diff --git a/README b/README.md
similarity index 86%
rename from README
rename to README.md
index 4ef40c7..99cf12d 100644
--- a/README
+++ b/README.md
@@ -1,141 +1,122 @@
 # SWH Tarball Loader
 
 The Software Heritage Tarball Loader is a tool and a library to
 uncompress a local tarball and inject into the SWH dataset its tree
 representation.
 
 ## Configuration
 
 This is the loader's (or task's) configuration file.
 
 *`{/etc/softwareheritage | ~/.config/swh | ~/.swh}`/loader/tar.yml*:
 
 ```YAML
 extraction_dir: /home/storage/tmp/
 storage:
-  cls: local
+  cls: remote
   args:
-    db: service=swh-dev
-    objstorage:
-      cls: pathslicing
-      args:
-        root: /home/storage/swh-storage
-        slicing: 0:2/2:4/4:6
-
-send_contents: True
-send_directories: True
-send_revisions: True
-send_releases: True
-send_occurrences: True
-
-content_packet_size: 10000
-content_packet_block_size_bytes: 104857600
-content_packet_size_bytes: 1073741824
-directory_packet_size: 25000
-revision_packet_size: 100000
-release_packet_size: 100000
-occurrence_packet_size: 100000
+    url: http://localhost:5002/
 ```
 
 ## API
 
 Load tarball directly from code or python3's toplevel:
 
 ``` Python
     from swh.loader.tar.tasks import LoadTarRepository
 
     # Fill in those
     tarpath = '/some/path/to/blah-7.8.3.tgz'
     origin = {'url': 'some-origin', 'type': 'dir'}
     visit_date = 'Tue, 3 May 2017 17:16:32 +0200'
     revision = {}
     occurrence = {}
 
     # Send message to the task queue
     LoadTarRepository().run((tarpath, origin, visit_date, revision, [occurrence]))
 ```
 
 ## Celery
 
 Load tarball using celery.
 
 Providing you have a properly configured celery up and running, the
 celery worker configuration file needs to be updated:
 
 *`{/etc/softwareheritage | ~/.config/swh | ~/.swh}`/worker.yml*:
 
 ``` YAML
 task_modules:
   - swh.loader.tar.tasks
 task_queues:
   - swh_loader_tar
 ```
 
 cf. https://forge.softwareheritage.org/diffusion/DCORE/browse/master/README.md
 for more details
 
 
 ## Tar Producer
 
 Its job is to compulse from a file or a folder a list of existing
 tarballs. From this list, compute the corresponding messages to send
 to the broker.
 
 ### Configuration
 
 Message producer's configuration file (`tar.yml`):
 
 ``` YAML
 # Mirror's root directory holding tarballs to load into swh
 mirror_root_directory: /srv/storage/space/mirrors/gnu.org/gnu/
 # Url scheme prefix used to create the origin url
 url_scheme: http://ftp.gnu.org/gnu/
 type: ftp
 
 # File containing a subset list tarballs from mirror_root_directory to load.
 # The file's format is one absolute path name to a tarball per line.
 # NOTE:
 # - This file must contain data consistent with the mirror_root_directory
 # - if this option is not provided, the mirror_root_directory is scanned
 # completely as usual
 # mirror_subset_archives: /home/storage/missing-archives
 
 # Randomize blocks of messages and send for consumption
 block_messages: 250
 ```
 
 ### Run
 
 Trigger the message computations:
 
 ```Shell
 python3 -m swh.loader.tar.producer --config ~/.swh/producer/tar.yml
 ```
 
 This will walk the `mirror_root_directory` folder and send encountered
 tarball messages for the swh-loader-tar to uncompress (through
 celery).
 
 If the `mirror_subset_archives` is provided, the tarball messages will
 be computed from such file (the `mirror_root_directory` is still used
 so please be consistent).
 
 If problem arises during tarball message computation, a message will
 be outputed with the tarball that present a problem.
 
 It will displayed the number of tarball messages sent at the end.
 
 ### Dry run
 
 ``` Shell
 python3 -m swh.loader.tar.producer --config-file ~/.swh/producer/tar.yml --dry-run
 ```
 
 This will do the same as previously described but only display the
 number of potential tarball messages computed.
 
 ### Help
 
 ``` Shell
 python3 -m swh.loader.tar.producer --help
 ```