On Sunday 14 Oct 2012 22:40:55 Arne Babenhauserheide wrote: > Am Samstag, 13. Oktober 2012, 05:41:25 schrieb xor: > > I strongly feel that this might be just a different model of the same stuff > > we do with current web of trust systems > > I thought the same at the beginning, but I now think it might actually have a > different domain. > > In a web of trust, an update is cheap, but it propagates slowly because it is > polling based. More exactly: The polling is done by the downloaders. > > In the group keys, it seems that the cost of the polling will be pushed to > the > upload side, because the stored keys will be updated more often: The > verification has to be done by the nodes accepting the upload (on the storage > layer). > > Imagine a group of 5 people who all post with roughly the same frequency. > Everytime one of the posts, that’s one poll on the key and then one upload. > > Imagine on the other hand a group of 20 000 people where only 5 post. Then > the > upload checking would waste much less time. > > It would be pretty similar to a WoT where every user also publishes the 5 > most > recent updates from people he/she trusts. That could even be done context- > based: Just include the revisions of the last 5 posters in the 5 forum boards > we last wrote to. > > > Effectively it seems to be key-merging in the storage layer to avoid > intermediate fetching from and reuploading into freenet. > > That *could* actually be an algorithmic advantage, because the speed of > freenet changes with the number of accesses, and that number might not enter > the merging on the storage-level.
Hmmm, good point. 20,000 users polling one key stream is pretty cheap. 20,000 users polling so many key streams that the gap between key fetches is larger than the ULPR interval (~ 30min to 1 hour) on the other hand is pretty inefficient. (And yes I know that's an optimistic figure!) > > Does that sound about right? If we have a board-level announcement protocol and only poll the active posters then Freetalk scales with the number of active posters on your boards. Key merging means we can have up to a few hundred posters in a group (thresholds yet to be certain), which is limited by the size of update traffic when keys are changed, and by the level of fragmentation of WoT trust. In theory we might even be able to use multi-moderator groups to optimise WoT, in which case such groups could be huge, but the cost of fragmenting on trust changes might be significant. > > Best wishes, > Arne
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Devl mailing list Devl@freenetproject.org https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
