...

• Enable a user to securely share a dataset with other users.
• It must be possible to limit access to datasets based on user-defined groups.
• It must be possible to use a normal BitTorrent client both for sharing and for downloading datasets.
• An API is needed for integration with other systems

...

Dataflow

The following picture illustrates the flow of data in Lobber:

...

There are two types of data involved: torrent files that describe data about datasets and the datasets themselves. In order to share a dataset a user creates a torrent file - either directly using a BitTorrent client or by using the Lobber java webstart client launched from the Lobber web application. The torrent-file is a relatively small file which is uploaded to the Lobber web application. The dataset itself is served from the users BitTorrent client (eg the java webstart client or the Lobber dropbox application, cf below).

As soon as the torrent file is registered with the Lobber web application a notification is sent out over the Lobber notification bus. Currently we use stomp but there are plans to replace this with pubsubhubbub. Information about new torrent is also available as RSS feeds (ATOM will also be available when we change to pubsubhubbub for notifications).

At this point one or more Lobber storage nodes will receive the notification and start to download the dataset from the users BitTorrent client (the choice of storage-node follows the lobber authorization model, cf below). As soon as a storage node finishes the download the client can stop serving up the dataset.

...