paper-as-proposal for storing data on IPFS at the library

[luizirber]

(originally from #1 (comment))

@camillescott suggestion: let's write a grant/proposal for storing data on IPFS at the library.

(We can link this with the DataRescueDavis effort)

Other pointers:

• The interplanetary wayback machine archive: https://github.com/oduwsdl/ipwb

[luizirber]

more links:

• Proposing some tooling for datasets (ipfs-pack and stuff) ipfs/notes#205
• http://www.researchobject.org/
• http://specs.frictionlessdata.io/data-packages/
• http://bjoern.brembs.net/2017/02/data-structures-for-open-science/

[eocarragain]

hi @luizirber . Did you ever take this work further or is any of your team working on using IPFS for sharing scientific/research data?

[luizirber]

hello @eocarragain! The paper never happened, but we do use IPFS in our projects. sourmash has support for storing the search indexes in IPFS (but need better docs and a more streamlined workflow in the main CLI). Poster talking more about how this works.

wort, a database for sourmash signatures of public genomic datasets, also published the data via IPFS. Here is an example dataset. I point the HTTP IPFS link to the cloudflare IPFS gateway. Poster with more details.

[eocarragain]

@luizirber wow. really interesting work. Thanks for the links.

I just gave a general talk this week on how ipfs might fit with more established persistent identifiers for scholarly outputs: https://doi.org/10.5281/zenodo.2548922 (or on IPFS bundled as a DataCrate package: https://ipfs.io/ipfs/QmdhEQrRvvo6mY4iyNz544xChM6nmrJbefvnYNEtVTQzz5/CATALOG.html)

[luizirber]

@eocarragain this is so cool, thanks for sharing! And I didn't know about DataCrate, reading more now =]

[eocarragain]

If you're into that sort of thing, here is a big long rough list of similar "research data packaging" projects: https://docs.google.com/document/d/155lA2BcixTl-zwJHGfLkxsmg7WmQbBK00QWyP8QggkE/edit

Certainly seems to be a common enough pattern across different fields. As a bonus, the static, self-describing, portable aspects of them plays very well with things like IPFS