donovan@julius.csl.sri.com (Donovan Hsieh) (06/01/90)
The following messages are posted for Alice Schafer of Mitre Corp "als@mbunix.mitre.org". Please respond to her directly. ************************************************************************ ************ We are looking at the task of specifying data models for a very large repository of scientific measurement data, metadata which indexes and describes the data, and scientific reports, algorithms, and resultant data based on the original measurements. Questions about models includes issues of what "language" the abstract data model should be created in, (e.g. extended Entity-Relationship, functional data models,object-oriented, petri-nets,etc) with resultant implications for what can be modelled in that language, as well as choices of logical db models, and choices for a common data interchange standard. I am interested in hearing about people's experience in using different models both as their high- and low-level design language. What have you used? In what domain? What would you reccommend as a modeling language and why? References to documents would be appreciated. A major issue is the smoothness of the translation from abstract layer to various logical database models-what has been tried? what worked? what are the pitfalls? Dovovan Hsieh mentioned that there are algorithms to translate the E/R model to normalized relations. where would one find these algorithms? any software? Has anyone translated E/R or other models into an objected-oriented model? I know that an object-oriented requirements analysis is extendable into an object-based system much more smoothly than a functional-based requirements analysis. Other issues of interest are: -the minimum common access capabilities- designed, i should think, based on the abstract data model & constrained by the heterogeneous underlying databases and their capabilities and data content; -common data interchange standards-; and -usability- can a knowledgeble non-programmer understand the data and access the desired information? do the models help, get in the way, or are simply irrelevant because user's interface ignores model? Discussions on these and related issues as well as references to studies would be appreciated. Alice Schafer Mitre Corp MS-A156 Burlington Rd Bedford, MA 01730 email: als@mbunix.mitre.org fax:617-271-8752