Conversion into Normalized Record Types
for every data flow which either enters or emanates from a data store (in the leaf level DFDs), the integral data elements are identified
for every data store, a list of the data elements which are entering and emanating are drawn up
then the dependencies among all the data elements are analyzed, and the normalization rules are applied in steps so that at every step a given relation is split into more “simple” relations
- every relation has a key which consists of one or more data elements
- every non-key data element functionally depends on that entire key and not on part of it
- no non-key data element depends on any other non-key data element in the relation (there are no transitive dependencies)