How to handle metadata issues

[faroit]

I guess finding missing links and metadata can be quite some work resulting in some discussion... Can we agree one issue per missing metadata? Maybe we make it more readable by introducing a few nice labels?

e.g.

---

Issue Title: ACM_MIRUM
label: urlerror
text: 403 erro

[ejhumphrey]

1 - great idea! it is a bunch of work, but once every error has an issue, and we're sure we've covered our bases, I'd be comfortable rattling some cages via the community mailing list. Folks are generally motivated to make their data accessible, since link rot can undermine their citation counts and such.

2 - what if we change broken URLs to point to the dataset's corresponding issue on github? 🤔then the general public could upvote or contribute on the most sought-after datasets.

3 - is there an easy way to at least semi-automate the process of creating this? I'm thinking maybe some light parsing of the build error logs to produce markdown in whatever format we decide for (a) easy + consistent copy-pasting into github, and (b) keeping track of which ones we've created? Could even collaborate through a google doc to share state?

[ejhumphrey]

gah, quickly realized (2) won't really work, because then that would break our URL testing ... but maybe a slightly smarter URL checker would do, e.g. fail if not found OR github issue at this repository.

[faroit]

i like your idea, lets do that.

concerning (3) we can use the github api from python to automatically create issues. I've done that before.