Show if the data has been "retired" aka is no longer updated

[ErieMeyer]

An idea via twitter: https://twitter.com/bendystraw/status/603352975004667904

Maybe if they haven't been updated in 13 months?

[bsweger]

Thanks for opening this! I was thinking specifically about data that's definitively, well-and-truly retired. Which seems hard to know for sure unless an agency or subject matter expert weighs in.

For example, the Census Bureau dropped CFFR and Federal Aid to States in the FY 2012 budget, so the latest available data is FY 2010.

In contrast, some agencies publish data on an irregular schedule (IRS zip code income taxes) or publish survey data that isn't updated annually.

Above exceptions aside, it could be useful to track "suspiciously old-looking" data, but that risks casting aspersions on information that's not actually out of date. Ideally, agencies would report past submissions as retired.

Many data.gov improvement ideas (like this one) would involve submitters considering their inventory as a holistic entity and not as individual datasets, which seems like a more involved process issue.

[rebeccawilliams]

I think this makes sense for agencies to proactively fill out in their data.json and have added an issue here: project-open-data/project-open-data.github.io#483

On the Data.gov end, would UI that says something like:

"Looks like this dataset hasn't been updated in a year. Please contact [Dataset Maintainer] to help us confirm if this dataset needs updating."

[rebeccawilliams]

Or:

"Looks like this dataset hasn't been updated in a year. Please request data at http://www.data.gov/data-request/ to help us confirm if this dataset needs updating."

[JJediny]

Related to #166

Should consider how to add some automated metrics that average the datePublished/Updated vs dateToday to display a datasets concurrency (i.e. freshness) on average to others for comparison/trends/indicators as part of the CKAN QA extension. In addition should consider a dataset comment thread to identify more recent data published (using connect.gov/myusa.gov to authenticate)... We could also leverage collections to relate periodic released datasets to couple past releases with most recent... if the average of those is 1 year or quarterly then we could possibly identify should-be/soon-to-be releases to stay on top of agencies to make sure they are not neglecting the addition of new releases