I am a huge fan of David Weinberger's writings and blog, but his recent Metadata and Desire article seems pretty vague. In general, the idea that data and metadata are similar, but different, is often confusing, and I think David's arguments about the limitations of metadata fail to clarify what he really means by metadata.

In a closed system of data, say a table listing first name and last name, there can be metadata—for example, a catalog which lists that this table is called NameTable and that it lives on my C drive. The table is data, the table definition is metadata. Clean boundary.

If you don't have a strictly closed system, you can still draw boundaries and talk about what is maybe inside (data) or outside (metadata). But, if you don't make clear what your boundaries are (and, in most cases, where there are multiple ways to look at it, which set of boundaries you are talking about), then talking about metadata having some issues different than data issues can be pretty vague.

For example, David writes:

But the most important fact about metadata is also its limitation. For metadata to work, it has to be part of a "schema," a set of categories and relationships that computers can make sense of.

How is this different than data. Is the numeral 23 not part of a "schema"? It is, in fact, part of a schema called a number system. Same with alphabets, dates, addresses (at least to some degree), etc.

I think what David is getting at is that, in an open system like the Internet, creating new metadata schemas and sharing them is not easy. But, I think the same can be said for data schemas as well—and I am not sure there is that big a difference.

Is it really a bigger philosophical dilemma if people try to use FOAF than if they try to use a two-digit year date format? Is it really a bigger philosophical issue converting from one metadata schema to another than converting from one date format to another?

One of the reasons boundaries are made between data and metadata is that, typically, data is the stuff that is easy to convert and manipulate and metadata is less so. It is easier to display a date in a different format than to restructure a database—in a database. Maybe in some other systems, the opposite is true.

In an open system, if people work with metadata like its data, well, then it is like data to them. I think folks who query RDF data using ontologies to map between schemas basically treat schemas like data (from the point of view of the ontology metadata).

To me, the real issue is that all metadata should be allowed to be data, and made accessible as data. The semantic web might envision something like this, except it seems to keep requiring more and more abstractions—in other words, in order for something to be more metadata, it has to be expressed in a more abstract system than the data it describes.

I think David's critique is actually in reaction to this. As he aptly concludes his article:

But that means that metadata, an abstraction of an abstraction, is directly and intimately tied to human projects and human desire. And what's desire? Nothing but the way we're pulled into the world, over and over, against our will and in ways that constantly surprise us. So, the increasing need for metadata pulls us out of the world as our desire continuously pulls us into the world.

Again, I think he could say the same about data. But, I also think that the reason for the increasing abstraction is perhaps the desire for increasing abstraction, rather than need for metadata itself.