Computation-tuple sequences and object histories

Reviewer: Krishnarao Gururao Kulkarni

There are many database applications where keeping track of the historical data for the objects in the database is important. Data models have, so far, treated this aspect by explicitly modeling time as one of the modeling constructs, although no implementations based on such data models exist. The authors of this paper take a different approach. They replace the notion of time by orderings among object occurrences (called “computation-tuple sequences” here). Much of the paper is concerned with formalizing the notion of “valid” computation-tuple sequences defined by a computation-tuple sequence scheme. The authors introduce the notion of “local” computation-tuple sequence schemes and show that such schemes are simpler for maintaining the validity of a computation-tuple sequence. The authors then introduce the concept of “bisimulatability,” which is concerned with the issue of establishing whether two computation-tuple sequences convey the same information by “simulating” each other. They then go on to establish some results on “local” bisimulatability which is a constrained form of bisimulatability. Though the paper's main concern is theoretical in nature, the topic it discusses is very practical. The authors have done a good job of explaining the relevant concepts and the data model is also simple to follow. It would be interesting to see how the theoretical results of this paper can be extended to an object-oriented data model as opposed to a record-oriented data model, on which this paper is based. Though it is probably too early to expect, I would hope that this work leads to a well-defined data model and language constructs which are practical in nature, i.e., amenable to efficient implementations.