0.7.0-beta1
 * sstable versioning (CASSANDRA-389)
 * switched to slf4j logging (CASSANDRA-625)
 * access levels for authentication/authorization (CASSANDRA-900)
 * add ReadRepairChance to CF definition (CASSANDRA-930)
 * fix heisenbug in system tests, especially common on OS X (CASSANDRA-944)
 * convert to byte[] keys internally and all public APIs (CASSANDRA-767)
 * ability to alter schema definitions on a live cluster (CASSANDRA-44)
 * renamed configuration file to cassandra.xml, and log4j.properties to
   log4j-server.properties, which must now be loaded from
   the classpath (which is how our scripts in bin/ have always done it)
   (CASSANDRA-971)
 * change get_count to require a SlicePredicate. create multi_get_count
   (CASSANDRA-744)
 * re-organized endpointsnitch implementations and added SimpleSnitch
   (CASSANDRA-994)
 * Added preload_row_cache option (CASSANDRA-946)
 * add CRC to commitlog header (CASSANDRA-999)
 * removed multiget thrift method (CASSANDRA-739)
 * removed deprecated batch_insert and get_range_slice methods (CASSANDRA-1065)
 * add truncate thrift method (CASSANDRA-531)
 * http mini-interface using mx4j (CASSANDRA-1068)
 * optimize away copy of sliced row on memtable read path (CASSANDRA-1046)
 * replace constant-size 2GB mmaped segments and special casing for index 
   entries spanning segment boundaries, with SegmentedFile that computes 
   segments that always contain entire entries/rows (CASSANDRA-1117)
 * avoid reading large rows into memory during compaction (CASSANDRA-16)
 * added hadoop OutputFormat (CASSANDRA-1101)
 * efficient Streaming (no more anticompaction) (CASSANDRA-579)
 * split commitlog header into separate file and add size checksum to
   mutations (CASSANDRA-1179)
 * avoid allocating a new byte[] for each mutation on replay (CASSANDRA-1219)
 * revise HH schema to be per-endpoint (CASSANDRA-1142)
 * add joining/leaving status to nodetool ring (CASSANDRA-1115)
 * allow multiple repair sessions per node (CASSANDRA-1190)
 * add dynamic endpoint snitch (CASSANDRA-981)
 * optimize away MessagingService for local range queries (CASSANDRA-1261)
 * make framed transport the default so malformed requests can't OOM the 
   server (CASSANDRA-475)
 * significantly faster reads from row cache (CASSANDRA-1267)
 * take advantage of row cache during range queries (CASSANDRA-1302)
 * make GCGraceSeconds a per-ColumnFamily value (CASSANDRA-1276)
 * keep persistent row size and column count statistics (CASSANDRA-1155)
 * add IntegerType (CASSANDRA-1282)
 * page within a single row during hinted handoff (CASSANDRA-1327)
 * push DatacenterShardStrategy configuration into keyspace definition,
   eliminating datacenter.properties. (CASSANDRA-1066)
 * optimize forward slices starting with '' and single-index-block name 
   queries by skipping the column index (CASSANDRA-1338)
 * streaming refactor (CASSANDRA-1189)


0.6.5
 * fix key ordering in range query results with RandomPartitioner
   and ConsistencyLevel > ONE (CASSANDRA-1145)
 * fix for range query starting with the wrong token range (CASSANDRA-1042)
 * page within a single row during hinted handoff (CASSANDRA-1327)
 * fix compilation on non-sun JKDs (CASSANDRA-1061)
 * remove String.trim() call on row keys in batch mutations (CASSANDRA-1235)
 * Log summary of dropped messages instead of spamming log (CASSANDRA-1284)


0.6.4
 * avoid queuing multiple hint deliveries for the same endpoint
   (CASSANDRA-1229)
 * better performance for and stricter checking of UTF8 column names
   (CASSANDRA-1232)
 * extend option to lower compaction priority to hinted handoff
   as well (CASSANDRA-1260)
 * log errors in gossip instead of re-throwing (CASSANDRA-1289)
 * avoid aborting commitlog replay prematurely if a flushed-but-
   not-removed commitlog segment is encountered (CASSANDRA-1297)
 * fix duplicate rows being read during mapreduce (CASSANDRA-1142)
 * failure detection wasn't closing command sockets (CASSANDRA-1221)
 * cassandra-cli.bat works on windows (CASSANDRA-1236)
 * pre-emptively drop requests that cannot be processed within RPCTimeout
   (CASSANDRA-685)
 * add ack to Binary write verb and update CassandraBulkLoader
   to wait for acks for each row (CASSANDRA-1093)
 * added describe_partitioner Thrift method (CASSANDRA-1047)
 * Hadoop jobs no longer require the Cassandra storage-conf.xml
   (CASSANDRA-1280, CASSANDRA-1047)
 * log thread pool stats when GC is excessive (CASSANDRA-1275)
 * remove gossip message size limit (CASSANDRA-1138)
 * parallelize local and remote reads during multiget, and respect snitch 
   when determining whether to do local read for CL.ONE (CASSANDRA-1317)
 * fix read repair to use requested consistency level on digest mismatch,
   rather than assuming QUORUM (CASSANDRA-1316)
 * process digest mismatch re-reads in parallel (CASSANDRA-1323)
 * switch hints CF comparator to BytesType (CASSANDRA-1274)


0.6.3
 * retry to make streaming connections up to 8 times. (CASSANDRA-1019)
 * reject describe_ring() calls on invalid keyspaces (CASSANDRA-1111)
 * don't reject reads at CL.ALL (CASSANDRA-1152)
 * reject deletions to supercolumns in CFs containing only standard
   columns (CASSANDRA-1139)
 * prefer sun jdk to openjdk in debian init script (CASSANDRA-1174)
 * detect partioner config changes between restarts and fail fast 
   (CASSANDRA-1146)
 * use generation time to resolve node token reassignment disagreements
   (CASSANDRA-1118)
 * restructure the startup ordering of Gossiper and MessageService to avoid
   timing anomalies (CASSANDRA-1160)
 * detect incomplete commit log hearders (CASSANDRA-1119)
 * force anti-entropy service to stream files on the stream stage to avoid
   sending streams out of order (CASSANDRA-1169)
 * remove inactive stream managers after AES streams files (CASSANDRA-1169)
 * add JMX metrics for row-level bloom filter false positives (CASSANDRA-1212)
 * added a redhat init script to contrib (CASSANDRA-1201)
 * use midpoint when bootstrapping a new machine into range with not
   much data yet instead of random token (CASSANDRA-1112)
 * kill server on OOM in executor stage as well as Thrift (CASSANDRA-1226)
 * remove opportunistic repairs, when two machines with overlapping replica
   responsibilities happen to finish major compactions of the same CF near
   the same time.  repairs are now fully manual (CASSANDRA-1190)
 * add ability to lower compaction priority (default is no change from 0.6.2)
   (CASSANDRA-1181)


0.6.2
 * fix contrib/word_count build. (CASSANDRA-992)
 * split CommitLogExecutorService into BatchCommitLogExecutorService and 
   PeriodicCommitLogExecutorService (CASSANDRA-1014)
 * add latency histograms to CFSMBean (CASSANDRA-1024)
 * make resolving timestamp ties deterministic by using value bytes
   as a tiebreaker (CASSANDRA-1039)
 * Add option to turn off Hinted Handoff (CASSANDRA-894)
 * fix windows startup (CASSANDRA-948)
 * make concurrent_reads, concurrent_writes configurable at runtime via JMX
   (CASSANDRA-1060)
 * disable GCInspector on non-Sun JVMs (CASSANDRA-1061)
 * fix tombstone handling in sstable rows with no other data (CASSANDRA-1063)
 * fix size of row in spanned index entries (CASSANDRA-1056)
 * install json2sstable, sstable2json, and sstablekeys to Debian package
 * StreamingService.StreamDestinations wouldn't empty itself after streaming
   finished (CASSANDRA-1076)
 * added Collections.shuffle(splits) before returning the splits in 
   ColumnFamilyInputFormat (CASSANDRA-1096)
 * do not recalculate cache capacity post-compaction if it's been manually 
   modified (CASSANDRA-1079)
 * better defaults for flush sorter + writer executor queue sizes
   (CASSANDRA-1100)
 * windows scripts for SSTableImport/Export (CASSANDRA-1051)
 * windows script for nodetool (CASSANDRA-1113)
 * make repair of RF==1 a no-op (CASSANDRA-1090)
 * improve default JVM GC options (CASSANDRA-1014)
 * fix SlicePredicate serialization inside Hadoop jobs (CASSANDRA-1049)
 * close Thrift sockets in Hadoop ColumnFamilyRecordReader (CASSANDRA-1081)


0.6.1
 * fix NPE in sstable2json when no excluded keys are given (CASSANDRA-934)
 * keep the replica set constant throughout the read repair process
   (CASSANDRA-937)
 * allow querying getAllRanges with empty token list (CASSANDRA-933)
 * fix command line arguments inversion in clustertool (CASSANDRA-942)
 * fix race condition that could trigger a false-positive assertion
   during post-flush discard of old commitlog segments (CASSANDRA-936)
 * fix neighbor calculation for anti-entropy repair (CASSANDRA-924)
 * perform repair even for small entropy differences (CASSANDRA-924)
 * Use hostnames in CFInputFormat to allow Hadoop's naive string-based
   locality comparisons to work (CASSANDRA-955)
 * nodes with IPv6 (and no IPv4) addresses could not join cluster
   (CASSANDRA-969)
 * Retrieve the correct number of undeleted columns, if any, from
   a supercolumn in a row that had been deleted previously (CASSANDRA-920)
 * fix index scans that cross the 2GB mmap boundaries for both mmap
   and standard i/o modes (CASSANDRA-866)
 * expose drain via nodetool (CASSANDRA-978)


0.6.0-RC1
 * JMX drain to flush memtables and run through commit log (CASSANDRA-880)
 * Bootstrapping can skip ranges under the right conditions (CASSANDRA-902)
 * fix merging row versions in range_slice for CL > ONE (CASSANDRA-884)
 * default write ConsistencyLeven chaned from ZERO to ONE
 * fix for index entries spanning mmap buffer boundaries (CASSANDRA-857)
 * use lexical comparison if time part of TimeUUIDs are the same 
   (CASSANDRA-907)
 * bound read, mutation, and response stages to fix possible OOM
   during log replay (CASSANDRA-885)
 * Use microseconds-since-epoch (UTC) in cli, instead of milliseconds
 * Treat batch_mutate Deletion with null supercolumn as "apply this predicate 
   to top level supercolumns" (CASSANDRA-834)
 * Streaming destination nodes do not update their JMX status (CASSANDRA-916)
 * Fix internal RPC timeout calculation (CASSANDRA-911)
 * Added Pig loadfunc to contrib/pig (CASSANDRA-910)


0.6.0-beta3
 * fix compaction bucketing bug (CASSANDRA-814)
 * update windows batch file (CASSANDRA-824)
 * deprecate KeysCachedFraction configuration directive in favor
   of KeysCached; move to unified-per-CF key cache (CASSANDRA-801)
 * add invalidateRowCache to ColumnFamilyStoreMBean (CASSANDRA-761)
 * send Handoff hints to natural locations to reduce load on
   remaining nodes in a failure scenario (CASSANDRA-822)
 * Add RowWarningThresholdInMB configuration option to warn before very 
   large rows get big enough to threaten node stability, and -x option to
   be able to remove them with sstable2json if the warning is unheeded
   until it's too late (CASSANDRA-843)
 * Add logging of GC activity (CASSANDRA-813)
 * fix ConcurrentModificationException in commitlog discard (CASSANDRA-853)
 * Fix hardcoded row count in Hadoop RecordReader (CASSANDRA-837)
 * Add a jmx status to the streaming service and change several DEBUG
   messages to INFO (CASSANDRA-845)
 * fix classpath in cassandra-cli.bat for Windows (CASSANDRA-858)
 * allow re-specifying host, port to cassandra-cli if invalid ones
   are first tried (CASSANDRA-867)
 * fix race condition handling rpc timeout in the coordinator
   (CASSANDRA-864)
 * Remove CalloutLocation and StagingFileDirectory from storage-conf files 
   since those settings are no longer used (CASSANDRA-878)
 * Parse a long from RowWarningThresholdInMB instead of an int (CASSANDRA-882)
 * Remove obsolete ControlPort code from DatabaseDescriptor (CASSANDRA-886)
 * move skipBytes side effect out of assert (CASSANDRA-899)
 * add "double getLoad" to StorageServiceMBean (CASSANDRA-898)
 * track row stats per CF at compaction time (CASSANDRA-870)
 * disallow CommitLogDirectory matching a DataFileDirectory (CASSANDRA-888)
 * default key cache size is 200k entries, changed from 10% (CASSANDRA-863)
 * add -Dcassandra-foreground=yes to cassandra.bat
 * exit if cluster name is changed unexpectedly (CASSANDRA-769)


0.6.0-beta1/beta2
 * add batch_mutate thrift command, deprecating batch_insert (CASSANDRA-336)
 * remove get_key_range Thrift API, deprecated in 0.5 (CASSANDRA-710)
 * add optional login() Thrift call for authentication (CASSANDRA-547)
 * support fat clients using gossiper and StorageProxy to perform
   replication in-process [jvm-only] (CASSANDRA-535)
 * support mmapped I/O for reads, on by default on 64bit JVMs 
   (CASSANDRA-408, CASSANDRA-669)
 * improve insert concurrency, particularly during Hinted Handoff
   (CASSANDRA-658)
 * faster network code (CASSANDRA-675)
 * stress.py moved to contrib (CASSANDRA-635)
 * row caching [must be explicitly enabled per-CF in config] (CASSANDRA-678)
 * present a useful measure of compaction progress in JMX (CASSANDRA-599)
 * add bin/sstablekeys (CASSNADRA-679)
 * add ConsistencyLevel.ANY (CASSANDRA-687)
 * make removetoken remove nodes from gossip entirely (CASSANDRA-644)
 * add ability to set cache sizes at runtime (CASSANDRA-708)
 * report latency and cache hit rate statistics with lifetime totals
   instead of average over the last minute (CASSANDRA-702)
 * support get_range_slice for RandomPartitioner (CASSANDRA-745)
 * per-keyspace replication factory and replication strategy (CASSANDRA-620)
 * track latency in microseconds (CASSANDRA-733)
 * add describe_ Thrift methods, deprecating get_string_property and 
   get_string_list_property
 * jmx interface for tracking operation mode and streams in general.
   (CASSANDRA-709)
 * keep memtables in sorted order to improve range query performance
   (CASSANDRA-799)
 * use while loop instead of recursion when trimming sstables compaction list 
   to avoid blowing stack in pathological cases (CASSANDRA-804)
 * basic Hadoop map/reduce support (CASSANDRA-342)


0.5.1
 * ensure all files for an sstable are streamed to the same directory.
   (CASSANDRA-716)
 * more accurate load estimate for bootstrapping (CASSANDRA-762)
 * tolerate dead or unavailable bootstrap target on write (CASSANDRA-731)
 * allow larger numbers of keys (> 140M) in a sstable bloom filter
   (CASSANDRA-790)
 * include jvm argument improvements from CASSANDRA-504 in debian package
 * change streaming chunk size to 32MB to accomodate Windows XP limitations
   (was 64MB) (CASSANDRA-795)
 * fix get_range_slice returning results in the wrong order (CASSANDRA-781)
 

0.5.0 final
 * avoid attempting to delete temporary bootstrap files twice (CASSANDRA-681)
 * fix bogus NaN in nodeprobe cfstats output (CASSANDRA-646)
 * provide a policy for dealing with single thread executors w/ a full queue
   (CASSANDRA-694)
 * optimize inner read in MessagingService, vastly improving multiple-node
   performance (CASSANDRA-675)
 * wait for table flush before streaming data back to a bootstrapping node.
   (CASSANDRA-696)
 * keep track of bootstrapping sources by table so that bootstrapping doesn't 
   give the indication of finishing early (CASSANDRA-673)


0.5.0 RC3
 * commit the correct version of the patch for CASSANDRA-663


0.5.0 RC2 (unreleased)
 * fix bugs in converting get_range_slice results to Thrift 
   (CASSANDRA-647, CASSANDRA-649)
 * expose java.util.concurrent.TimeoutException in StorageProxy methods
   (CASSANDRA-600)
 * TcpConnectionManager was holding on to disconnected connections, 
   giving the false indication they were being used. (CASSANDRA-651)
 * Remove duplicated write. (CASSANDRA-662)
 * Abort bootstrap if IP is already in the token ring (CASSANDRA-663)
 * increase default commitlog sync period, and wait for last sync to 
   finish before submitting another (CASSANDRA-668)


0.5.0 RC1
 * Fix potential NPE in get_range_slice (CASSANDRA-623)
 * add CRC32 to commitlog entries (CASSANDRA-605)
 * fix data streaming on windows (CASSANDRA-630)
 * GC compacted sstables after cleanup and compaction (CASSANDRA-621)
 * Speed up anti-entropy validation (CASSANDRA-629)
 * Fix anti-entropy assertion error (CASSANDRA-639)
 * Fix pending range conflicts when bootstapping or moving
   multiple nodes at once (CASSANDRA-603)
 * Handle obsolete gossip related to node movement in the case where
   one or more nodes is down when the movement occurs (CASSANDRA-572)
 * Include dead nodes in gossip to avoid a variety of problems
   and fix HH to removed nodes (CASSANDRA-634)
 * return an InvalidRequestException for mal-formed SlicePredicates
   (CASSANDRA-643)
 * fix bug determining closest neighbor for use in multiple datacenters
   (CASSANDRA-648)
 * Vast improvements in anticompaction speed (CASSANDRA-607)
 * Speed up log replay and writes by avoiding redundant serializations
   (CASSANDRA-652)


0.5.0 beta 2
 * Bootstrap improvements (several tickets)
 * add nodeprobe repair anti-entropy feature (CASSANDRA-193, CASSANDRA-520)
 * fix possibility of partition when many nodes restart at once
   in clusters with multiple seeds (CASSANDRA-150)
 * fix NPE in get_range_slice when no data is found (CASSANDRA-578)
 * fix potential NPE in hinted handoff (CASSANDRA-585)
 * fix cleanup of local "system" keyspace (CASSANDRA-576)
 * improve computation of cluster load balance (CASSANDRA-554)
 * added super column read/write, column count, and column/row delete to
   cassandra-cli (CASSANDRA-567, CASSANDRA-594)
 * fix returning live subcolumns of deleted supercolumns (CASSANDRA-583)
 * respect JAVA_HOME in bin/ scripts (several tickets)
 * add StorageService.initClient for fat clients on the JVM (CASSANDRA-535)
   (see contrib/client_only for an example of use)
 * make consistency_level functional in get_range_slice (CASSANDRA-568)
 * optimize key deserialization for RandomPartitioner (CASSANDRA-581)
 * avoid GCing tombstones except on major compaction (CASSANDRA-604)
 * increase failure conviction threshold, resulting in less nodes
   incorrectly (and temporarily) marked as down (CASSANDRA-610)
 * respect memtable thresholds during log replay (CASSANDRA-609)
 * support ConsistencyLevel.ALL on read (CASSANDRA-584)
 * add nodeprobe removetoken command (CASSANDRA-564)


0.5.0 beta
 * Allow multiple simultaneous flushes, improving flush throughput 
   on multicore systems (CASSANDRA-401)
 * Split up locks to improve write and read throughput on multicore systems
   (CASSANDRA-444, CASSANDRA-414)
 * More efficient use of memory during compaction (CASSANDRA-436)
 * autobootstrap option: when enabled, all non-seed nodes will attempt
   to bootstrap when started, until bootstrap successfully
   completes. -b option is removed.  (CASSANDRA-438)
 * Unless a token is manually specified in the configuration xml,
   a bootstraping node will use a token that gives it half the
   keys from the most-heavily-loaded node in the cluster,
   instead of generating a random token. 
   (CASSANDRA-385, CASSANDRA-517)
 * Miscellaneous bootstrap fixes (several tickets)
 * Ability to change a node's token even after it has data on it
   (CASSANDRA-541)
 * Ability to decommission a live node from the ring (CASSANDRA-435)
 * Semi-automatic loadbalancing via nodeprobe (CASSANDRA-192)
 * Add ability to set compaction thresholds at runtime via
   JMX / nodeprobe.  (CASSANDRA-465)
 * Add "comment" field to ColumnFamily definition. (CASSANDRA-481)
 * Additional JMX metrics (CASSANDRA-482)
 * JSON based export and import tools (several tickets)
 * Hinted Handoff fixes (several tickets)
 * Add key cache to improve read performance (CASSANDRA-423)
 * Simplified construction of custom ReplicationStrategy classes
   (CASSANDRA-497)
 * Graphical application (Swing) for ring integrity verification and 
   visualization was added to contrib (CASSANDRA-252)
 * Add DCQUORUM, DCQUORUMSYNC consistency levels and corresponding
   ReplicationStrategy / EndpointSnitch classes.  Experimental.
   (CASSANDRA-492)
 * Web client interface added to contrib (CASSANDRA-457)
 * More-efficient flush for Random, CollatedOPP partitioners 
   for normal writes (CASSANDRA-446) and bulk load (CASSANDRA-420)
 * Add MemtableFlushAfterMinutes, a global replacement for the old 
   per-CF FlushPeriodInMinutes setting (CASSANDRA-463)
 * optimizations to slice reading (CASSANDRA-350) and supercolumn
   queries (CASSANDRA-510)
 * force binding to given listenaddress for nodes with multiple
   interfaces (CASSANDRA-546)
 * stress.py benchmarking tool improvements (several tickets)
 * optimized replica placement code (CASSANDRA-525)
 * faster log replay on restart (CASSANDRA-539, CASSANDRA-540)
 * optimized local-node writes (CASSANDRA-558)
 * added get_range_slice, deprecating get_key_range (CASSANDRA-344)
 * expose TimedOutException to thrift (CASSANDRA-563)
 

0.4.2
 * Add validation disallowing null keys (CASSANDRA-486)
 * Fix race conditions in TCPConnectionManager (CASSANDRA-487)
 * Fix using non-utf8-aware comparison as a sanity check.
   (CASSANDRA-493)
 * Improve default garbage collector options (CASSANDRA-504)
 * Add "nodeprobe flush" (CASSANDRA-505)
 * remove NotFoundException from get_slice throws list (CASSANDRA-518)
 * fix get (not get_slice) of entire supercolumn (CASSANDRA-508)
 * fix null token during bootstrap (CASSANDRA-501)


0.4.1
 * Fix FlushPeriod columnfamily configuration regression
   (CASSANDRA-455)
 * Fix long column name support (CASSANDRA-460)
 * Fix for serializing a row that only contains tombstones
   (CASSANDRA-458)
 * Fix for discarding unneeded commitlog segments (CASSANDRA-459)
 * Add SnapshotBeforeCompaction configuration option (CASSANDRA-426)
 * Fix compaction abort under insufficient disk space (CASSANDRA-473)
 * Fix reading subcolumn slice from tombstoned CF (CASSANDRA-484)
 * Fix race condition in RVH causing occasional NPE (CASSANDRA-478)


0.4.0
 * fix get_key_range problems when a node is down (CASSANDRA-440)
   and add UnavailableException to more Thrift methods
 * Add example EndPointSnitch contrib code (several tickets)


0.4.0 RC2
 * fix SSTable generation clash during compaction (CASSANDRA-418)
 * reject method calls with null parameters (CASSANDRA-308)
 * properly order ranges in nodeprobe output (CASSANDRA-421)
 * fix logging of certain errors on executor threads (CASSANDRA-425)


0.4.0 RC1
 * Bootstrap feature is live; use -b on startup (several tickets)
 * Added multiget api (CASSANDRA-70)
 * fix Deadlock with SelectorManager.doProcess and TcpConnection.write
   (CASSANDRA-392)
 * remove key cache b/c of concurrency bugs in third-party
   CLHM library (CASSANDRA-405)
 * update non-major compaction logic to use two threshold values
   (CASSANDRA-407)
 * add periodic / batch commitlog sync modes (several tickets)
 * inline BatchMutation into batch_insert params (CASSANDRA-403)
 * allow setting the logging level at runtime via mbean (CASSANDRA-402)
 * change default comparator to BytesType (CASSANDRA-400)
 * add forwards-compatible ConsistencyLevel parameter to get_key_range
   (CASSANDRA-322)
 * r/m special case of blocking for local destination when writing with 
   ConsistencyLevel.ZERO (CASSANDRA-399)
 * Fixes to make BinaryMemtable [bulk load interface] useful (CASSANDRA-337);
   see contrib/bmt_example for an example of using it.
 * More JMX properties added (several tickets)
 * Thrift changes (several tickets)
    - Merged _super get methods with the normal ones; return values
      are now of ColumnOrSuperColumn.
    - Similarly, merged batch_insert_super into batch_insert.



0.4.0 beta
 * On-disk data format has changed to allow billions of keys/rows per
   node instead of only millions
 * Multi-keyspace support
 * Scan all sstables for all queries to avoid situations where
   different types of operation on the same ColumnFamily could
   disagree on what data was present
 * Snapshot support via JMX
 * Thrift API has changed a _lot_:
    - removed time-sorted CFs; instead, user-defined comparators
      may be defined on the column names, which are now byte arrays.
      Default comparators are provided for UTF8, Bytes, Ascii, Long (i64),
      and UUID types.
    - removed colon-delimited strings in thrift api in favor of explicit
      structs such as ColumnPath, ColumnParent, etc.  Also normalized
      thrift struct and argument naming.
    - Added columnFamily argument to get_key_range.
    - Change signature of get_slice to accept starting and ending
      columns as well as an offset.  (This allows use of indexes.)
      Added "ascending" flag to allow reasonably-efficient reverse
      scans as well.  Removed get_slice_by_range as redundant.
    - get_key_range operates on one CF at a time
    - changed `block` boolean on insert methods to ConsistencyLevel enum,
      with options of NONE, ONE, QUORUM, and ALL.
    - added similar consistency_level parameter to read methods
    - column-name-set slice with no names given now returns zero columns
      instead of all of them.  ("all" can run your server out of memory.
      use a range-based slice with a high max column count instead.)
 * Removed the web interface. Node information can now be obtained by 
   using the newly introduced nodeprobe utility.
 * More JMX stats
 * Remove magic values from internals (e.g. special key to indicate
   when to flush memtables)
 * Rename configuration "table" to "keyspace"
 * Moved to crash-only design; no more shutdown (just kill the process)
 * Lots of bug fixes

Full list of issues resolved in 0.4 is at https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&&pid=12310865&fixfor=12313862&resolution=1&sorter/field=issuekey&sorter/order=DESC


0.3.0 RC3
 * Fix potential deadlock under load in TCPConnection.
   (CASSANDRA-220)


0.3.0 RC2
 * Fix possible data loss when server is stopped after replaying
   log but before new inserts force memtable flush.
   (CASSANDRA-204)
 * Added BUGS file


0.3.0 RC1
 * Range queries on keys, including user-defined key collation
 * Remove support
 * Workarounds for a weird bug in JDK select/register that seems
   particularly common on VM environments. Cassandra should deploy
   fine on EC2 now
 * Much improved infrastructure: the beginnings of a decent test suite
   ("ant test" for unit tests; "nosetests" for system tests), code
   coverage reporting, etc.
 * Expanded node status reporting via JMX
 * Improved error reporting/logging on both server and client
 * Reduced memory footprint in default configuration
 * Combined blocking and non-blocking versions of insert APIs
 * Added FlushPeriodInMinutes configuration parameter to force
   flushing of infrequently-updated ColumnFamilies