http://dl.acm.org/citation.cfm?id=500091
SIGMOD ‘75
Because the user interface in a relatonal data base management system may be decoupled from the storage representation of data, novel, powerful and efficient integrity control schemes are possible. This paper indicates the mechanism being implemented in one relational system to prevent integrity violations which can result from improper updates by a process. Basically each interaction with the data is immediately modified at the query language level to one guaranteed to have no integrity violations. Also, a similar modification teohnique is indicated to support the use of “views,” i.e. relations which are not physically present In the data base but are defined in terms of ones that are.
Integrity of stored data can be corrupted in at least two ways: 1) By concurrent update by two or more processes; 2) byinadvertant,improper or malicious update by a process. The first mechanism is a well known operating system problem [1] which is addressed in [2,3] in the context of a relational data base system.
The solution of these problems at the user language level should be contrasted with lower level solutions.
a (complete) query language which frees the programmer from concern for how data structures are implemented and what algorithms are operating on stored data …
Intuitively, a one variable aggregate free assertion specifies a condition which is true or false for each tuple in a relation. Hence, integrity assurance reduces to checking each tuple that is inserted or modified to ensure the truth of the assertion.