Changes

Changes 0.9.6.12 -> 0.9.6.13
 * Changed license to a permissable zlib/png-style

Changes 0.9.6.11 -> 0.9.6.12
 * NPE when server sent unmatched hop-by-hop-identifier (e.g. due to server bug or client timeout)
 
Changes 0.9.6.10 -> 0.9.6.11
 * Handle 3GPP mismatch between announced applications in CER/CEA and actual
   messages.

Changes 0.9.6.9 -> 0.9.6.10
 * Allow bogus PRNG when initializing DWR jitter
 * Eliminate (harmless) race condition when shutting down under high load
   leading to "WARNING: Got a DPA. This is not expected"
 * Mismatch between relative and absolute timeouts in request times. (should
   be relative)

Changes 0.9.6.8 -> 0.9.6.9
 * Fixed IllegalMonitorStateException when shutting down

Changes 0.9.6.7 -> 0.9.6.8
 * Added timeout capability to SimpleSyncClient and NodeManager
 * Logs destination IP address before initiating outgoing connections
 * Prevent double-start of stack (would lead to erratic behaviour)
 * Support TCP source port range

Changes 0.9.6.6 -> 0.9.6.7
* Expanded CC example server
* Fixed npe on outgoing sctp connections that fail
* Support replaced logger implementation, eg. in resin and other application
  servers.

Changes 0.9.6.5 -> 0.9.6.6
* AVP_Time.queryDate() returned wrong values de to 32-bit arithmic.

Changes 0.9.6.4 -> 0.9.6.5
* NodeManager could not find matching oustanding requests if the state object
  was null.
* Node.waitForConnectionTimeout() and NodeManager.waitForConnectionTimeout()
  methods added.
* Handle more than 2^31 sessions connectly when generating Session-IDs
* Support some unusual vendor-specific-application-id AVP formats.
* AVP_Time.queryDate() returned wrong value (Date object creation with seconds
  instead of milliseconds)
* AVP.queryPayload() made public.

Changes 0.9.6.3 -> 0.9.6.4
* Fix regression in 0.9.6.3:
  - ConcurrentModificationException when talking to silent peer
  - Bugfix when connect() connects imediately. (solaris)
  Both regressions were due to my fat fingers and subversion happily
  accepting changes to tags.

Changes 0.9.6.? -> 0.9.6.3
* Made more robust when calling Node.findConnection/sendMessage/... when the
  node has not been started.

Changes 0.9.6.1 -> 0.9.6.2
* If a peer accepted the connection, but never replied to CER, then a
  ConcurrentModificationException was thrown when the connection timed out.
* Local connections on Solaris lead to corner case where NIO behaves
  "interestingly" and the stack became non-functional.

Changes 0.9.6 -> 0.9.6.1
* File descriptor leak when a persistent peer had a name that could not be
  looked up in DNS.
* DCC examples no longer announce credit-control application as a supported
  accounting application.

Changes 0.9.5 -> 0.9.6
* Handle SCTP association restart correctly.
* Node.findConnection() could return connection keys for connections there were
  not in the "open" state, potentially leading to NPE if a client immediately
  tried using it.
* Logging in Node changed a bit to make failure to load SCTP support look less
  dangerous.
* Do not attempt to send DWR on non-ready connections.
* Timeout&close non-ready connections.

Changes 0.9.4 -> 0.9.5
* SCTP support
* CER election phase could easily fail.
* Improved handling of M-bit in grouped AVPs in dk.i1.diameter.Utils
* Implemented DW interval jitter as per rfc3539 section 3.4.1 item 1

Changes 0.9.4 -> 0.9.4.1
* M-bit was not set on AVPs inside Vendor-Specific-Application-Id

Changes 0.9.3 -> 0.9.4
* Examines result-code in CEA and acts on it.
* Error-bit was not set in error responses for protocol errors.
* Vendor-specific accounting applications in the CE messages were incorrectly
  sent as vendor-specific authentication applications.
* Log vendor-specific applications during capability negotiation.
* Implemented peer authentication (see NodeValidator interface).
* CER with election-lost was missing the Result-Code AVP.

Changes 0.9.2 -> 0.9.3
* On shutdown peers were sent a DPR with disconnect-cause='busy' instead of the
  (correct) 'rebooting' cause.
* It is now possible to have a (small) grace period for making a clean
  shutdown of connections waiting for reception of DPA. This has mostly
  academic interest. The default is to send DPR and close immediately.
* An CEA was erroneously sent as response to an unhandled CEA
* Idle timeout changed from 1 hour to 7 days.
* NodeSettings.???watchdoginterval() and NodeSettings.???idleTimeout() added.
* NPE when new inbound connection lost election
* Support for connecting to relays
* Connecting to peers on non-standard ports resulted in not being able to
  send messages to them due to port mismatch when looking up the peer.
* Added note about restoring hop-by-hop-identifier when forwarding answers
  with NodeManager.forwardAnswer()
* Added example relay to show per-message state handling in NodeManager.
* Message.prepareAnswer() and Messageheader.prepareAnswer() was added. They
  are identical to prepareResponse() but more appropriately named.

Changes 0.9.1 -> 0.9.2
* Reset TCP connection when non-diameter traffic is encountered (RFC3588
  section 8). Previously, the connection was simply closed.
* waitforConnection() was not working right.
* Moved session-id generation to Node
* Added credit-control example
* Raw packets were not logged with log-level=FINEST
* AVP.setM() method added
* Shutting down a node could in some cases result in a NullPointerException
  in "Diameter node reconnect thread"

Changes 0.9.0 -> 0.9.1
* Node.waitForConnection() and NodeManager.waitForConnection() added.
* AVP.vendor_id was not being reset on reused AVPs when calling decode()
* Made dk.i1.diameter.Messageheader constructors public
* With certain malformed origin-host-id values from peers the Node thread
  would encounter a NullPointerException because java.net.URI could not
  parse the URI.
* Potential race condition fix with sending messages to a peer after it
  was marked ready but before a CEA was sent.
* Faster detection of non-diameter traffic on sockets
* CER, DWR, DPR always had the hop-by-hop-identifier set to 0.