Le jeudi 07 mai 2009 21:41:44, Thomas Bruderer a ?crit : > > I vote for this option, to be exact, i vote for the way, how FMS basicly > > works: > > > > - 1 trustlist for every application: With this value, you set, if you > > want to read someones messages, see someones files, freesites or whatever > > your WoT-based app uses. No value by default, if trusted identities have > > set a value for an identity, it is used for trust calculation (details > > for the calculation can be found on the FMS freesite). If you yourself do > > set a value, it may either be part of the trust calculation or > > (optionally) overwrite the calculated trust. If no value is set or there > > exists a value and it is bigger than a locally set limit, you download > > his data, else it is ignored. > > > > - 1 trustlist trust: This does set, if you trust the person to introduce > > identities and to vote "right" on people. If you set this value and it is > > higher than your local limit, then you download his trustlists and the > > values in them are used in the application trust value of every known > > identity. > > > > This system is relativly simple, you can understand it and follow it, if > > you want and it does use the "little world theory" and the idea of a > > friends network. And, just as a reminder: This idea and the basic concept > > was created and agreed upon between all actively writing frost identities > > during the beginning of the massive frost spam. > > I absolutly dislike the Idea of two lists (FMS is much too complicated > for my taste). In fact I am not sure if there is the need of different > namespaces, the likelyhood that you like someones uploads when he doesnt > like his Messages is very unlikely, but if someone spams on Freetalk its > pretty likely he also spams other applications. If someone needs > different Identities for different Application, he should just generate > 2 different ones. > > But if we stick to namespaces! Just applicate any Metric on all the > namespaces available. This could be something simple like Min() or > Median() or whatever Metric you like. Just calculate this Metascore and > handle it accordingly. >
I agree with this solution. Just compute trusts for one given context, and fetch the identity accordingly. With that solution, you're sure not to download any spam identity (well, not too much) for a given context. You need to compute more trust (one per context), but it avoids the (imho) ugly trustlist trust solution. > BTW: The scale is from -100 to 100, but as someone already pointed out > this doesnt indicate if we allow negative trust, which I would suggest > to not use. If you can gain something of generating a new ID (i.e trust > 0 in contrast to -100) the whole scale below 0 will be useless. > > greets > Apophis / Thomas > _______________________________________________ > Devl mailing list > Devl at freenetproject.org > http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
