Extensibility?
What does it mean to have extensible XML suipport?
This is a question that came up somewhat obliquely in the latest OpenDocument Metadata SC conference call, where I was presenting my draft requirements for the bibliographic use case, one of which was the need for extensilbility. XML, after all, is an acronym for eXtensible Markup Language. Given my focus on metadata, I’ll restrict myself more to that realm.
It seems to me there are largely two views on this question. One perspective—I’ll call it the “document-based” view—says that extensibility is defined first through the simple ability to create new languages, and second within those languages to create strategic extension points.
Another view—I’ll call it the “module” view—sees metadata not fundamentally in terms of documents and complete schemas, but rather in terms of modules of descriptions that can be plugged together, mixed up, or otherwise interact, mostly independently.
This first view suggests to me an image of a book, complete with introduction and conclusion, index, and covers. It’s a more hermetic view of metadata.
The second view is, I think, the view of the web and hyperlinks, RDF, and more recently microformats. Why invent invent elaborate new schemas, this view says, when you can instead mix-and-match from a rich set of existing alternatives?
So when we at the Metadata SC talk about “extensibillity,” then, as a requirement, what do we mean?
I can only really speak for myself, but to me—a partisan of the second view—extensibility has to mean both that one can add custom XML markup and that the markup conforms to some rules such that ad hoc mixing and interaction is possible.
Simply allowing anything-goes addition of arbitrary content achieves little that is useful. While there may well be use cases for this sort of thing—Microsoft’s custom schema functionality surely must be valuable in some contexts—it seems to me it would be counterproductive to not insist on some minimal expectations of interoperability across a document format’s metadata format.
This is not to say that all conforming applications must fully understand extension structures, but it is to insist on the need for at least minimal legibility (for example, the ability to display any foreign content).
