It's astonishing to me how difficult it (still) is to design a syncable local-first data model.
I keep thinking I've found a decent way, then realizing its flaws, then despondently noticing that the flaws were already discussed in Ink & Switch's article: inkandswitch.com/local-first.ht
Conversation
Arguably things are even worse than discussed in that article. Take Firebase, for instance: it implements offline caching, but that's very different from sync. You have to design a whole replication strategy on top to get something like "a synced file format."
1
20
CouchDB seems like the closest solution, if you can design a conflict-free model. But I spent the last week getting into the details of actually operating a multi-user service, and I am now quite thoroughly spooked!
8
30
Happy to discuss my approach with :) Uses CRDT-based data that is stored in local sqlite that is query-able with normal sql queries. Seamlessly syncs in background. Uses hybrid local clocks & merkle tries to verify.
Video here:
2
9
Yeah, that's roughly the approach I'm using… but jeez, so much complexity—so painful.
There's definitely space for a this to be abstracted away. imho, the tradeoff of this complexity is a powerful model for features like undo. Worth it for me.
Hard to get this abstracted away though. Lots of perf & other implications.
1
2
At least with CRDTs, I'm not really scared about data inconsistencies. The theory behind them is quite simple which makes syncing understandable at least.