A mathematician from Yekaterinburg cooked “Broth” from wiki and reputations.

On June 20, the first public server of the “Broth” network, a distributed P2P system of massive joint filtering, was launched for testing. As the developer Victor Grishchenko explains, this is something like a wiki with reputations, but only here another engine is used. “At the heart of the“ Broth ”is the oc-co engine, which uses the XMPP protocol (Jabber). This engine, using social connections and opinions of participants, collects XML documents from distributed pieces. It uses peer-to-peer, not client-server algorithms, unlike a simple wiki. ”



Victor Grishchenko, a software mathematician from Yekaterinburg, studies reputation metrics and free-scalable graphs ( scale-free networks), and also teaches a special course on this topic at Ural State University. The Internet project called "Broth" is part of the scientific work of Victor. Now anyone can check this system in action, since on June 20, 2006 the first public server started working.



The core of the distributed system “Broth 2.0” is an external component for the Jabber server and is implemented in Java 1.5. The client part is implemented in Ajax technology, the client is the browser (currently only Firefox is supported, preferably version 1.5 and higher). The server part is created in Java (Berkeley DB JE, ​​servlets).



“Broth” is a scientific project and released under the GPL license. According to the description of the author, the project is similar to a distributed wiki system with reputations, where the page is searched for by title, and its content is the result of co-filtering the entire social network of users. At the same time, the “Broth” servers are a single network with a single address space. That is, “Broth” is an editable web without spam.

')

The foundation of the “Broth” network is a social network of participants formed by their IM contacts (using the Jabber IM network). When you read a page, “Broth” interrogates your friends about its contents and assembles the received answers (“pieces”). When you make edits, the piece you edited will be stored and provided on request. All requests-answers, naturally, occur automatically, imperceptibly for you.



The main "highlight" - a system for calculating reputation. “There are no standard and generally accepted formulas in this field of science,” says Grishchenko. - The algorithm for calculating the reputation, however, is quite simple and close to the naive probabilistic understanding, that is, reputation is the probability of obtaining a relevant opinion, calculated from the results of previous interactions. As if you were throwing a dice with white and black faces, and the dice’s reputation would be the percentage of whites. ”



Each friend (contact) is assigned a reputation, which is also the “reputation distance”. The pieces are ranked according to the distance from which they came - the further the author or the recommender of the piece, the smaller, faded type this text is displayed on the web page. You can “confirm” a page or a piece, and “delete” garbage. When you confirm the relevance of a piece, the reputation of those friends who recommended it grows. When you delete - the reputation of referees falls.



It is unlikely that "Broth" claims to carry some practical benefits. The project is of purely scientific interest. Its creator himself while for sure does not know exactly where the system will find application: “It is difficult to say for now what ecological niche will be occupied by” Broth ”, since the release was just that. The engine itself works with arbitrary XML, so there are already proposals to apply it not only for the real-time P2P wiki, but also for shared bookmarking, tagging photos like Flickr , etc. To a greater or lesser degree, all modern "social services" such as del.icio.us , citeUlike, and the like can be implemented on the oc-co engine. The global plus of this option is that the same social network is used. Modern social services have to rebuild the network every time, the user cannot “pick up” his network from one site to another. It is possible to create a new type of Internet forums, in which "all of their own" and there is no problem of littering. In short, there are many possibilities. ”