On Thu, Aug 17, 2000 at 11:44:44PM +0700, Oskar Sandberg wrote:
> On Thu, Aug 17, 2000 at 10:53:09AM -0500, david at aminal.com wrote:
> > 
> > My preference would be to create an identifier that is suited to
> > the needs of Freenet, and just use the Internet to get from hither
> > to yon, without relying on it to tell us anything about who or what
> > is *at* hither or yon.
> 
> That is my personal preference too, but many things are my personal
> preference that just can't happen. There are a lot things I keep hearing
> we must do:
> 
> - we must resolve addresses from an internal node address space
> - we must have some form micro payment system
> - we must get perfect anonymity
> - we must have a moderation systems for search results / namespace
> - we must have links rated by capacity
> etc etc ad infinitum
> 
> but nobody ever bothers to tell us how any of it can possibly be done.
> 
> As far as the address space is concerned, we have these options:
> 
> - use DNS or another centralized database.  
> - use the ARK system I proposed which should allow for some recovery as
> long as addresses shift at a reasonable rate.

> - combination of the above
> 
> Ideally I would love to have a system were a node is identified only by
> it's public key and each node does a lookup through some perfect system to
> figure out how to send to it. But for those of us whose noses are actually
> in the code of this system trying to make it work ideally is a long way
> off.
>

What's the problem?

What - a document

who - a node identifier

where - how do I talk to the node


What is covered, documents have keys.

where is covered, there is currently how to get there info in the
form of the tcp address and the port.

we need a who. Who can be *anything unique and unchanging*.

Documents are stored on nodes. Documents have keys, nodes
have identifiers. When a node wants to reveal itself by
associating with a document, it needs to announce all three
pieces. The node identifier, the document key, and a means you
can use to reach the node at the time of the announcement.

When you store this stuff, it all relates to the node
identifier. A node identifier can have multiple ways of
being reached attached to it. If the tcp address doesn't
work, try another one. Try reaching it via its modem
and telephone number. In other words, A node identifier
should be able to appear in the data store with a number
of different ways of reaching it attached to the
identifier, just like it can appear with a number of
different document keys.

Now all you need to do is to prevent the node identifier
from being spoofed, and I thought that was what signatures
were for, but maybe I'm misunderstanding. I really don't
know much about crypto, so that wouldn't suprise me. 


David Schutt
















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

Reply via email to