diff --git a/Dockerfile b/Dockerfile
index 34e23b14df..1c07b9024c 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -21,7 +21,7 @@ run	apt-get install -y -q reprepro
 # Install s3cmd 1.0.1 (earlier versions don't support env variables in the config)
 run	apt-get install -y -q python-pip
 run	pip install s3cmd
-run	/bin/echo -e '[default]\naccess_key=$AWS_ID\nsecret_key=$AWS_KEY\n' > /.s3cfg
+run	/bin/echo -e '[default]\naccess_key=$AWS_ACCESS_KEY\nsecret_key=$AWS_SECRET_KEY\n' > /.s3cfg
 # Download dependencies
 run	PKG=github.com/kr/pty REV=27435c699;		 git clone http://$PKG /go/src/$PKG && cd /go/src/$PKG && git checkout -f $REV
 run	PKG=github.com/gorilla/context/ REV=708054d61e5; git clone http://$PKG /go/src/$PKG && cd /go/src/$PKG && git checkout -f $REV
diff --git a/release.sh b/release.sh
index 58e6dd3ceb..4d589e3126 100755
--- a/release.sh
+++ b/release.sh
@@ -14,15 +14,33 @@ set -e
 
 # Print a usage message and exit.
 usage() {
-	echo "Usage: $0 VERSION BUCKET"
-	echo "For example: $0 0.5.1-dev sandbox.get.docker.io"
+	cat <<EOF
+To run, I need:
+- to be in a container generated by the Dockerfile at the top of the Docker
+  repository;
+- to be provided with the name of an S3 bucket, in environment variable
+  AWS_S3_BUCKET;
+- to be provided with AWS credentials for this S3 bucket, in environment
+  variables AWS_ACCESS_KEY and AWS_SECRET_KEY;
+- a generous amount of good will and nice manners.
+The canonical way to run me is to run the image produced by the Dockerfile: e.g.:"
+
+docker run -e AWS_S3_BUCKET=get-staging.docker.io \\
+              AWS_ACCESS_KEY=AKI1234... \\
+              AWS_SECRET_KEY=sEs3mE... \\
+              f0058411
+EOF
 	exit 1
 }
 
-VERSION=$1
-BUCKET=$2
-[ -z "$VERSION" ] && usage
-[ -z "$BUCKET" ] && usage
+[ "$AWS_S3_BUCKET" ] || usage
+[ "$AWS_ACCESS_KEY" ] || usage
+[ "$AWS_SECRET_KEY" ] || usage
+[ -d /go/src/github.com/dotcloud/docker/ ] || usage
+cd /go/src/github.com/dotcloud/docker/ 
+
+VERSION=$(cat VERSION)
+BUCKET=s3://$AWS_S3_BUCKET
 
 setup_s3() {
 	# Try creating the bucket. Ignore errors (it might already exist).