DBpedia is often described as a Linked Data hub. Its about page even describes it as a Nucleus for the Web of Data. A glance at the LOD cloud diagram shows DBpedia front and centre with lots of datasets pointing in.

The implication is if you have some book data and I have some book data, we can both link to DBpedia as a central identifier hub and it becomes possible to triangulate between our datasets. Which is fine. At least if the identifiers don't wobble.

But I wonder what happens if you have some book data and I have some music data or you have some health data and I have some nutrition data etc. DBpedia has data from lots of different domains. By showing it as a single bubble on the LOD cloud it's implied that it's internally coherent, heavily interlinked and homogenous. I wonder if that's true.

Lots of people have looked at the link density within Wikipedia but DBpedia relationships are only the subset it's possible to auto extract (mainly infoboxes and categories). And most of the links within Wikipedia are inline in the article text. So how interlinked is that subset of Wikipedia links. Would it be possible to follow your nose across domains from punk to the Sex Pistols to Jamie Reid to the Situationists like you can in Wikipedia for example? Or not?

Or if you made a diagram similar to the LOD one but only depicting link density inside DBpedia would it look like an homogenous data space or more like a microcosm of the LOD cloud with lots of sparsely connected islands. Is it a hub or a set of hubs that happens to be extracted from the same source and hosted in the same place? If it's the latter it kinda implies that we don't have one Linked Data cloud but several?!?