dathub

[okdistribute]

Hosting this here now: https://github.com/karissa/dathub

Vision

We want to empower people to use dat to foster collaboration with data. The primary experience will be viewing information about and interacting with a particular dat owned by an individual or organization.

Key Functions

People will participate so that they can share their data with others and retrieve data from others. They will use the registry as a way to:

1. create a central data repository for their data in a persistent, scalable, and easily sharable way
2. create a citation for data in scientific papers
3. find data to use for their own analyses and to reproduce research

The dat data registry is NOT:

• A data hosting service (yet)

Initial proposal for features

As a user, I can

1. create an account/login
   • login with github
2. pubilsh a dat: create a registry entry for my hosted dat and get a static link
   • UI: with URL to a file/folder
   • CLI: dat publish URL [options...]
   • with metadata: tag, keywords, license, schema
3. view a published dat
4. view/search for published dats
   • by name
   • by user
   • by keyword
5. view dat-edtior of a published dat
6. do operations on a published dat
   • fork
   • PR

As an owner of a published dat, I can

1. add keywords to my published dat
2. change the name of my published dat (which does not change the static link)
3. add/edit a README
4. add schema (column definitions/descriptions)

Unresolved questions:

• how will we connect a user on the site w/ a user on the command line?

[finnp]

Cool to see that you are working on this. What does "format-agnostic api" mean? Can people just link their csv files or should the data be hosted through dat?

[okdistribute]

Hey @finnp , thats a good question. It might be better to make everything a dat so we have uniform behavior (i.e., can use dat-editor to view). For small csv files, we can offer an uploading service and do the heroku magic for them.

Yeah, "format-agnostic" might not be the best wording. I just mean it shouldn't matter how the data is stored (essentially, use dat!)