diff --git a/Changelog.md b/Changelog.md
index b7d2363..63828da 100644
--- a/Changelog.md
+++ b/Changelog.md
@@ -4,6 +4,10 @@
 * Adds `types::Value` for dynamic column types.
 * Adds support for user-defined aggregate functions (behind the existing `functions` Cargo feature).
 * Introduces a `RowIndex` trait allowing columns to be fetched via index (as before) or name (new).
+* Introduces `ZeroBlob` type under the `blob` module/feature exposing SQLite's zeroblob API.
+* Adds CI testing for Windows via AppVeyor.
+* Fixes a warning building libsqlite3-sys under Rust 1.6.
+* Adds an unsafe `handle()` method to `Connection`. Please file an issue if you actually use it.
 
 # Version 0.6.0 (2015-12-17)
 
diff --git a/README.md b/README.md
index a813174..05635c7 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,7 @@
 # Rusqlite
 
-[![Build Status](https://api.travis-ci.org/jgallagher/rusqlite.svg?branch=master)](https://travis-ci.org/jgallagher/rusqlite)
+[![Travis Build Status](https://api.travis-ci.org/jgallagher/rusqlite.svg?branch=master)](https://travis-ci.org/jgallagher/rusqlite)
+[![AppVeyor Build Status](https://ci.appveyor.com/api/projects/status/github/jgallagher/rusqlite?branch=master&svg=true)](https://ci.appveyor.com/project/jgallagher/rusqlite)
 
 Rusqlite is an ergonomic wrapper for using SQLite from Rust. It attempts to expose
 an interface similar to [rust-postgres](https://github.com/sfackler/rust-postgres). View the full
@@ -72,45 +73,6 @@ features](http://doc.crates.io/manifest.html#the-features-section). They are:
 * [`blob`](http://jgallagher.github.io/rusqlite/rusqlite/blob/index.html)
   gives `std::io::{Read, Write, Seek}` access to SQL BLOBs.
 
-### Design of Rows and Row
-
-To retrieve the result rows from a query, SQLite requires you to call
-[sqlite3_step()](https://www.sqlite.org/c3ref/step.html) on a prepared statement. You can only
-retrieve the values of the "current" row. From the Rust point of view, this means that each row
-is only valid until the next row is fetched.  [rust-sqlite3](https://github.com/dckc/rust-sqlite3)
-solves this the correct way with lifetimes.  However, this means that the result rows do not
-satisfy the [Iterator](http://doc.rust-lang.org/std/iter/trait.Iterator.html) trait, which means
-you cannot (as easily) loop over the rows, or use many of the helpful Iterator methods like `map`
-and `filter`.
-
-Instead, Rusqlite's `Rows` handle does conform to `Iterator`. It ensures safety by
-performing checks at runtime to ensure you do not try to retrieve the values of a "stale" row, and
-will panic if you do so. A specific example that will panic:
-
-```rust
-fn bad_function_will_panic(conn: &Connection) -> Result<i64> {
-    let mut stmt = try!(conn.prepare("SELECT id FROM my_table"));
-    let mut rows = try!(stmt.query(&[]));
-
-    let row0 = try!(rows.next().unwrap());
-    // row 0 is valid now...
-
-    let row1 = try!(rows.next().unwrap());
-    // row 0 is now STALE, and row 1 is valid
-
-    let my_id = row0.get(0); // WILL PANIC because row 0 is stale
-    Ok(my_id)
-}
-```
-
-There are other, less obvious things that may result in a panic as well, such as calling
-`collect()` on a `Rows` and then trying to use the collected rows.
-
-Strongly consider using the method `query_map()` instead, if you can.
-`query_map()` returns an iterator over rows-mapped-to-some-type. This
-iterator does not have any of the above issues with panics due to attempting to
-access stale rows.
-
 ## Author
 
 John Gallagher, johnkgallagher@gmail.com
diff --git a/src/lib.rs b/src/lib.rs
index d15b0e0..f81926b 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -496,6 +496,18 @@ impl Connection {
         self.db.borrow_mut().load_extension(dylib_path.as_ref(), entry_point)
     }
 
+    /// Get access to the underlying SQLite database connection handle.
+    ///
+    /// # Warning
+    ///
+    /// You should not need to use this function. If you do need to, please [open an issue
+    /// on the rusqlite repository](https://github.com/jgallagher/rusqlite/issues) and describe
+    /// your use case. This function is unsafe because it gives you raw access to the SQLite
+    /// connection, and what you do with it could impact the safety of this `Connection`.
+    pub unsafe fn handle(&self) -> *mut ffi::Struct_sqlite3 {
+        self.db.borrow().db()
+    }
+
     fn decode_result(&self, code: c_int) -> Result<()> {
         self.db.borrow_mut().decode_result(code)
     }
@@ -977,6 +989,9 @@ pub type SqliteRows<'stmt> = Rows<'stmt>;
 ///
 /// ## Warning
 ///
+/// Strongly consider using `query_map` or `query_and_then` instead of `query`; the former do not
+/// suffer from the following problem.
+///
 /// Due to the way SQLite returns result rows of a query, it is not safe to attempt to get values
 /// from a row after it has become stale (i.e., `next()` has been called again on the `Rows`
 /// iterator). For example:
@@ -988,7 +1003,7 @@ pub type SqliteRows<'stmt> = Rows<'stmt>;
 ///     let mut rows = try!(stmt.query(&[]));
 ///
 ///     let row0 = try!(rows.next().unwrap());
-///     // row 0 is value now...
+///     // row 0 is valid for now...
 ///
 ///     let row1 = try!(rows.next().unwrap());
 ///     // row 0 is now STALE, and row 1 is valid
@@ -1002,12 +1017,6 @@ pub type SqliteRows<'stmt> = Rows<'stmt>;
 /// (which would result in a collection of rows, only the last of which can safely be used) and
 /// `min`/`max` (which could return a stale row unless the last row happened to be the min or max,
 /// respectively).
-///
-/// This problem could be solved by changing the signature of `next` to tie the lifetime of the
-/// returned row to the lifetime of (a mutable reference to) the result rows handle, but this would
-/// no longer implement `Iterator`, and therefore you would lose access to the majority of
-/// functions which are useful (such as support for `for ... in ...` looping, `map`, `filter`,
-/// etc.).
 pub struct Rows<'stmt> {
     stmt: &'stmt Statement<'stmt>,
     current_row: Rc<Cell<c_int>>,