Goal
Reduce ledger execution-detail frame size by embedding the selected universal Zstd dictionary, omitting redundant per-frame dictionary IDs, and making the superblock format version the compatibility boundary.
Why
The ledger bitcode-encodes Option<ExecutionDetails> and compresses each value as an independent Zstd frame. The selected curated hybrid dictionary was trained and validated against a deterministic 60,000-sample corpus with identifier-churn scenarios to retain useful structural and runtime vocabulary without overfitting to incidental program IDs.
On the untouched random holdout, dictionary-backed level-3 Zstd reduced output from 16.39% to 10.59% of the uncompressed bitcode size, a 35.37% saving relative to level-3 Zstd without a dictionary. Savings remained 27.27% for future-20% approved-program traffic, 25.37% with zero approved-program traffic, and 30.28% with high-50% approved-program traffic. The universal score is 27.39%.
Scope
- Keep the selected 1,280-byte artifact at
assets/execution-details.dict with SHA-256 9f9f91c856f04c0899b3b5fb3174a1b020608f321d3a7d2bd42491dd310c4c9d.
- Embed it in the ledger crate with
include_bytes!() and use an explicit compression level of 3.
- Load the dictionary once when each reusable appender and reader Zstd context is initialized; preserve the existing bitcode layout and direct
compress_to_buffer file-buffer path.
- Disable the Zstd frame dictionary-ID flag and verify representative frames omit dictionary IDs while matching Zstd and bitcode contexts still round-trip.
- Store immutable ledger version
1 first in every superblock.meta and reject unsupported versions before opening or creating the superblock index and data files.
- Treat the dictionary and codec as part of the versioned ledger format. Any future dictionary or codec change requires a ledger-version bump and explicit compatibility handling; no dictionary-less fallback or migration is included.
Goal
Reduce ledger execution-detail frame size by embedding the selected universal Zstd dictionary, omitting redundant per-frame dictionary IDs, and making the superblock format version the compatibility boundary.
Why
The ledger bitcode-encodes
Option<ExecutionDetails>and compresses each value as an independent Zstd frame. The selected curated hybrid dictionary was trained and validated against a deterministic 60,000-sample corpus with identifier-churn scenarios to retain useful structural and runtime vocabulary without overfitting to incidental program IDs.On the untouched random holdout, dictionary-backed level-3 Zstd reduced output from 16.39% to 10.59% of the uncompressed bitcode size, a 35.37% saving relative to level-3 Zstd without a dictionary. Savings remained 27.27% for future-20% approved-program traffic, 25.37% with zero approved-program traffic, and 30.28% with high-50% approved-program traffic. The universal score is 27.39%.
Scope
assets/execution-details.dictwith SHA-2569f9f91c856f04c0899b3b5fb3174a1b020608f321d3a7d2bd42491dd310c4c9d.include_bytes!()and use an explicit compression level of 3.compress_to_bufferfile-buffer path.1first in everysuperblock.metaand reject unsupported versions before opening or creating the superblock index and data files.