The Unreasonable Effectiveness of Simple Binary Encoding (SBE)

yafetn.github.io

8 points·yafetn··10 comments

It's always worth noting that JSON compresses the hell out of repetitive key names, just as it did for the even more repetitive XML. It's nicer to have a bit packing format that means you get even better space usage, and you skip a compression step, but it works for a huge number of cases.

This reminds me a lot of protobuf, just packing serial data based on type, specified width combined with a specified length (like an array) when 9-n items are possible.

I've used SBE and Aeron, can recommend.

The difference with SBE vs JSON is there's no need to serialize/deserialize anything. In SBE you literally just write in the value and read the value. You aren't turning bytes into text, which takes time and might require intermediate state (depends on what you're implementing).

It's almost like a non-format, because what could be simpler?

> It's almost like a non-format, because what could be simpler?

Something that specifies endianness and field types?

Yeah I'm not saying there's nothing there at all, just that it's minimal.

This isn't what "unreasonable effectiveness" is. Why worry about AI beliefs being disconnected from source meaning. Humans already abuse language.

What is unreasonable here?

Yeah. The title is clickbait and the post describes what every other binary protocol used to be like before JSON took over.

Honestly I can't see what the point of the post is. If you create a protocol that is optimized for bytes, you will get a protocol that is optimized for bytes. It's a trivial tautology.

Clickbait article, save your time. Could be generated by ChatGPT.

I'm so over these Unreasonable titles

Reply on news.ycombinator.com