``History teaches everything, even the future.''
-Lamartine, speech at Macon 1847.
The Time Capsule File System (TCFS) project was started by Dr. Alan Bawden to address the problem of preserving archival data for MIT's Artificial Intelligence Lab and Laboratory for Computer Science. The project's members currently consist of Dr. Bawden, Pandora Berman, and Brian Zuzga. To date, Dr. Bawden and I have established a universal framework for preserving any archival files in a platform- and medium-independent fashion and Ms. Berman and I have begun the migration of approximately 77 gigabytes of the Labs' Incompatible Timeshare System (ITS) backup tapes to the TCFS format, as proof of the concept. I have prototyped tools to help read and manipulate this body of TCFS files so users will be able manipulate and use this vast amount of data.
The motivations for this project are the extremely poor condition of the old backup tapes and the vanishing base of people who understand how to decode them. We are losing data every day, and some of this data may prove valuable for historical, legal, and personal reasons. Also, a profusion of formats exist at the Labs, making it difficult to track and maintain them all. A single universal format will reduce the chances that valuable data would become stranded in some esoteric format. We know that some of these tapes are extraordinarily valuable right now, but we do not know which ones. There are also tapes that are not valuable now, but may be in the future.
This document contains three basic sections. The first section, consisting of chapters two and three, deals with the overall TCFS framework that has been established for archiving any file worthy of preservation, regardless of its format. The second section, chapter four, addresses issues encountered specifically while processing the ITS backup tapes. The third section, chapters five and six, discusses tools to help users deal with the massive quantity of TCFS files that have been and will be created. These tools are not application-specific, but neither are they a core component of the TCFS system, since they are not held to the same standard of robustness over time.