Uses SQLite with no way to scale, and bad schema, nothing can go wrong
and also centralized
Discussion
they had to last-second hack together Postgres support
I'm gonna obselete the primal cache. It's so retarded.
I am working my own alternative cache server. Replacing crappy DVMs and the buggy protocol that is with proper algorithms too
🤝
But do you write tests tho?
As you can see I hate people who don't test.
I need to do that for a few projects :)
But I do a lot of verification and manual testing + making the code actually readable. Doesn’t scale too much though.
Yes and especially those who advertise explicitly that they don't test XD and yet somehow that's what gets all the type. not boring old tested software
FoundationDB is peak testing-driven development.
The entire thing is designed to run in a reproducible simulator.
TDD is not my thing I'll admit, I suppose more BDD. I think anytime you refuse to acknowledge the forest regularly during development it becomes a limiting factor, but I guess it heavily depends on the project. A database design has a pretty fixed high level architecture theses days.
I'm butchering relay subscription filters to do graph lookups. I'm sure the devs will love that.
Please do so Semi. Dvms were never meant to be the content discovery thing. It just became the content discovery thing because there was nothing else, besides centralized trending algos on some big platform. Maybe you build something better. Go for it.
The protocol can be significantly reworked to allow direct connection between feed provider and client (reducing latency) and add certain things like pagination etc
Doesn’t need to be bound by event size limits either, and improved privacy
Events can be encrypted, clients just don't do it. Latency is very low when you use a proper relay. But as I said, feel free to make something better.
Adding an intermediate hop is always higher latency, and ideal != what commonly happens
Encryption also increases event sizes (base64 encoding)