Isn't there bottleneck at your egress - I thought sqlite was fast at reads but only handles one read at a time - so if someone is downloading a gigabyte binary out of your db every other connection just has to wait - am I wrong?
Something I want to try is using sqlite as my filesystem but just storing content hashes that would point to an S3-compatible object store, so you get the atomicity and rollbackability and all but you also get massive parallelization and multiregionality of an object store
Edit: I googled it again and find that multiple processes can read sqlite concurrently so shouldn't be a problem
There’s still a bottleneck in that you have to host it on the same machine and egress from the same network interface. Classic tradeoff between consistency and availability though.
Something I want to try is using sqlite as my filesystem but just storing content hashes that would point to an S3-compatible object store, so you get the atomicity and rollbackability and all but you also get massive parallelization and multiregionality of an object store
Edit: I googled it again and find that multiple processes can read sqlite concurrently so shouldn't be a problem
https://stackoverflow.com/questions/4060772/sqlite-concurren...