On Sat, Oct 13, 2012 at 05:57:43AM +0200, xor wrote: > On Saturday, October 13, 2012 05:41:25 AM xor wrote: > > I strongly feel that this might be just a different model of the same stuff > > we do with current web of trust systems > > Well, I should admit that what I said here is a bit over simplified. > Maybe I should only say that the algorithmic complexity might be the same and > both systems are some kind of web of moderators.
The order of magnitude is probably the same with PSKs and without PSKs. However it is likely a massive linear speedup, depending on the complexity of the local WoT, tradeoffs on group sizes, how many groups we join, and so on. And that's worthwhile. But we can make WoT+Freetalk scale anyway (if all we care about is the order of magnitude scaling of performance). (And then we can make it fast by a linear speedup from PSKs) Polling only those we explicitly trust is problematic because we would not see other posts, at least not quickly. IMHO it would be too slow. Polling regular posters to our boards would scale reasonably well, because in practice the set of regular posters to any given board is likely to be limited. All we need for this to work is some way to detect newbies, so we see posts from new posters (and from those who haven't posted in ages). There are 4 options AFAICS: 1. Poll everyone so we pick up newbies immediately regardless of who they announce to. (Does not scale!) 2. Rely on hints from regular posters. (SLOW!) 3. Require newbies to announce to the regular posters on the board, and hope that the regular posters publish puzzles, answer them, don't arbitrarily censor newbies and so on. (This isn't great either; a board can easily be a "closed shop"; e.g. digger3 prefers WoT to centralised moderation because you *can* see everything and make your own choices if you want to) 4. Some sort of board-specific (context-specific) announcement protocol, preferably visible to everyone. The final option is the best IMHO, but I'm not sure whether it is feasible without hashcash. The basic principle is either: A) Puzzles are automatically generated, and not posted by anyone. E.g. puzzle is to find X such that the last N bits of H(X) = the N'th hash of the board name. This is quite workable with hashcash, but hashcash is problematic, as the gap between an attacker's system and a newbie with a slow computer's system is around 1000x! B) Puzzles are posted by regular posters, but the solutions can still be seen by anyone. E.g. a long CAPTCHA that can be verified by hashcash, with WoT-based mechanisms for dealing with posters who publish intractable puzzles. The problems with hashcash are actually very similar to the problems with CAPTCHAs: Many users find them difficult, but they can usually be cracked by a computer, or solutions bought very cheaply in bulk. Bitcoin would be a possibility but again has its own problems. Announcement is an unresolved problem but it's a problem that the internet at large has too - it's just a bit harder for us, because e.g. we have to allow multiple guesses at the same puzzle (since we're using them as a key), and we allow more time than a website would. :( We need a Freenet-specific solution, e.g. leveraging the darknet, and maybe we'll build one long term (e.g. Scarce SSKs: special inserts rate limited on every link). Meantime context-specific announcement looks feasible.
signature.asc
Description: Digital signature
_______________________________________________ Devl mailing list Devl@freenetproject.org https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
