Alternate file-based sync mechanism

[musirisriram]

Hi,

I use this tool in a network where many of my servers can't access the public internet, and I'd have to work to find a place that's universally reachable where I can reliably host a server.

Can I have a mode where the tool streams history records to a well-known directory, in a simple append-only format with rotation? Then I can use external tools like scp/rsync to keep the directories synced across machines, even when my network topology is multi-hop. (There would need to be a mechanism to auto-import new records as well, so it's not entirely trivial.)

[ddworken]

Hey! This is an interesting FR. I'm generally not opposed to this as a feature, but it seems like a pretty specific use case that I'm not sure too many other people have so I'm unlikely to prioritize this. If anyone else wants this feature, please chime in! Otherwise, I'd encourage you to either use hishtory without syncing (see "Offline Install" in the readme) or to consider trying to implement this yourself.

[achrstl]

I like the idea to have a file-based sync option, using it with a file storage and synchronization service like gdrive would be a good combination.

[kachunkachunk]

Chiming in, it already seems to work nicely on infrastructure where home directories are centrally stored on NFS, from limited testing. But this could be nice to have over the existing self-host option, which I'm about to explore.

Edit: Ah yeah the self-hosted route is really not suitable if your workplace isn't keen on users running servers on the network either. So no-sync mode it is. Arbitrary file storage would be nice, if it can handle multiple sessions from the same user (or multiple users perhaps?) It isn't sounding so different from a self-hosted SQLite approach there, admittedly.