I think this is a very interesting discusion and I have changed sides a
couple of times.
Basicly I think you have both succesfully pointed out weaknesses in each
others propopsals in such a way I think neither Idea will do.
I think the 'slashdot' effect Oskar's idea will cause is *really* bad.
Currently requests should hardly ever reach their local epicentre because
popularity will have formed an insulating ring of cashed copies around it.
Excactly what makes Freenet 'slashdot' effect proof however the follow
through requests would penetrate this cushion.
As for Ian's proposal I don't think it has a single point of failure like
Oskar's, but I'm not sure it will be succesfull enough. As long as the
information is not too old I think it could work. But as soon as links
between several epicentres are severed (wich I asume will happen quite
quickly if we introduce performance criteria into the routing) it will fail
to reach significant portions of the cashed data.
This is the price you pay for anonymity everything points away from the
insert and towards the local epicentres.
I can't see a away around these problems, wich makes me sad.
Neil
_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
