[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [freehaven-dev] freehaven broadcast channels
On Mon, 21 Feb 2000, Joseph Sokol-Margolis wrote:
> I see no reason we need to provide a mechanism to ensure everyone
> get's the same broadcast. As long as a node can verify the
> authenticity of the message, we should be okay. (the mnessage's
> contents are tied to server, so if it's bogus there's a trust hit)
Cool. I think dropping this requirement will save us *a lot* of
communication complexity. For instance, I guess that it is this
requirement which leads to the O(n^2) messages in the algorithm
Adam Smith mentioned. Because otherwise I can envision this algorithm :
* Node sends message to all other nodes with a request for return receipt.
The return receipt could take either the form of a message sent to
a particular (newly generated?) reply block, or a public usenet posting,
or something else. Doesn't necessarily have to identify the sender.
* All good nodes which receive the message send a return receipt.
2n messages, n number of destination nodes. All good nodes are identified,
and nonresponding nodes noted...at least as long as we have a timeout.
I can't be sure about this being the main reason for the O(n^2) until I
get my hands on that distributed algorithms book or the Dolev paper.
Fortunately, I think the book is required reading for one of the courses
here this term, so there will be a bunch of copies floating around. If
that fails, the library has that paper.
Note that everyone gets the same message, but now they don't know that
they're all getting the same message. Which is where the nasty extra
messages would come in.