The Tribler team recently published version 5.1 of it's Bittorrent client with a much-improved search function. Tribler, which is developed by around two dozen scientists at the TU Delft, features P2P-based search to make web-based Bittorrent index sites and search engines obsolete.

tribler 5.1 ui

P2P-based search obviously has the advantage that it can't be taken down, but Tribler competes on another field as well: It aims to be much faster than any web-based index.

Tribler's lead scientist Johan Pouwelse told me that the client now responds to 99% of all queries within 1.5 seconds. Response times like these are enabled by the architecture of the client, which uses semantic clustering to reply to search requests in a timely manner.

tribler response time

I tested the new client a little bit today and can confirm that search results appear almost immediately. However, the overall number of search results still seems to be really small, probably because not too many people actually use Tribler. In fact, I didn't get any search results at all a few times - probably because Tribler wasn't connected to its overlay network anymore.

Tribler has been developed as a research project at the TU Delft for a number of years now, and the team behind the client has certainly done a lot to advance Bittorrent file sharing. A previous version of the client featured a P2P-based, Last.fm-like recommendation system that would discover new content for you based on your download history.

However, Pouwelse and his team decided to deemphasize this feature concentrate on improving search. Some of the social features are still used under to hood to optimize users' search results, but Pouwelse told me that these functions will become more explicit in future versions of the client. One example will be content-based channels that Pouwelse describes as "P2P RSS."

His overall quest however is too keep things simple. "Bittorrent is ideal for video, but was always lacking this Grandma-friendlyness," he told me. "Central to the ease of use is fast (...) search."

Tags: , , , , ,