Compliance first: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.
Home / Guidelines / Open data policy guideline 23
Open data policy guideline 23

Permanent Access: Open Data Policy Guideline 23

Permanent Access: Open Data Policy Guideline 23

Guideline 23 asks that published government data stay available at a stable location indefinitely, with changes versioned and archived, so that permanence attaches to the data itself rather than to whichever website hosts it.

Links rot, portals get replaced and vendors change. A policy that only promises a website will eventually lose data. This guideline asks for permanence written into the rule itself.

What the guideline means

Once a dataset is public, it should stay public and findable, at a stable address or through an archive, in perpetuity. Guideline 23 stresses that the commitment should attach to the data, not to a particular portal. Platforms come and go; the obligation to keep the information accessible should survive them.

Permanence also covers change. When a dataset is corrected, restructured or superseded, the guideline recommends version tracking and archiving so that users can see what changed and when. These rules work best when they build on existing records-management law (Guideline 2).

Model policy language patterns

Paraphrased permanence provisions include:

  • Datasets published under this policy shall remain accessible and shall not be removed without public notice and a documented reason.
  • A department that changes or ends collection of a dataset, or removes it from public access, shall report the change.
  • Superseded versions shall be archived and remain available for download.
  • Datasets shall be published at persistent addresses that do not change when the portal is redesigned.

Where it appears in adopted policies

Permanence language is less common than portal or format provisions. In this index it appears in:

New York State's 2013 executive order on open data also contained permanence language.

Common pitfalls

  • Platform migrations that break every link. Moving portals without redirects or stable identifiers wipes out citations and integrations.
  • Quiet removals. Datasets that disappear without notice erode trust in the whole program.
  • Overwriting history. Replacing a file in place destroys the time series users relied on (see also Guideline 17).
  • Tying the promise to a vendor contract. If access ends when a subscription lapses, permanence was never guaranteed.

A practical checklist for permanence provisions

  • Attach the permanence commitment to the data, not to a specific website or vendor.
  • Assign stable identifiers or addresses that survive platform changes.
  • Require public notice and a documented reason before a dataset is withdrawn.
  • Archive superseded versions and keep them downloadable.
  • Maintain a change log describing corrections and structural changes.
  • Align the policy with existing records-retention schedules so published copies are covered by the same legal obligations as the originals.

What permanence teaches about the value of operational data

Data buyers value continuity. A model builder who licenses a dataset wants confidence that the history is complete, that past records will not silently change, and that definitions are documented across versions. Those are the same properties Guideline 23 asks of governments.

For a business, that means keeping historical operational records intact rather than purging them when systems change, and versioning extracts so each delivery can be reconciled with the last. Well-kept history often accounts for much of a dataset's value. See how to value my data and the de-identified data guide for how history and privacy work together.

We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Our compliance promise: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Frequently asked questions

Can a government ever remove a published dataset?

Sometimes it must, for instance if sensitive information was released by mistake. The guideline's concern is unannounced, undocumented removal. Good policies require notice, a stated reason, and where possible an archived copy of what remains publishable.

What is version tracking for open data?

Version tracking means recording when a dataset changed, what changed, and keeping earlier versions available. It lets researchers reproduce past analyses and lets users distinguish corrections from new information.

How does permanent access relate to records retention laws?

Retention schedules already require governments to keep many records for set periods. Permanence provisions work best when aligned with those schedules, extending existing obligations to the published, digital copy rather than inventing a separate system.

Policies in the index that reflect guideline 23

PlaceAdoptedInstrumentGuidelines
Montgomery County, MDDec 17, 2012County legislation (Bill 23-12, Open Data Act)457811131822
San Francisco, CAMar 28, 2013Ordinance34578101118
Nashville, TNMay 12, 2014Executive order (No. 043)2345781011
Asheville, NCOct 13, 2015Council resolution13489101118
Naperville, ILNov 15, 2016Council resolution45789101112
Syracuse, NYJul 26, 2017Executive order1456781113

Sources

  1. Sunlight Foundation Open Data Policy Guidelines (historical reference)
  2. Sunlight Foundation open data policy collection (public repository)

Updated December 31, 2024.