> curl -Ls http://bit.ly/gh-install-package | sudo sh

Yes let us pipe a script from a changeable uri controlled by a third-party.

But then again Homebrew, Chocolatey, GitLab, etc is also guilty of this!

GitLab only does this on https sites as far as I know and offer a link so people that are concerned can copy the whole script.

You can just download the files, check it and then execute

I do something similar with aufs and snapraid.

Aufs is responsible for the storage pool (union mount). I look forward to replace it with something integrated in the Linux kernel (e.g overlayfs).

Snapraid is responsible for the redundancy. It kind of implements an offline RAID5 scenario per file —instead of per disk block—, it only maintains a parity disk. Of course since it is offline, you have to update the parity every time you change a file. Usually once per day or per week is ok for long term storage. This approach offers many advantages. Every data disk just works everywhere since it has a common filesystem in it. Via parity data you can undelete files, find duplicate files and correct read errors. You won't lose all drives if you lose more than parity+1 drives. It allows scrubbing and rebuilding.

Given the restrictions (no large files, no open (for long time) files, not lots of small files, nothing you need accessible 100% of time) what use case does this fill?

Wow, that's um, not awesome. And why would anyone use this instead of ZFS? Anyone? Anybody?

This is not a good HN comment. There's nothing wrong with being substantively critical, but there's no information here, just a slag.

https://news.ycombinator.com/newsguidelines.html

Because it's distributed? Kind of a big difference, there.