Matthew Toseland wrote:
> Perhaps by requesting the data from another node?

Possibly... I guess it would have to be done probabilistically - every 
node along the path wants to know whether its downstream neighbour 
forwarded the request, but we don't want every node along the path to 
send out a request for the data. Also what happens if the downstream 
neighbour hoards the data instead of forwarding it, and answers the 
audit requests itself? I guess it's not a problem as long as the data's 
available...

So for each neighbour, we keep track of the inserts we've sent it and 
periodically request one of them... if a neighbour has an unusually low 
success rate then we penalise it? This sounds promising...

Cheers,
Michael

Reply via email to