Representing information about files
Jeffrey C Mogull · International Conference on Distributed Computing Systems · 1986
Intelligent use of files depends on having sufficient knowledge about them: their purposes, structures, and contexts. Humans have traditionally made do by using their own methods for capturing and manipulating such knowledge, but this is not available to programs, nor is it necessarily convenient for humans. Computer file systems must make this knowledge more widely available, by formalizing its storage and providing efficient manipulations. Explicit and careful treatment of knowledge about files makes possible the effective solution of otherwise intractable problems. These include the sharing of files in heterogeneous distributed systems, and the location of files not by path names but by predicates on their properties. The thesis proposes a model of files that includes the concept of file properties. It then discusses how file properties are used in systems with inadequate property support, and how they could be used in a system that treats them explicitly. Interface specifications are presented for a file system and a directory system based on the model, and it is shown how these systems can be efficiently implemented. Performance measurements demonstrate that such systems are not significantly slower than traditional systems, and are dramatically faster for certain functions.