On Tue, Jul 03, 2001 at 06:46:21PM +0100, Adam Langley wrote:
> I wonder if the announcment protocol couldn't be made to work off
> multiple node addresses.
I have coded it accordingly in Fred (it is run once for every seed
node). Currently it would use the same UniqueID to avoid collisions - we
might need to add a branch number to the announcementrequest message to
make sure there are no issues with this of course.
> (Oskar please don't kill me for this) At the moment the Alice contacts
> her one known node Bob and Bob forwards the announcement, forming a
> chain. The messages pass up and down this chain a number of times. If
> Bob is Evil he can simulate the whole announcment and MITM Alice
> forever.
>
> If Alice could get more than 1 reference she could:
>
> Order them into Nodes[n]
> for i in Nodes:
> contact i and say "You are part of an announcement chain and X is
> before you and Y is after you"
> Start the annoucement at Nodes[0]
>
> The announcment gets forwarded along the chain Alice setup and if it
> runs off the 'end' (a node Alice didn't give a next-in-chain noderef
> to) the last node forwards randomly until HTL runs out (as it does
> currently).
>
> Only 1 node in the chain has to be good for the agreed keyspace to be
> random and the one of the annoucement replies could be signed by all
> the nodes in the chain who Alice setup - thus she can check that all
> the nodes she setup took part.
>
> We could then be less strict about the init noderef because it's not
> so dependant on 1 node (assuming Alice gets a good spread of initrefs
> from web sites, IRC, whatever).
I think just initializing n chains is just as well. Since I am not a big
believer in the "single node single epicenter" theory regarding the way
that Freenet topology develops, I don't think giving alice n (which will
be small, < 5 or so) initial reference values is a problem.
>
> AGL
>
> --
> Whenever anyone says, "theoretically," they really mean, "not really."
--
'DeCSS would be fine. Where is it?'
'Here,' Montag touched his head.
'Ah,' Granger smiled and nodded.
Oskar Sandberg
[EMAIL PROTECTED]
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://lists.freenetproject.org/mailman/listinfo/devl
