Blog post: #Dropserver progress for August 2025. Was a bit slow getting this one out (halfway through September, already?!?) but it's a cathartic one because I spent the month finding things I need to change, add and rip out as the project evolves.

https://olivierforget.net/blog/2025/dropserver-progress-august-2025/

#opensource #selfhosting

Dropserver Progress - August 2025

Olivier Forget's Personal Blog

@teleclimber One thing we thought about for backup (but have never built) is that if your backup is platform-aware you do not have to back up much most of the time. The vast majority of my Sandstorm grains have not been touched in... years. If my backup was aware of the platform it would only need to back up grains (or appspaces) which had been modified since the previous backup was made.

@ocdtrekkie Yep definitely. You could get finer-grained (heh) backups while using less space. You could also keep multiple recent backups for some appspaces/grains if it's an important one.

Another interesting thing to think about is how to integrate (or somehow work with) an OS level or third party file backup system. How can you backup the latest version while also ensuring you have file integrity? I[m thinking in particular of instances running on Mac with Time Machine copying things in the background.

@teleclimber One thing that really surprised me is how many things that people use to copy or sync files which do not ever bother checking hashes by default. A lot of times it's an option, left off in the name of speed.

Ian had been working on a specification for backup tools to use: https://github.com/hashsplit/hashsplit-spec and was building a designed-for-Sandstorm implementation: https://github.com/zenhack/ocap-merkledag that unfortunately was not completed.

GitHub - hashsplit/hashsplit-spec

Contribute to hashsplit/hashsplit-spec development by creating an account on GitHub.

GitHub