- Added parallelization for node preloads during audit proof generation
- Removed
parallel_azksfeature in favor of explicit parallelism configuration object allowing the setting of static or dynamic parallelism across insertion and preloads
- Added parallelization for node preloads during insertion
- Added support for tracing
- Fixed a bug in key history API that can incorrectly return HistoryProofs without UpdateProofs.
- Fixed a bug in key history API that occurs when a proof is being generated while a transaction is being committed
- Updated test vectors to the latest version
- Added an efficiency optimization for history proof generation and verification (incompatible with previous logic)
- Fixing a preloads ordering bug
- Only instantiate metrics attribute in storage manager when runtime metrics is enabled. This avoids an unnecessary import of atomic types, which not all architectures support.
- Upgrades history proof generation to properly support the MostRecent parameter (giving a limited history proof)
- Eliminates a rare bug that can result in an aZKS being overwritten during Directory initialization
- Updated storage tombstone API params to be more ergonomic
- Renamed HistoryParams enum variants to highlight caveats
- Added NCC audit report to README
- Improved documentation
- Updated deploy script to support pre-release version strings
- Added error-handling for various edge-cases when performing akd_core verification
- Updated dependencies
- Updated VRF checks and public key validation
- Added duplicate entries check in publish
- Added zeroize on drop for VRFExpandedPrivateKey
- Reduced visibility of membership verification functions
- Added error checks for NodeLabel::try_from and get_bit_at functions
- Clarified dual-licensing wording
- Updated documentation and dependencies
- Renamed public-tests feature to public_tests
- Added a Configuration trait for specifying different cryptographic configurations of the tree
- Reorganized local_auditor, poc, akd_client, and others into an examples/ directory
- Added lookup verify check for version number exceeding epoch
- Added an opt-out optimization on batch-retrieval of nodes for lookup proof generation
- Contains optimizations for audit proof generation
- Adds support for parallelization of the audit proof generation along with disabling cache cleaning in audit proof generation
- More naming cleanups + reorganization
- Cleanup of unwrap()s in the codebase
- Use AzksValue as node value instead of Digest
- Preloading optimizations in history proofs
- Big naming cleanups
- Ability to mock the Database trait
- Fix in history proof validations
- Moving in-memory testing/audit database to DashMap for better parallel performance
- A auditor user-interface based on iced in Rust
- Code-coverage CI
- Avoid looking up previous version of nodes that are new
- Fixed bug in longest common prefix calculation
- Added sha512/256 support
- Updated Direction from Option to u8 enum
- Updated VRF label derivation and commitment generation
- Updated serialization to use big endian
- Optimized protobuf encoding
- Optimized VRF computations
- Optimized node preloading
- Optimized node insertion
- Parallelized calculations of VRF labels
- Parallelized node insertion
A slew of bug fixes (transactions not cleaning up, cache management, etc) Reductions in loggers Tests on WASM generation + proof verification Fixing some namespace clashes (on proto module name) General cleanups
- Remove defunct get_root_hash_at_epoch functionality
- Minimize the dependencies on tokio such that it minimizes imported crates into the dependency tree
- Major rewrite to remove all winter_* crates such that all verification functions and types are shared in a common akd_core crate
- Adapt best-practices for generated code in akd_core [for protobuf]
- Cleaned up documentation
- Updated history checks for validity
- Hasher generic moved from method calls to Directory struct
- Update VRF to latest specification
- VRF's enabled always
- API cleanups
- Large rewrite of storage management. Cache + transaction logic abstracted to common logic so database implementations are much more basic
- Protobuf serialization support for all over-the-wire types for efficient, backwards compatible encoding
- Leaning down of akd_client to reduce dependencies when feature flags not set
- Serialized proof validation in akd_client for embedded usages
- Independent auditor REPL for easy validation of proofs from a public storage account
- Breadth-first-search preloading for audit proofs
- Fix the cache concurrency problem where new node construction caused a database write-through and cache miss
- Fix handling of transactions with batch_set in the memory-based implementations such that caches weren't getting populated properly
- Added initial solution for public auditing from an S3 bucket (including backwards compatible protobuf serialization of the append-only proofs)
- oZKS construction removing the need for history node states, greatly reducing overall storage requirements
- Concurrent access (reading while writing) is supported, assuming atomic row operations
- Create an akd_test_tools crate to house test utilities that clients can import
- Check for sequential history in client key history proof validation
- Check values against last published value and skips duplicates in publish
- Migrate POC app from structopt to clap3
- Add tree fixture generator tool for tests
- Note: This release contains backward-incompatible changes to the Storage trait
- AKD dependency tree adjustments. No logical changes in the code
- A re-publish of 0.5.2
- "Lean" client supports key history proof validation
- Support tombstone's for expired value-states (GDPR style compliance)
- Bulk lookup proofs
- Cleanup of error handling patterns & flow
- Fixing the MySQL testing infra
- Verifiable random functions (VRFs) for node labels
- Better support for distributed environment (many read-nodes, single writer node)
- "Lean" client which is only dependent on the core hashing functionality. This is a new crate akd_client
- Full serialization of public structs
- Cleaning up risky unwrap() calls within the lib to minimize panic! risks
- Error and error code cleanups in the library
- Proof structs prefer vec of structs rather than paired vectors
- Fix CI pipeline publish
- Standardize AKD label & value partners
- Refactor storage interface
- Remove unbounded storage of epochs in HistoryTreeNode
- Deprecate the location field in HistoryTreeNode in favor of using the label directly to identify node position in storage
- POC db drop and bench-lookup operations
- Version bump
- Changed visibility of structs that are publicly exposed
- Upgrade winter_* to 0.2
- Other minor cleanups
- Pinning to tokio 1.10 instead of specifically 1.10.2
- Specific version is not required for the akd_mysql crate since major versions didn't change
- Require absolute URLs to reference information within crate documentation see
- Move publish pipeline to run checks on both packages before starting publish
- Versioning information changes & proper links for public release on crates.io
- Cleanup of DB layer and more performance improvements for MySQL storage.
- Additional documentation and general cleanups
- Removed MySQL dependencies on the core library functionality so if a user uses a different data-layer they are not dependent on MySQL
- Proof-of-concept (REPL) application + integration tests in CI pipeline
- Migration from tokio 0.2 to 1.X and mysql_async from 0.23.1 to 0.28.1
- Added more crate-level documentation
- Added a proof-of-concpt application that interacts with the storage layer
- Initial release