Persistence of Data
Records, page 6
In many cases you may need to persist groups of records, where the records are similar to each other in structure. These often take the form of tables of records, where each record contains the same fields. You may need to work with different types of records in the same file (multiple tables). In many cases, you will want the information to be accessible to multiple users or applications at once. Clearly, this is the type of information that is often stored using a database system. This might include use of a specialized database command set or the database controls and commands built into Visual Basic.
You may be thinking that this is a somewhat convoluted way of talking about databases. Why go to all of the trouble to differentiate between "saving records" and building a database? Because it is important to distinguish between the goal and the solution. The goal is to persist a certain type of data: records that are similar to each other. One solution is a database - but it is not the only solution.
You have already seen that it is possible to store records in a document file. However, if a database is clearly designed for record storage, why would you ever want to store records in a document file? You might want to do so if you do not need the database's searching and filtering capability, if you have little or no need to share the file and if the records are simple and of fixed length. In that case not only might the document approach be easier to program - it will probably be faster and will allow you to avoid the overhead of including a database engine with your application.
The case for application configuration information is easier to see. A database would almost always be extreme overkill for this type of data persistence. Even storage of per user application information is more easily accomplished using any of the other techniques.
What about documents? Is it appropriate to use a database to store document information - a word processing document, for example? It certainly can be done. You can imagine a complex database structure that includes one table for the lines of the document (perhaps with formatting information), and other tables that contain different types of objects, perhaps in blob format to hold images, video, sound, and so forth. These tables can be linked through the relational capabilities of the database.
This approach would work - but it can become quite complex - easily as complex as implementing the document using a proprietary format and performing your own disk operations. The database approach also can have a significant impact on performance and involves significant overhead.
