Discussion: Ideas for file uploading

[timwis]

Out of the box, JKAN currently supports linking to files hosted somewhere via their URL, so if you have an FTP server, or dropbox, or google drive, etc. you can just paste in the URL. But I imagine we can make that experience a little more seamless. Some ideas:

• Use a /files directory in the jkan repo. A drag & drop interface commits binary files to it. (Would probably increase build times)
• Have users create a separate jkan-files repo and use a drag & drop interface to commit binary files to it there in the gh-pages branch so they'll be served over a CDN
• Filestack stores files for you and provides a JS lib for the interface. Free tier allows 250 uploads per month and 3GB bandwidth. Not open source.

Any other ideas? The basic requirements are that there be at least a decent free tier and a way to upload via HTTP.

[mheadd]

My vote would be to keep JKAN as a directory, rather than a repository, unless there is a compelling reason to do so.

Google Drive, Dropbox and some other services that make it easy to set up FTP servers should meet most people's needs. And this allows you to keep JKAN simple and easy to use.

[JJediny]

Agree with @mheadd best to keep it simple and avoid offering an uploader via JKAN... as users can use any publically hosted file as a source, but it might make sense to offer a JS lookup of files in a specific directory within the repo like files. This would allow users to associate/link to an already available file in the repo?

This would also make it easier to re-associate/test/reconstruct links (independent of forks) while simplifying the experience for the user with a dropdown list?

[waldoj]

I don't have any clever ideas to contribute about how to do this, I just want to say that I think that this is a fine idea, but one that I'd hold off on for a v2.0. Right now, JKAN is a catalog instead of a repository (a distinction, BTW, that I'd never seen articulated before), and if was me, I'd focus on being a great catalog first. This seems like it could be a big time sink.

[timwis]

Good points guys, thanks. Just want to add in @aborruso's comment from #34 as it's quite relevant here:

About github and publishing using "official" open data schema I think it's useful to look ad git-data-publisher

Example generated data page

[iltempe]

Actually i link filetto that i store in another github dataset. I prefer separate data from website...

[mheadd]

@timwis Git Data Publisher looks pretty boss. I will say that the auto generated datapackage.json is a nice benefit as well. makes it super easy to push to a full data portal with API if desired.