Consider the caption from the perspective of the archivist — its value is related to its provenance. The first is created by the archivist at the point it is committed to the digital space. The second — the inherited caption — is inserted at some intermediary point between the creation of the photograph and its commitment to the archive.

Let me elaborate a little on this. The caption is a piece of text attached to an object that may require some manner of description. The caption may not always be a descriptive, but for the sake of this article, let us consider it is. Clive Scott provides a very detailed discussion on what the caption (or the title) is, and how it engages with the image. For the purposes of this article, I will refer to the caption in relation to the photographic image, though it may be possible to extend the argument to include other kinds of media as well.

The word caption faces dilution — the caption referred to a curatorial value; auto-caption generation, and other object recognition techniques lead us away from the authority of the caption. Then again, authority itself is a difficult topic to talk about in the context of the caption. The archivist attributes a caption based on the state of her knowledge — the little that can be articulated about the cultural object is provided. By no means should we think that the attribution of captions is without error: an archive I used to work with collects North Indian classical music. A large amount of their holdings is acquired from private collections; the objects that come into their purview would sometimes bear labels, and sometimes they would arrive unidentified. It was the task of the archivist to identify the artist and verbalise particulars of that specific piece. But, the task of the archivist wasn’t limited to that: often, the labelled pieces of would bear errors. The archivist would then need to not only identify the mistake, but also amend the incorrect data. At this point, let us stop and reflect on that erroneous label; what are we to do with it? Should we throw it away, discard it for its inaccuracy? Or, should we preserve it, as the label, once entered into the archive, becomes part of the archive and is, thus, of cultural and historical importance? After all, the mistaken label is someone’s perspective on the contents of the object; can that be rejected as easily as that? Things get more complicated when we realise that the issue might not be about an error and a correction. It could be about a viewpoint that has become obsolete over time; or, it could be the caption of a photograph that was changed to manipulate its reading. Traditional cataloging systems do not provide means to version the caption. That is the root of all our problems.

It’s fairly obvious that modern metadata systems borrow much of their shape and structure from the catalogue. The fields of descriptors are very similar to catalogue fields. No matter how elaborate, and defined the fields are, they still lack the ability to demonstrate versions. And while we are thinking of metadata, it might be prudent to remember that the need for versioning does not only pertain to captions.

Let me approach this from the distance.

Metadata standards have existed for about thirty odd years. Their need and ability to describe the digital object has changed over time. The digital push, with it’s incredible production of content, and it’s need to preserve the material as digital, has seen an exponential growth in importance of metadata. Every digital image is imprinted with machine-attributed metadata. Every webpage requires metadata. Heck, even metadata requires metadata. Data needs to be searchable and accessible: metadata provides the means to do that for non textual objects.

There will soon come a time when the digital data within archives would need to be archived — it is an certainty of time. Why do we do then? Do we take the objects, reject the metadata? Do we take both the object and the metadata together and contemplate a way to articulate them as one being? Should we archive them separately? Or, should we version the metadata for the image? At the face of it, it seems a simpler solution.

For some time now, we have been versioning of computer code. It is something most computer-programmers are familiar with. It is a tested, and functional system. Is it then possible to use a similar system for versioning metadata? That is something to ponder.