diff --git a/README.md b/README.md
index 877ae37..1a51220 100644
--- a/README.md
+++ b/README.md
@@ -1,15 +1,13 @@
-# NE DB (Node Embedded DataBase)
+# NeDB (Node Embedded DataBase)
-Embedded persistent database for Node.js, with no dependency (except npm modules of course). The API is the same as MongoDB.
+Embedded persistent database for Node.js, with no dependency (except npm
+modules of course). You can think of it as a SQLite for Node.js, which
+can be installed and used in less than 30 seconds. The API is a subset of MongoDB's.
-**It's still experimental!** I'm still stabilizing the code. The API will not change though. Don't hesitate to file issues/PRs if you find bugs.
-
-## Why?
-I needed to store data from another project (Braindead CI). I needed the datastore to be standalone (i.e. no dependency except other Node modules) so that people can install the software using a simple `npm install`. I couldn't find one without bugs and a clean API so I made this one.
## Installation, tests
```javascript
-npm install nedb
+npm install nedb --save // Put latest version in your package.json
make test
```
@@ -164,19 +162,29 @@ db.remove({ system: 'solar' }, { multi: true }, function (err, numRemoved) {
## Performance
### Speed
-It is pretty fast on the kind of datasets it was designed for (10,000 documents or less). On my machine (3 years old, no SSD), with a collection with 10,000 documents:
+It is pretty fast on the kind of datasets it is designed for (10,000
+documents max). On my machine (3 years old, no SSD), with a collection
+containing 10,000 documents:
* An insert takes 0.1ms
* A read takes 5.7ms
-* An update takes 62ms
-* A deletion takes 61ms
-
-Read, update and deletion times are pretty much non impacted by the number of concerned documents. Inserts, updates and deletions are non-blocking. Read will be soon, too (but they are so fast it is not so important anyway).
+* An update takes 58ms
+* A deletion takes 57ms
You can run the simple benchmarks I use by executing the scripts in the `benchmarks` folder. They all take an optional parameter which is the size of the dataset to use (default is 10,000).
### Memory footprint
-For now, a copy of the whole database is kept in memory. For the kind of datasets expected this should not be too much (max 20MB) but I am planning on stopping using that method to free RAM and make it completely asynchronous.
-
+A copy of the whole database is kept in memory. This is not much on the
+expected (20MB for 10,000 2K documents). If requested, I'll introduce an
+option to not use this cache to decrease memory footprint (at the cost
+of a lower speed).
+
+
+## Use in other services
+* connect-nedb-session is a session store for
+Connect and Express, backed by nedb
+* I'm planning on making an export tool to get all your data in an nedb
+ database in a Mongo database
## License