Archiving data
Mathematical publications have a well developed infrastructure for preservation over long time scales. Libraries worldwide provide the ability to find and read articles from a century ago.
Datasets do not fit trivially into this framework, since they are not published in journals or books, and sometimes contain interactive elements. In order to preserve this kind of mathematical knowledge, creators of datasets should take active measures. Fortunately, there are now good options for doing so.
Zenodo is hosted by CERN and provides long term storage for scientific datasets. They will currently host up to 100 files, each up to 50GB. Consider submitting a copy of your data to Zenodo, even if you are actively hosting it elsewhere.
It is also worth mentioning that Github provides archival hosting as well; the SymbolicData project chose to transfer their content there when they ended the project in 2020.
Software Heritage is focused on source code more than data, but is also worth considering for some archival needs.
This page is a stub. If you are interested in writing a better version, please contact an editor or submit a pull request.