Should we have key fields for random?

[matko]

I wonder if we should have key fields for random.

The use case I have in mind is one where you have a bunch of fields which you consider to be 'core properties' of a document, but which are not unique. For example, supposing we store people without any unique identifiable id. While there's nothing truly unique about such records, people still have a name which distinguishes them from most (but not all) other people. So it might make sense to include this in the id generation, followed by some random suffix.