Data Conversion: Doing it Right the First Time - High-Performance Web Databases: Design, Development, and Deployment - page 593

could be attributes based on more than one domain. Not all domain prob-
lems will create conversion difficulties, but they may be problematic later
if it cannot be proven that these were preexisting anomalies and not a re-
sult of the conversion efforts.
Wrong Cardinality
The old data could contain cardinality violations. For example, the
structure may say that each employee has only one job record, but in fact
some may have five or six. These sorts of problems make database design
difficult.
Wrong Optionality
Another common problem is the absence of a record when one should
be there. It may be a rule that every employee has at least one record of ap-
pointment, but for some reason 1% of old records show no job for an em-
ployee. This inconsistency must be resolved by the client.
Orphaned Records
In many cases, a record is supposed to refer back to some other record
by making reference to the key value for that other record. In many badly
designed system, there is no key to refer back to, at least not one that
uniquely identifies the record. Technically, there is no primary key. In some
cases, there is no field available to make this reference, which means that
there is no foreign key. In other cases, the key structure is fine, but the ac-
tual record referred back to does not exist. This is a problem with referen-
tial integrity. This record without a parent is called an orphan.
Inconsistent Redundancy
If each data item is determined fully by its key, there will be no undesir-
able redundancy, and the new database will be normalized. If attempts at
normalization are made where there is redundant information, the DCT will
be unable to make consistent automated choices about which of the redun-
dant values to select for the conversion.
On badly designed systems, there will be a great deal of undesirable re-
dundancy. For example, a given fact may be stored in multiple places. This
type of redundancy wastes disk storage, but in some cases may permit fast-
er queries.
The problem is that without concerted programming efforts, this redun-
dant information almost certainly is going to become inconsistent. If the
old data has confusing redundancies, it is important to determine whether
they are due to historical changes in the business rules or historical chang-
es in the values of fields and records.