Release Notes: This version adds support for degree 1 isolation (dirty reads), automated maintenance of secondary indices on tables, native two-phase commit for applications that don't want to use a transaction monitor for distributed transaction processing, and the ability to fetch multiple rows in a single call to the Berkeley DB API (bulk return). This version can survive transient disk space exhaustion without a restart. It adds support for Lineo's Embedix platform.
Release Notes: The first 5.3 release came out a few month ago. One of the main highlights was support for the Android platform. 5.3.21 is the second release. This is mainly a bugfix release, but it does include one feature which is potentially very interesting, support for Oracle JDK7.
Release Notes: New features include an online backup command for SQL, VACUUM (which does not require the database to be quiescent), encryption, BFILE, ADO.NET, iOS/iPhone/iPad support (in addition to existing support for Android), a new command line tool called "db_replicate" which will turn non-HA applications into HA/replicated applications with no new code, performance improvements, and lots of other features.
Release Notes: Increased Performance: improved concurrency for multi-threaded applications on SMP systems. A new db_sql tool: automatic generation of Berkeley DB application code from SQL for faster development and migration. New APIs: support for C#/.NET and the C++ Standard Template Library (STL).
Release Notes: The lock manager may now be partitioned, improving performance on some SMP and CMP systems. Optional pre-zero log files. Replications across different endian CPUs. Replication can manage the master node via leases. The Java/JNI API now supports the Direct Persistence Layer (DPL). failchk() and other methods related to post-process failure database cleanup have been improved.
Release Notes: Non-stop upgrades for replicated systems: customers can upgrade/patch the database without taking the entire system down. Multi-version concurrency control (MVCC) allows read transactions to proceed without being blocked by write transactions, therefore improving concurrency and performance. Replication framework is a pre-built, out of the box framework that makes it faster and easier for developers to build replicated apps.