Blog: David Loshin

But that's like saying we need world peace or to end poverty. People have been talking about metadata as if it is the cure-all to everything, but every presentation I've seen has either been down in the weeds of metadata repository architecture or so superficial it has no impact.

I think it would be better to say we need more abstraction in everything we do. Data warehouse metadata seems to help no one except developers and MAYBE provide a rollover definition to interactive end-users, but that isn't getting us anywhere. We don't need definitional metadata, we need "relational" metadata, and I don't mean relational database. A standalone atomic definition of a data element is a baby step. How does it relate to other elements? Where does it appear in models? What are the various versions of its semantics?

If that is what Gartner meant, then I think they're on to something, but I'd like some suggestions on how to pull it off.

Your point essentially makes explicit my implicit criticism of the talk. Essentially, there is no doubt that metadata (and I am talking about *real* metatdata that incorporates operational cosntraints, associated business rules, correlation to enterprise information policies, semantics, etc.) is a key enabling component of any forward-think BI activity, but I (as have others) have been preaching this for years in various talks, books, and articles. So why, in 2007, does it appear as a "future trend" for Gartner?