// writing
Engineering notes from the crate.
Posts about Zstandard dictionaries, how to measure compression ratios that actually hold up, and the structural quirks of Solana data we exploit.
-
Why generic compressors leave bytes on the table for chain data
Vanilla zstd is excellent at what it does. What it does is rediscover common byte patterns on every compress call. Here is why that matters for Solana data, and what a hardcoded dictionary actually buys you.
engineeringcompression -
Measuring compression: ratios that matter for archive node operators
The headline 60:1 in our README is the top of one preset's range on archival data. Here is how we measure compression ratios honestly, why the lower bound matters more than the upper, and how to benchmark on your own workload.
benchmarksoperations -
Reading the algorithm: how we exploit chain-specific structure
A walk through the actual source of the Solana preset — the dictionary patterns, the Zstd construction, the preset-to-level mapping. Nothing magical, mostly a small constant table feeding a well-tested encoder.
architectureengineering