diff --git a/.gitignore b/.gitignore
index e44e8f90..a9d37c56 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,2 @@
 target
 Cargo.lock
-# Ignore the dumps for now until the format stabilizes
-assets/*
diff --git a/Cargo.toml b/Cargo.toml
index d0f8c165..88c7f421 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -13,6 +13,7 @@ bitflags = "^0.4"
 plist = "0.0.13"
 rustc-serialize = "0.3"
 bincode = "0.5"
+flate2 = "*"
 
 # [profile.release]
 # debug = true
diff --git a/Readme.md b/Readme.md
index 65b0d5b2..1e9096e4 100644
--- a/Readme.md
+++ b/Readme.md
@@ -6,13 +6,16 @@ If you are writing a text editor (or something else needing highlighting) in Rus
 
 It is currently mostly complete and can parse, interpret and highlight based on Sublime Text syntax and `tmTheme` files.
 
-## Goals
+## Features/Goals
 
-- Work with many languages (accomplished through using existing grammar formats)
-- Be super fast, both in terms of highlighting and startup time
-- API that is both easy to use, and allows use in fancy text editors with piece tables and incremental re-highlighting and the like
-- Expose internals of the parsing process so text editors can do things like cache parse states and use semantic info for code intelligence
-- High quality highlighting, supporting things like heredocs and complex syntaxes (like Rust's).
+- [x] Work with many languages (accomplished through using existing grammar formats)
+- [ ] Highlight super quickly, as fast as Sublime Text (not there yet but matching most editors)
+- [x] Load up quickly, currently in around 23ms but could potentially be even faster.
+- [ ] Include easy to use API for basic cases
+- [x] API allows use in fancy text editors with piece tables and incremental re-highlighting and the like.
+- [x] Expose internals of the parsing process so text editors can do things like cache parse states and use semantic info for code intelligence
+- [x] High quality highlighting, supporting things like heredocs and complex syntaxes (like Rust's).
+- [x] Include a compressed dump of all the default syntax definitions in the library binary so users don't have to manage a folder of syntaxes.
 
 ## Screenshots
 
@@ -30,7 +33,7 @@ There's currently an example program called `syncat` that prints one of the sour
 - [x] Parse TextMate/Sublime Text theme files
 - [x] Highlight a scope-annotated iterator into a colour-annotated iterator for display.
 - [x] Ability to dump loaded packages as binary file and load them with lazy regex compilation for fast start up times.
-- [ ] Bundle dumped default syntaxes into the library binary so library users don't need an assets folder with Sublime Text packages.
+- [x] Bundle dumped default syntaxes into the library binary so library users don't need an assets folder with Sublime Text packages.
 - [ ] Add nice API wrappers for simple use cases. The base APIs are designed for deep high performance integration with arbitrary text editor data structures.
 - [ ] Make syncat a better demo, and maybe more demo programs
 - [ ] Document the API better and make things private that don't need to be public
@@ -59,7 +62,7 @@ The current perf numbers are below. These numbers should get better once I imple
     - Vim is instantaneous but that isn't a fair comparison since vim's highlighting is far more basic than the other editors.
     - These comparisons aren't totally fair, except the one to Sublime Text since that is using the same theme and the same complex defintion for ES6 syntax.
 - ~220ms to load and link all the syntax definitions in the default Sublime package set. This is ~60% regex compilation and ~35% YAML parsing.
-    - but only ~16ms to load and link all the syntax definitions from a pre-made binary dump with lazy regex compilation.
+    - but only ~23ms to load and link all the syntax definitions from an internal pre-made binary dump with lazy regex compilation.
 - ~1.9ms to parse and highlight the 30 line 791 character `testdata/highlight_test.erb` file. This works out to around 16,000 lines/second or 422 kilobytes/second.
 - ~250ms end to end for `syncat` to start, load the definitions, highlight the test file and shut down. This is mostly spent loading.
 
diff --git a/assets/default_newlines.packdump b/assets/default_newlines.packdump
new file mode 100644
index 00000000..f201e244
Binary files /dev/null and b/assets/default_newlines.packdump differ
diff --git a/assets/default_nonewlines.packdump b/assets/default_nonewlines.packdump
new file mode 100644
index 00000000..a1b5e2f5
Binary files /dev/null and b/assets/default_nonewlines.packdump differ
diff --git a/benches/loading.rs b/benches/loading.rs
index 7dbfa732..9f58eed1 100644
--- a/benches/loading.rs
+++ b/benches/loading.rs
@@ -14,6 +14,14 @@ fn bench_load_syntax_dump(b: &mut Bencher) {
     });
 }
 
+#[bench]
+fn bench_load_internal_dump(b: &mut Bencher) {
+    b.iter(|| {
+        let ps = PackageSet::load_defaults_newlines();
+        test::black_box(&ps);
+    });
+}
+
 #[bench]
 fn bench_load_syntaxes(b: &mut Bencher) {
     b.iter(|| {
diff --git a/examples/syncat.rs b/examples/syncat.rs
index 2acb1856..7a4ab6ce 100644
--- a/examples/syncat.rs
+++ b/examples/syncat.rs
@@ -12,7 +12,7 @@ use std::path::Path;
 use std::fs::File;
 
 fn main() {
-    let ps = PackageSet::load_from_folder("testdata/Packages").unwrap();
+    let ps = PackageSet::load_defaults_nonewlines();
     let highlighter = Highlighter::new(PackageSet::get_theme("testdata/spacegray/base16-ocean.\
                                                               dark.tmTheme")
         .unwrap());
diff --git a/src/dumps.rs b/src/dumps.rs
index 1fa167ef..1539310b 100644
--- a/src/dumps.rs
+++ b/src/dumps.rs
@@ -1,25 +1,62 @@
 use bincode::SizeLimit;
 use bincode::rustc_serialize::*;
 use std::fs::File;
-use std::io::BufReader;
+use std::io::{BufReader, BufWriter};
 use package_set::PackageSet;
 use std::path::Path;
+use flate2::write::ZlibEncoder;
+use flate2::read::ZlibDecoder;
+use flate2::Compression;
 
 impl PackageSet {
+    /// Instantiates a new package set from a binary dump of
+    /// Sublime Text's default open source syntax definitions and then links it.
+    /// These dumps are included in this library's binary for convenience.
+    /// This method loads the version for parsing line strings with no `\n` characters at the end.
+    ///
+    /// This is the recommended way of creating a package set for
+    /// non-advanced use cases. It is also significantly faster than loading the YAML files.
+    ///
+    /// Note that you can load additional syntaxes after doing this,
+    /// you'll just have to link again. If you want you can even
+    /// use the fact that SyntaxDefinitions are serializable with
+    /// the bincode crate to cache dumps of additional syntaxes yourself.
+    pub fn load_defaults_nonewlines() -> PackageSet {
+        let mut ps = Self::from_binary(include_bytes!("../assets/default_nonewlines.packdump"));
+        ps.link_syntaxes();
+        ps
+    }
+
+    /// Same as `load_defaults_nonewlines` but for parsing line strings with newlines at the end.
+    /// These are separate methods because thanks to linker garbage collection, only the serialized
+    /// dumps for the method(s) you call will be included in the binary (each is ~200kb for now).
+    pub fn load_defaults_newlines() -> PackageSet {
+        let mut ps = Self::from_binary(include_bytes!("../assets/default_newlines.packdump"));
+        ps.link_syntaxes();
+        ps
+    }
+
     pub fn dump_binary(&self) -> Vec<u8> {
         assert!(!self.is_linked);
-        encode(self, SizeLimit::Infinite).unwrap()
+        let mut v = Vec::new();
+        {
+            let mut encoder = ZlibEncoder::new(&mut v, Compression::Best);
+            encode_into(self, &mut encoder, SizeLimit::Infinite).unwrap();
+        }
+        v
     }
 
     pub fn dump_to_file<P: AsRef<Path>>(&self, path: P) -> EncodingResult<()> {
-        let mut f = try!(File::create(path).map_err(EncodingError::IoError));
-        encode_into(self, &mut f, SizeLimit::Infinite)
+        let f = BufWriter::new(try!(File::create(path).map_err(EncodingError::IoError)));
+        let mut encoder = ZlibEncoder::new(f, Compression::Best);
+        encode_into(self, &mut encoder, SizeLimit::Infinite)
     }
 
     /// Returns a fully loaded and linked package set from
     /// a binary dump. Panics if the dump is invalid.
-    pub fn from_binary(v: Vec<u8>) -> PackageSet {
-        let mut ps: PackageSet = decode(&v[..]).unwrap();
+    pub fn from_binary(v: &[u8]) -> PackageSet {
+        let mut decoder = ZlibDecoder::new(v);
+        let mut ps: PackageSet = decode_from(&mut decoder, SizeLimit::Infinite).unwrap();
         ps.link_syntaxes();
         ps
     }
@@ -28,8 +65,8 @@ impl PackageSet {
     /// a binary dump file.
     pub fn from_dump_file<P: AsRef<Path>>(path: P) -> DecodingResult<PackageSet> {
         let f = try!(File::open(path).map_err(DecodingError::IoError));
-        let mut reader = BufReader::new(f);
-        decode_from(&mut reader, SizeLimit::Infinite)
+        let mut decoder = ZlibDecoder::new(BufReader::new(f));
+        decode_from(&mut decoder, SizeLimit::Infinite)
     }
 }
 
@@ -42,7 +79,7 @@ mod tests {
         ps.load_syntaxes("testdata/Packages", false).unwrap();
 
         let bin = ps.dump_binary();
-        let ps2 = PackageSet::from_binary(bin);
+        let ps2 = PackageSet::from_binary(&bin[..]);
         assert_eq!(ps.syntaxes.len(), ps2.syntaxes.len());
     }
 }
diff --git a/src/lib.rs b/src/lib.rs
index a2c28d86..da331455 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -9,6 +9,7 @@ extern crate bincode;
 extern crate rustc_serialize;
 #[macro_use]
 extern crate bitflags;
+extern crate flate2;
 pub mod syntax_definition;
 pub mod yaml_load;
 pub mod package_set;
diff --git a/src/package_set.rs b/src/package_set.rs
index 030209d8..8b59be3e 100644
--- a/src/package_set.rs
+++ b/src/package_set.rs
@@ -15,6 +15,13 @@ use std::mem;
 use std::rc::Rc;
 use std::ascii::AsciiExt;
 
+/// A package set holds a bunch of syntaxes and manages
+/// loading them and the crucial operation of *linking*.
+///
+/// Linking replaces the references between syntaxes with direct
+/// pointers. See `link_syntaxes` for more.
+/// Linking, followed by adding more unlinked syntaxes with `load_syntaxes`
+/// and then linking again is allowed.
 #[derive(Debug, RustcEncodable, RustcDecodable)]
 pub struct PackageSet {
     pub syntaxes: Vec<SyntaxDefinition>,
@@ -152,6 +159,7 @@ impl PackageSet {
     /// It is necessary to do this before parsing anything with these syntaxes.
     /// However, it is not possible to serialize a package set that has been linked,
     /// which is why it isn't done by default, except by the load_from_folder constructor.
+    /// This operation is idempotent, but takes time even on already linked package sets.
     pub fn link_syntaxes(&mut self) {
         for syntax in self.syntaxes.iter() {
             for (_, ref context_ptr) in syntax.contexts.iter() {