Many things seem feasible with competitive object storage pricing. Still needs a little a bit of local caching to reduce read requests and origin abuse.
I think rclone mount can do the same thing with its chunked reads + cache, wonder what's the memory overhead for the process.
> Looks great! How is the migration story when schema changes are needed?
We have an admin API you can use to write migration scripts. The process is a bit manual though, and a more integrated solution is on the roadmap
> How do you deal with old clients?
Instant treats the the backend the source of truth. If there's an inconsistent cache, we drop the cache and fill it from scratch. We tend to write code that's backwards compatible, and suggest the GraphQL ethos: make sure when changing schema that active clients won't break
reply