Git version control

Git version control

The scholarly publishing infrastructure doesn’t handle updates well. The version-of-record ideology freezes a single instance in time. Corrections aside, that version remains the annointed “article” or “book” forever. There’s a carve out, in the book world, for “editions,” but it’s normally reserved for the lucrative textbook market. The predominant practice, overwhelmingly so, is to suspend scholarly works in amber.

There’s really no reason for one-offism in a digital landscape. It’s true that the current system tethers citations and CV lines to the idea of a singular record. The ballast of convention is weight enough to keep things more or less as they were for offset printing: a final textual arrangement designated as the publication. But the durability of the single-version ideal is a problem—especially in kinetic disciplines like media studies.

When techno-social churn meets the drawn-out publishing process—peer review, revise-and-resubmit, copy editing, the issue queue—the result is already-outdated work. The current system tolerates the instant-obsolescence for want of an alternative; it’s the cost, in effect, of doing scholarly business. But there really could be an alternative—an open versioning standard that makes updating a routine facet of scholarly publication. mediastudies.press was created, in part, to work on this issue.

Karin Wulf has written that citation “requires some measure of fixity, a version of record.” But the two clauses need separating: Yes, citation depends on fixity of a kind, but not a one-and-only version of record. A proper versioning system would enable scholars to cite specific, traceable versions—each with their own time-stamped version label and unique identifier.

One model is the software world’s version control systems, designed to track changes to programs. The best-known approach is Linus Torvalds’ Git, which undergirds popular services like GitHub. But calls to adopt Git or GitHub wholesale deserve scrutiny. There are a cascade of assumptions about truth and knowledge built into the very language of software development. Any scholarly versioning system must be self-conscious about its borrowings, including inherited metaphors and meanings. Still, the software world’s version-tracking system is a promising foundation, or at the very least an inspiration.

Example of Zenodo versioning

Example of Zenodo versioning

One approach with plain roots in software diff-tracking is Zenodo’s deployment of DOI versioning. (Zenodo is a nonprofit European repository, intended for open works and data supported by European Commission funds but free to all scholars.) Zenodo’s clever solution to the problem of updates is to issue, at the moment of creation, not one but two DOIs. One of them is, in effect, an umbrella identifier, representing all versions of the Zenodo item. The other resolves to that initial, v1.0 version. When an item—a dataset or an article—is updated to v1.1, Zenodo issues a third DOI specific to that version. The result is that the first, overarching DOI—Zenodo calls it the “Concept DOI”—remains a version-of-record proxy. It’s fixed and stable, and always refers to the item as a whole. But the additional, version-specific DOIs permit granular and accurate citation.

Right now the discussion on versioning seems centered on datasets in particular. That makes sense, as updates to datasets are, presumably, more routine; the need for a citable solution is acute. But there’s no reason that Zenodo’s approach can’t work for publication too, and it’s the tack we plan to take at mediastudies.press.