AFAICT, we have two options:
-
Move to maintained bindings, such as https://github.com/zaidoon1/rust-rocksdb/. I was hesitant to do this prior, for the obvious supply-chain security reasons, but the adoption of cargo vet would likely solve that now?
-
Remove our usage of RocksDB.
While I unfortunately don't believe sled became production-ready, fjall may be. While I'd generally be incredibly cautious to the point of not even suggesting it, their 3.0 announcement was honestly just incredibly convincing they knew what they were doing and this could be considered.
There is also heed wrapping LMDB as another option.
librocksdb-sys (presumably embedding the entirety of RocksDB) also has the downside of being the largest single contributor to our cargo vet backlog at almost 900,000 lines of code.
AFAICT, we have two options:
Move to maintained bindings, such as https://github.com/zaidoon1/rust-rocksdb/. I was hesitant to do this prior, for the obvious supply-chain security reasons, but the adoption of
cargo vetwould likely solve that now?Remove our usage of RocksDB.
While I unfortunately don't believe
sledbecame production-ready,fjallmay be. While I'd generally be incredibly cautious to the point of not even suggesting it, their 3.0 announcement was honestly just incredibly convincing they knew what they were doing and this could be considered.There is also
heedwrapping LMDB as another option.librocksdb-sys(presumably embedding the entirety of RocksDB) also has the downside of being the largest single contributor to ourcargo vetbacklog at almost 900,000 lines of code.