diff --git a/CHANGELOG.mdown b/CHANGELOG.mdown
index a7989125..4b574105 100644
--- a/CHANGELOG.mdown
+++ b/CHANGELOG.mdown
@@ -1,3 +1,10 @@
+### 1.19.0 (2016-04-12)
+
+  * Break: StreamHandler will not close streams automatically that it does not own. If you pass in a stream (not a path/url), then it will not close it for you. You can retrieve those using getStream() if needed
+  * Added DeduplicationHandler to remove duplicate records from notifications across multiple requests, useful for email or other notifications on errors
+  * Added ability to use `%message%` and other LineFormatter replacements in the subject line of emails sent with NativeMailHandler and SwiftMailerHandler
+  * Fixed HipChatHandler handling of long messages
+
 ### 1.18.2 (2016-04-02)
 
   * Fixed ElasticaFormatter to use more precise dates
diff --git a/doc/02-handlers-formatters-processors.md b/doc/02-handlers-formatters-processors.md
index 95ba8e53..9170d6df 100644
--- a/doc/02-handlers-formatters-processors.md
+++ b/doc/02-handlers-formatters-processors.md
@@ -85,6 +85,16 @@
   when it happens you will have the full information, including debug and info
   records. This provides you with all the information you need, but only when
   you need it.
+- _DeduplicationHandler_: Useful if you are sending notifications or emails
+  when critical errors occur. It takes a logger as parameter and will
+  accumulate log records of all levels until the end of the request (or
+  `flush()` is called). At that point it delivers all records to the handler
+  it wraps, but only if the records are unique over a given time period
+  (60seconds by default). If the records are duplicates they are simply
+  discarded. The main use of this is in case of critical failure like if your
+  database is unreachable for example all your requests will fail and that
+  can result in a lot of notifications being sent. Adding this handler reduces
+  the amount of notifications to a manageable level.
 - _WhatFailureGroupHandler_: This handler extends the _GroupHandler_ ignoring
    exceptions raised by each child handler. This allows you to ignore issues
    where a remote tcp connection may have died but you do not want your entire