integrated library system (or ILS, essentially an enterprise resource planning system for libraries)—another layer of logistical media—can perhaps help channel the "float." Rue and Magaddino both sang the praises of a piece of software called collectionHQ, a "collection performance improvement solution" that enables librarians to analyze use patterns by branch, assess patrons' demands, and rebalance collections accordingly. The optimization and interoperability of library software would ideally allow for the creation of additional links within the place-based infrastructure.
For instance, the three library systems are moving toward a universal return system, which would allow patrons across the city to return materials at any of the three systems' branches. Universal borrowing—allowing patrons to request items from any of the three library systems' floating collections—is more of a challenge, Magaddino and Rue said, because of disconnects in their various "soft infrastructures": disparate circulation policies, different barcoding protocols, and so forth. That said, according to Sam Rubin, Chief of Staff at the New York Public Library, the new idNYC municipal ID card represents a "soft launch" of a city-wide library card, "since the ID can also serve as a library card for all systems."
BookOps is already the headquarters for another initiative that unites all three library systems, in partnership with the City's Department of Education: MyLibraryNYC serves over 500 public schools, whose teachers can order individual titles or tailored "teacher sets" from the libraries' collections and receive curriculum support for their use. BookOps's drivers drop off and pick up the materials. And students can use their library cards at both their school library and their local public libraries. Such initiatives demonstrate the potential for cross-system linkage as well as fruitful divisions of labor between libraries and other public infrastructures; it's possible to overcome the stubborn technical and bureaucratic barriers that undoubtedly stood in the way of logistical integration.***
Students and researchers, both within the metropolitan region and around the world, are the beneficiaries of another inter-networked logistical system. Fifty-six miles away from BookOps, on Princeton University's Forrestal Campus, sits the 164,000-square-foot, concrete-walled and steel-roofed Research Collections and Preservation Consortium (ReCAP), a book repository collaboratively owned and operated by the New York Public Library and Columbia and Princeton universities. Here the NYPL's logistical network is integrated with those of private academic libraries, and, as is the case with many urban infrastructures, its logistical geography extends into the ex-urban landscape.
The consortium began in 2000 when the three library systems acknowledged their shared concerns regarding collection management, preservation, and space. Up until then, said Jacob Nadal, ReCAP's Executive Director, all of them had been leasing commodity warehouse space to store their overflow materials. Whereas BookOps manages the public branch libraries' circulating collections—those materials that patrons can check out—ReCAP houses only research collections, which, at NYPL, typically don't circulate. And among those research materials, ReCAP handles only those that are very infrequently requested by patrons or that simply can't fit on libraries' already-packed shelves. (According to Nadal, less than 0.02% of the holdings have circulated more than once per year in the last decade.) If it turns out that materials held at ReCAP start circulating more frequently, ReCAP's staff might suggest that those materials return to campus for ready access.
ReCAP's seven buildings currently house over 12.5 million items—books, films, DVDs, microforms, archival materials—with capacity for up to 17 million and room on-site for additional construction. Heavy masonry and insulation, vapor barriers, sealed docks and doors, and a gas- and particulate-filtering HVAC system allow for precise climate-control, providing optimal temperatures and humidity for the paper, celluloid, and plastics that make up ReCAP's collection. Nadal said that storing material at ReCAP rather than in a standard library stack could extend its life by a factor of four. Sensors and an environmental monitoring service maintain prime conditions across changing seasons and weather while also helping to minimize energy use and costs. Meanwhile, a 5,000-panel solar array on ReCAP's roof generates a good portion of the energy required to power the HVAC system—a demand that physical-media repositories like ReCAP share with their digital-media counterpart, the data center.
ReCAP's staff processes 3,000 new items on an average day and handles over 250,000 requests annually. Each item's owning library will have assigned it a barcode used to situate the item within ReCAP's inventory system—on X tray, X shelf, and X row. This is not a collection for browsing; there's no thematic or topical rationale to its organization. The shelving and inventory systems' shared logic is instead one that prioritizes efficient storage and retrieval. When a patron requests a book through the library catalog, that request has to be mediated through a database to allow a staff member to find it.
This translation involves a negotiation between different software systems and organizational logics: one of which conceives of the book as a topical and material unit of knowledge, the other of which regards the book as a thing on a shelf—the tiniest bit of shelf possible. ReCAP consolidates requests three times daily, and its inventory management system generates a "pick list" that identifies the most efficient pattern by which staff can navigate forklifts among those shelves to find the target item—a process not unlike their "picker" counterparts' in the Amazon fulfillment center 20 miles south in Robbinsville, NJ. The retrieved items are packaged in barcoded totes, moved to the loading dock, and shipped out via courier: twice daily to Princeton, and overnight to Manhattan. When the borrowed items return, they're put back right where they came from, in the same box on the same shelf.
In its next phases of development, ReCAP might become more data-center-like in its storage logic, labor logistics, and ambience. Nadal said that they're investigating how to increase the existing facility's storage density, particularly through automated (i.e., robot-driven) retrieval systems like those employed in North Carolina State University's Hunt Library and the University of Chicago's Mansueto Library. And because robots don't need to breathe, the storage facility could be de-oxygenated to eliminate the risk of fire and prevent materials' oxidation-based decay—thus producing a storage system that would not only be illegible to humans, but climatically inhospitable to them, too.
While NYPL's, Columbia's, and Princeton's millions of items share the same storage facility, they're not universally accessible to all library patrons. Some items are coded to be available only to patrons of the specific library that owns them, like Columbia's rare books library, others can be delivered to any branch within the member institution's libraries, and still others are available to patrons at any of the three institutions. Many items are also available via interlibrary loan (ILL)—another layer of library logistics that is much more geographically expansive; when I visited, ReCAP's ILL staff was preparing shipments to researchers in Berlin and Mexico City.
ReCAP is currently working to enhance the interoperability of these myriad systems to "transform the nature of the partnership from management of a shared space … to management of a shared collection." Lack of software interoperability means the libraries can't take advantage of their collections' geographical proximity. They can't simply reach down the shelf to use one another's materials; they instead borrow from one another's collections via Interlibrary Loan. The next step, Nadal said, is to develop middleware that can negotiate between ReCAP's three catalogs and its own inventory management system. Patrons will be able to conduct a "common search" across the entire collection, and all partner library patrons—including card-carrying patrons of the NYPL's research libraries—can have access to these shared resources. Nadal estimates that shared search and access will add four to five million items to each member institution's offerings.
Up to this point, I've resisted characterizing ReCAP with a term that's proven conflagratory in recent debates about the future of libraries: off-site storage. It is the kind of operation that to detractors symbolizes the de-bookification (and thus the supposed de-humanization) of our libraries. Yes, managing a shared collection does commonly involve the "de-duplication" of periodicals or, less frequently, the culling of redundant books that far exceed public demand. And yes, it represents a possible extra day's wait for requested materials and a lost opportunity for serendipitous discovery in the stacks (if they were actually browsable in the first place). Yet much of the consternation over off-site storage is perhaps attributable to the general invisibility of library logistics, to the "unknown unknowns" in library systems' widely distributed geographies.
"Any library of any size has a space crisis," Nadal said. Regardless of whether a library is public or academic, urban or rural, land-rich or land-poor, there's never enough space for all the programmatic elements, including book storage, that need to squeeze into the library building. Libraries "operate in tectonic layers," said Nadal—some more or less public than others—and when controversies like the one surrounding the NYPL's Central Library Plan surface, which called for moving millions of items from its central branch in Manhattan to Princeton, long-hidden strata "emerge to the surface." (For more on this issue, see the full version of this essay.)
All the place-bound physical materials and seemingly placeless digital resources of the library follow distinct paths of acquisition, processing, delivery, and maintenance. And the logistical spaces, like BookOps and ReCAP, that carry out those processes mediate between the human and the machine: they make use of automated conveyance systems that structure, but ultimately depend upon, workers' manual labor.
They use bar codes and conveyor belts in tandem with analog systems—color-coded paper slips denoting a book's language and location, stacks of plastic bins—to route materials to their destinations. In collection development, data-driven analytics rub up against staff expertise and patrons' unpredictable desires to drive decisions regarding what materials to acquire and where to keep them. Patrons access books through digital interfaces with nested layers of software behind them. In all cases, middleware—not just mediating software, but also mediating spaces and people—has to reconcile the old and the new, the hard and the soft, the automated and the manual, the human-readable and computer-intelligible.
Networked libraries are essentially building a commons, drawing on the complementary offerings of its members and making those resources available to the widest possible public. But they're doing so by designing and utilizing systems driven by managerialist values. Both BookOps and ReCAP make use of hardware (i.e., conveyor and "sortation" systems) and software (barcodes and inventories) that originated in manufacturing, retail, and shipping—infrastructure that prioritizes efficiency by saving space, time, energy, money, human labor, and so forth. While these fields' operating principles differ greatly from the library's humanist ideals of education, public service, and the promotion and preservation of cultural heritage, those commercial tools needn't deterministically impose their logics and values, wholesale, on the library.
With careful management, libraries can ensure that logistical systems are customized to address their particular needs. And they can turn to designers and developers—architects, interaction designers, programmers—to create effective middleware, both soft and hard, to negotiate potential operational rifts between libraries' specific needs and logistical systems' standard functionality.
It just so happens that the efficiency-promoting systems also contribute to another kind of "saving": the preservation of library materials, a deeply humanist concern. What's more, by "outsourcing" many of their logistical operations and relying on resources housed in the cloud, libraries can allow their local physical infrastructures to do what they do best: facilitate the coming together of people and material knowledge. Saved space, time, and labor and "bandwidth" are, ideally, reallocated: from managing the collection to serving the public, who benefits from a more robust set of resources, a better functioning and more intuitive catalog to find them, and service and space that's more responsive to its local concerns.
Ideally, with careful design and planning, that's how it works. But that thoughtful planning depends on an informed public—and on librarians, designers, engineers, and public servants who actively engage in developing logistical systems that promote responsive public service. These systems have agency: alignments between their hard- and software structure the library's operations and service, expanding or limiting the publics that libraries can reach. Library logistics aren't simply about coordinating and controlling the movement of people and things; they're also about values and ethics—about how a library serves its constituents, promotes learning, and cultivates community.
Shannon Mattern is an Associate Professor in the School of Media Studies at The New School, where she writes and teaches about libraries, archives, maps, media infrastructures, and other media-spaces. You can find her at wordsinspace.net and @shannonmattern.
Visit Urban Omnibus to read the full version of this essay. For more on how New York's libraries are negotiating protocols, promoting interoperability, and shaping the digital sphere, see the author's blog post "Scaling Firewalls."