Gavia Libraria

Replacing MARC

The Loon has had Jonathan Rochkind’s jeremiad on the overhyping of linked data sloshing around the back of her head since he posted it. She can find very little to disagree with in Rochkind’s adduced evidence, nor in his suggested actions, yet she has not yet managed to agree with the post in toto.

She thinks the root of her unease is Rochkind’s implicit ascription of intentional, even (Rochkind hints) malicious, falsehood to linked-data advocates in libraries. In other words, the Loon doesn’t quarrel with the what of Rochkind’s post, but she has very, very different ideas about the why.

Let her try to elucidate them. To do so, she needs to start in a very different place from Rochkind: with MARC and its associated human and technological infrastructure.

In the Loon’s estimation, the parlous state of linked-data discourse is largely ascribable to librarianship’s inability to tell die-hard MARC catalogers once and for all that MARC is washed up and it is time for them to move on.

The Loon, evil ruthless fowl that she is, is not in the least shy about this: MARC is washed up. So are many practices associated with it; they do not provide anything like sufficient value for what the practices cost to implement. Indeed, they often get in the way of library-system usability for patrons, never mind efficiency overall. Catalogers who can do nothing but MARC and flatly refuse to learn to do anything else are a critically endangered species, and in the Loon’s beady black eyes, their refusal to learn earns them everything they get. She has no sympathy whatever; their learning-refusenikkism tarnishes her beloved profession, and she hates that.

So despite MARC’s washed-upness, it retains a large and vocal community of supporters who will throw up the biggest roadblock they can (which, in the consensus culture of many libraries, means “quite a large one indeed”) before any attempt to change away from it. Moreover, this MARC-refusenik community is not susceptible to arguments from usability in the Loon’s experience; to them, usability is not a thing that is tested, much less a moving target, but a thing that generates spontaneously from MARC’s received wisdom like Aristotelian scallops from sand. As for arguments from technical capacity—what computers can and cannot do with MARC records—honestly, the Loon was recently waylaid for three-quarters of an hour by a cataloger who was dogmatically convinced that computer programmers lie outright about what they and computers can do. This is not the Loon’s experience at all; library programmers are refreshingly candid about the limits of computation with respect to MARC and ISBD, and they show their work. Nonetheless—how can anyone get through to such people?

Yet MARC is thoroughly washed up. What to do? What to change? What to move to? How to convince the refusenik bloc?

“What to move to” (speaking strictly of serializations) is an intriguing question vis-à-vis Rochkind’s analysis. If not linked data, what? XML (much though the Loon loves it) is not-so-gradually losing its toehold on the web with the advent of HTML5, and even on its old stronghold the digital humanities with the advent of text mining. It is not a horse to bet a great deal on. Relational models and the databases that use them have been tarnished by FRBR’s slow-motion crash-and-burn, and in any case the Loon has trouble imagining a single ERD that any two catalog vendors would agree on, much less any two libraries.

Linked data may not be a great choice (the Loon agrees with Rochkind about RDF’s odd annoying crotchets), but it certainly seems the least bad choice available. (The Loon invites argument on this point! She may well have missed something.) It aligns tidily with librarian fondness for authority control (albeit with identifiers as an indirection layer) and controlled vocabularies. It is flexible enough to avoid repeating the problems MARC created as a “one standard to rule them all” that really did not fit anything much but print books. (Indeed, if MARC/AACR2 practices had not calcified so early, library catalogs, freed to experiment, might be in a far less parlous state than they are. A period of free-for-all data design coupled with good user testing—which, thanks to European national libraries, does appear to be happening—does not seem at all bad to the Loon.)

As for the modeling and vocabulary issues, Rochkind is correct about them, but where will the impetus for fixing them come from if not from the process of ditching MARC? It certainly won’t come from the refusenik bloc! It is not entirely fair to blame linked data for not solving these problems out of the box. Nothing else libraries could move to would either; no serialization can. It’s the motion between one serialization and another that bids fair to force at least some solutions, and again, linked data is as good a target serialization as any and better than some because of its emphasis on identifiers and vocabularies. Block linked data, even for reasonable reasons, block the motion. This will not, in the Loon’s view, lead to the results Rochkind wants.

Moreover, contra Rochkind, linked data is actually solving real interoperability problems in real libraries, archives, and museums. They’re just not catalog-specific problems, which may be why Rochkind did not discuss them. The Loon’s favorite example, though it is hardly the only of its type, is this Museums and the Web presentation, in which modeling various metadata stores as linked data enabled easier search-index and user-interface construction without forcing all the metadata into a single mold.

So. Linked data it is, then, will-we-nill-we. How to get from here to there?

The Loon thinks that the muddled communication and excess hype Rochkind rightly decries is an artifact of libraries’ and their vendors’ stasis dilemma. The actual impetus to move away from MARC (and in the process to solve the various other problems with library bibliographic-description practices) has to come from somewhere, and it must be a substantial impetus if it is to overcome the refusenik bloc’s objections. It is neither malicious lies nor thoughtless hype exactly; the Loon’s sense is that all this back-and-forth is testing the waters for a message, any message, that will catch fire. Mixed in with this is the usual can-I-dump-this-problem-on-the-vendors cluelessness from many less than ideally tech-savvy library administrators (this is an old, old song; the Loon can yodel many verses of it, and she doubts not Rochkind can as well).

What can we do about it? The Loon likes Rochkind’s final suggestion best: “Require linked data plans to produce iterative incremental value.” The Loon would break this apart into two somewhat shorter suggestions: “fix things” and “impress people.” We know in exhaustive detail where MARC and its practices are broken; it only makes sense (and is a shockingly low bar, frankly) to build non-MARC things that simply work better. As for “impress people,” when North Carolina State University did that with its Endeca-based catalog in the mid-2000s, not even the refusenik bloc could successfully prevent the rise of the discovery layer.

There is both good and bad in discovery layers, of course. But that is true of any change, including a change from MARC to linked data. The Loon still favors this particular change, because change is indeed necessary, and linked data is the best of a perhaps-unpromising crop of options. Let us move while the moving is good, and fix as much as we can along the way.