May 13, 2018

Baker’s Smudges

Open source collaborations hope to revive the insights once garnered from dirty catalog cards

In 1994, novelist Nicholson Baker published an article in The New Yorker called ‘Discards,’ a fervent defense of the library card catalog, then rapidly being replaced by the online catalog. He focused on the richness of the information contained on those cards, which often bore the traces of decades of painstaking notation and correction by catalog librarians. Much of that information was lost when cards were converted to online catalogs, a hollow victory of efficiency over scholarship, in Baker’s view, made more galling by the enthusiasm some librarians expressed for the retirement of their old catalogs.

If Baker’s article looks tame, it is because the intervening years have relegated the card catalog to historical artifact. Online catalogs have long since won, though curiously not, as Baker anticipated, because evolved online catalogs have changed the equation between card and computer file. On the contrary, online catalogs have proved as resistant to innovation and evolution as the card catalogs they replaced, whether one considers the content they include (or, more to the point, exclude) or their equally limited discovery capabilities.

Baker did get one essential point right. In the midst of a discussion of the alleged disinterest of librarians in their own catalogs, he points to the absence of thumb smudges on the cards of books about cataloging and then experiences a small epiphany: ‘Who knows what diligent researcher who photographed (from above, on a tripod) each close packed drawer of Harvard’s Widener catalogue with a high-contrast camera might find out, were he to correlate his spectrographic dirt-band records with the authors that, as distinct clumps, exhibited some darkening?’

Baker was right to call attention to the data lost in conversion to online catalogs: prestandard data such as dashed-on entries, local data, even incidental data such as dirt. That anyone would use card dirt to help make choices is a testament to both the need for choice-supporting data as much as the catalog’s failure to offer some better means of providing it. Online catalogs offered superior search strength, but the sterility of their records made their results less useful owing to information that didn’t transfer to MARC format. Librarians of my ilk systematically undervalued the dirty, prestandard information in our discarded card catalogs. If those smudges have relevance today, it is as a prompt to build such data back into our modern online systems.

Indicators of use

Even without the spectrographic dirt analyses that Baker proposed, it’s hard to imagine a serious objection to the premise that relatively good books got relatively high use and that their catalog cards got dirtier as a result. That said, card dirt is beset with the limitations native to other use-based measures such as circulation or citation count. All undervalue recent works or works in specialized fields. But the real trouble with dirt is that it’s dirty: qualitative judgment is the life blood of academia and much too important to allow such circumstantial evidence to be taken seriously.

Until today, that is. Immersed as we are with social web phenomena such as Wikipedia, Google relevance ranking, and Facebook, we now find evidence of collective judgment trumping gatekeeper authority at every turn. The means by which individual behavior is aggregated into collective judgment hardly matters: such is our confidence in the value of other people’s behavior, even when the evidence is circumstantial and we know nothing about their level of expertise. James Surowiecki describes this phenomenon in his book The Wisdom of Crowds (Doubleday, 2004), which makes the case that some types of decisions are actually better left to groups than individuals, like the Zagat’s Guide to restaurants. From a modern perspective, card dirt seems no more troublesome as a quality indicator than the recommendations presented in Amazon’s ‘customers who bought this book also bought’ feature.

Baker’s smudges also place the success of the card catalog in a more nuanced light: the catalog was exceptionally effective in meeting the ‘finding’ needs of its users, but it did not help them choose among alternatives. Therein lies the narrower significance of Baker’s smudges for future library systems: such systems need to give users many and effortless ways of moving from finding to choosing and back again.

There is no mystery about what it might mean for a library system to support choosing: we are all familiar with Amazon displays, which typically devote less than ten percent of their real estate to ‘finding’ and virtually all the rest to a dizzying array of choosing elements such as illustrations, reviews, sales data, page images, and so on. This is not to say that libraries need to ape Amazon. Libraries do not sell books, their content actively supports academic work, and they maintain far richer metadata – in a word, future library systems should support choice on their own terms.

The well-used card catalog

In a broader sense, a smudge represented a dimension of a catalog that owed nothing to the intent of its inventors. This leads to the second lesson of the smudge: future systems need to be designed so as to anticipate change, whether in data, architecture, or user needs.

Our systems need to adapt and grow in ways we cannot anticipate today. Future innovations won’t necessarily replace the integrated library system (ILS) in the near term, particularly in regard to mature technologies like cataloging or financial control modules. Many ILS vendors have new products intended to enhance searching by sitting on top of bibliographic databases, not by replacing them. It remains to be seen, however, whether these systems can avoid the proprietary code pitfalls that plague their ILS, like persistent bugs and the inability to customize or integrate with other systems.

Open source collaboration models

Some libraries have long since given up waiting for vendors to supply better systems and have instead begun taking matters into their own hands. One such example is the University of Rochester’s eXtensible Catalog (XC) project. A grant from the Andrew W. Mellon Foundation funds the writing of a project plan for the creation of an open source catalog that would contain all of a library’s legacy MARC metadata alongside metadata from any number of other sources. Part of this project plan is to be technological, but an important component identifies similar projects and then creates a framework for broad-based cooperation and community-based code development. Finally, the grant also provides for a systematic approach to incorporating student and faculty work practices into system design and usability testing to enhance user success. Both may help us avoid relying on Nicholson Baker to tell us what users need from future library systems.

The word extensible has two meanings in the context of this grant. First, the proposed system is intended to accommodate multiple metadata schemas so as to allow searches across any number of systems. The second meaning of extensible derives from the system’s open source nature, which will encourage smudgy community code building capable of taking the system wherever its members wish it to go, from portal or courseware integration to push technologies to quick and easy bug fixes. The energy behind both meanings of the word extensible comes from the abiding desire of libraries like ours to break free of the ‘get in the queue’ approach to development of our most important, even defining, online product.

The XC project will consider a variety of technological models before making its recommendations, but one approach would allow for smudgy functionality development by simply copying bibliographic records into a second database, in either XML or SQL format. The ILS would remain the database of record for the types of data it currently manages, leaving XC free to reinvent how that data works, how it integrates with non-ILS data, and how it appears to the public. XML catalogs are not new, with systems in production at the National Agriculture Library and University of Buffalo, among others, but the success of these forward-thinking libraries suggests much unrealized potential, particularly with regard to system integration, user-centered design, and community stewardship.

In the best spirit of Linux-like code building, however, it may not matter which technology the XC project plan recommends, or even which of librarianship’s many alternatives manage to attract a durable community. The most important aspect of the XC project may be its potential to galvanize libraries to try once more to ‘build it ourselves,’ this time with networked communities and open source code. Librarians face an urgent imperative to develop systems that can quickly and easily evolve along with both the work practices of users and with the scholarly communication environment. Academic work is at its best when students and faculty engage the best quality scholarship. Library system development will flourish so long as it remains mindful of the smudges: the evidence of new potential to contribute to the core mission of libraries.

Stanley Wilder is Associate Dean, River Campus Libraries, University of Rochester, NY

Catalog Annotations

If you are truly yearning for the old card catalog, the Ann Arbor District Library (AADL) online catalog features a nostalgic tip-of-the-hat to the low-tech library. Each item record has a link to its own old-style catalog card.

Using a combination of PHP and a graphics tool kit called GD (both are open source) and bibliographic record data, a photorealistic card can be generated dynamically. In addition, users can contribute marginalia to the card so that they can share their opinion of the material with the rest of the world. Even though the marginalia is typed in, randomly chosen handwriting fonts lend an air of authenticity.

The system also allows users to build their own personal card catalog collections. With a single click, a card can be added to a user’s personal collection. Users then have the option to share their card collections publicly (

I originally wrote this software as a bit of a lark but also as an informal experiment to gauge public response to social software in the OPAC. Usage has been phenomenal, and a number of people have reported a true emotional response to seeing these cards integrated with the AADL catalog. While the system does not recreate the feel of the card stock or the smell of the ink, it seems to be a gentle reminder that the library is a service by people, for people. [For more on Blyberg and AADL, see ‘Library 2.0,’ p. 40.]

To see the cards yourself, visit the AADL catalog (, do a search, and look for the ‘Card catalog image’ links.

John Blyberg is Network Administrator and Lead Developer, Ann Arbor Public Library, MI

Link List
Ann Arbor Public Library Catalog
Cards from the Rapidly
Disappearing Library Card Catalog
Creating a Virtual Card Catalog
eXtensible Catalog (XC)
Princeton Catalog with Card Images
Maker Workshop
In this two-week online course, you’ll create a maker program that aligns with your budget and community needs, with personal coaching from maker experts—from libraries and beyond—May 23 & June 6, 2018.