thearchivist.funHarvard's Library Innovation Lab releases a Data.gov archive
Harvard Law School Library's Innovation Lab announced a Data.gov archive on 6 February 2025. The release described a 16-terabyte collection containing more than 311,000 datasets harvested during 2024 and 2025, with access through Source Cooperative. These figures describe the announced collection at that time; they should not be read as a current inventory or as a claim that every US government data system was captured. The project paired preserved data with metadata and work on authentication and provenance. It also released open-source software and documentation so others could inspect or reproduce the approach. That combination is important: copying a dataset is only one part of making it useful later. A future reader needs to know where it came from, when it was collected and how the preserved version relates to the original publisher. The Lab's broader Public Data Project describes preservation, discovery and monitoring as connected tasks. A public dataset can change, move or disappear, and the archive's value includes documenting versions rather than merely keeping a link to the latest location. The initial collection became an example of library practice applied to a large, distributed set of digital government resources. For researchers, the practical lesson is to preserve the context required to interpret a number. Keep the dataset's schema, units, geographic coverage, time range and collection date with the data itself. A file that opens successfully can still be misread if those details are missing. Likewise, an unchanged checksum confirms the identity of a file, not the accuracy of the measurements inside it. This release belongs in internet history because it treats public data as an ongoing stewardship responsibility rather than assuming that online publication guarantees durable access.
Context and provenance
The primary announcement establishes the release date and reported size, while the project page supplies the continuing institutional context. Keeping both helps distinguish a launch snapshot from a living initiative. For a small research archive, the same method scales down: save a dataset version, its documentation and a short note explaining why it was selected. Record gaps honestly. A collection described as complete needs a defined boundary, such as the resources linked by a particular catalog at a particular time. Without that boundary, completeness becomes an impressive word that cannot be tested or maintained.
Why it matters
The release combined public-data preservation with provenance, reproducible tooling and a library's long-term stewardship model.
Sources and evidence
Editorial review: 2026-09-23. New evidence may change this account.