On Sun, Jun 03, 2001 at 03:40:58PM -0400, Tavin Cole wrote:
> > 1) is an inevitable consequence of any improvement to Freenet
> > functionality,
> True, but this is something that can be cleanly separated from the
> requirements for implementing a "Freenet" node.
Which is essentially semantics.
> > and is countered by the fact that itegrating this
> > functionality into Freenet means that there will be more nodes capable
> > of providing this search functionality, which will be good for the
> > network.
>
> I think you will have to prove that.
I think it is obvious that the more people who run nodes which support
this searching functionality the better it will be.
> This system will certainly require
> some kind of web of trust rankings system to provide useful search results.
Not nescessarily, both Gnutella and Napster get along without them.
> You won't be able to stop searches on the first hit like you can when
> searching for a Freenet key, so the topological requirements of the
> network may become different, and in any case, you will be dealing with
> a different set of challenges regarding bandwidth usage and caching.
I think that results will be interesting, but with the exception of
allowing multiple search responses (each of which will be cached
probabilistically I guess) it is largely the same as the current
mechanism, but with a more sophisticated closeness function. Of course
there are questions, as there were (and still are) with Freenet, but I
think that there is a good enough chance it will work to make it worth
trying.
> It's really a very different animal and it could be horrible for the network,
> just as it could be good.
It should be designed so as to have no effect on the current network,
good or bad (an isolated section of the datastore, or a separate
datastore altogether).
Ian.
PGP signature
