From: Ross Singer <[log in to unmask]>
Remember, when we talk about MARC data, we can't just talk about what's *possible* in MARC21, but what actually is in the records we have: since each record is a discrete standalone document, if the data isn't there, it's nearly impossible to improve upon that.

And that is why this:

Unless we know exactly how catalogers have used the MARC elements to record data about our items (regardless of which elements our automated systems use or ignore) we don’t know what we have to work with in our existing corpus. If you look, you may be surprised by what you find.