Hacker News Reader

Comment by pama | Hacker News Reader

They mention replication and optimistic concurrency to allow multiple writers to work in parallel.

The optimistic concurrency is a feature of SQLite, not their sqld server.

And I think there are other approaches to replication that still retain the embedded aspect; litefs, litestream.

I guess I'm thinking, if it looks like postgres, and it speaks postgres (wire protocol) and it has features or replication like postgres, why wouldn't you use postgres?

Obviously because you already run sqlite and need to access it remotely for some reason. You might want to standardize how applications like Datasette works and use a known working library for the postgres wire protocol.

>why wouldn't you just use postgres

Because postgres is not single file?

Why do you want something that writes to a single file?

The mental model of "a single DB file" is a easier to grog than a large and seemingly opaque DB server.

Most (all?) of the well-known RDBMS are designed to host the data needs of multiple separate applications, which makes configuration look a lot more complicated when all you want is n=1.

But then why do you need to write to the DB remotely?

It seems like the inflection point moving from "single-local-application" to "writing-to-remote-db" would be an appropriate point to switch from a single file DB to a larger RDBMS.

I can think of a few scenarios where it could make sense, like a shared small DB on a LAN that doesn't need a full fledged "server", but a primary/secondary relationship makes sense. But those are pretty few... and n>1.

Obviously someone wrote this because they felt a need for it, but I'm struggling to see the use-case. Maybe it's good for CI/or unit testing?

The use case mentioned for http is: you have edge services that need to be able to query a db and http might be the only open way to communicate. Exactly why you run on the edge or use sqlite is actually off topic. I mean I run User Mode Linux in production because it was just easier than the alternative when we first developed it.

I self host a lot of stuff at home. I don't need a large RDBMS. A lot of software is written for Postgres and I don't bother running it as I don't want to manage it. I'm really interested in this if I can start using those apps now, without spinning up postgresql.

> easier to grog than a large and seemingly opaque DB server.

Not really, with this solution now you need to manage a non-replicated, non-high-availability mock of Postgres where every user is 'postgres', dealing with the usual remote RPC call issues and latency.

If you need to manage a RDBMS anyways, you probably can manage Postgres; if you can't manage Postgres then it's unlikely that you will be able to manage this solution as well.

Reply on news.ycombinator.com