On Mon, May 18, 2015 at 7:35 PM, Roger Dunn <
tahoe...@gmail.com> wrote:
> The use case is that the normalized data will be fetched often, is very
> tabular in nature, and the additional data is more document centric and will
> be fetch less often, by quite a margin.