Oskar Sandberg wrote:
> Actually, the university problem can be solved by so called shy nodes. A shy
> node would not use the DataSource to autodiscover the any new nodes, and nodes
> passing DataReplies from shy nodes would always reset DataSource to 
> themselves.
> This provides a security behind which node operators can contribute to the
> future Freenet but limit their node to talk to trusted nodes (and not have 
> it's
> address sent accross the network, where a snooping enemy might see it), but
> it also provides a way for nodes that wish to talk to Freenet only through one
> or a couple of "gateways" to do so for speed reasons.

Hmm.  I see some problems here:

1. Most universities and ISPs are not going to run freenet gateways, due
in part to liability fear but primarily laziness and conservatism. 
They'll run their mail servers and that's about it.

2. How to you get users to set up shy nodes, especially since shy nodes
will probably perform worse than regular nodes since they're forced to
talk to/through overloaded gateways?  Most people will download the
standard freenet node and run it.  

3. This doesn't help locality much unless the gateways have huge
caches.  You could have a thousand (or more) freenet nodes at a
university, and potentially millions on a large ISP.  These should be
able to cache for each other without relying on an overloaded central
resource.

4. The gateways become single points of failure, both technically and
legally.  Shutting down the gateway effectively erases all files stored
on the \nodes "behind it."

_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev

Reply via email to