> In my personal opinion, an important thing we did (tried to do) > was start pinning down what "anonymous" means with respect to file storage > and availability systems. What does it actually mean to say "anonymous for > all practical purposes," and how is it that Gnutella can claim to be > "anonymous"? What grounds can we give for the statement "Freenet is more > anonymous than Gnutella," and how should we reason about this anonymity? I do agree that this is valuable. I found it very interesting when someone recently posted to the dev list saying that they wanted to introduce Freenet into China, and what were the threats. There were a number of comments detailing the resources that would be required to identify authors and readers of information on Freenet. I feel that this is a good way to measure anonymity, ideally a dollar value could be placed on the cost to achieve different things such as breaking reader anonymity, and breaking writer anonymity. Of course, this would need to take into account likely future improvements in technology (halve this figure for every year after 2000 etc etc). I am probably not as concerned about anonymity as many people on this mailing list, and I am convinced that all the anonymity (however defined) in the world won't be of any use unless the system is usable to the general public (for example Oskar, I would imagine, would take a more "if they can't be bothered then they don't deserve anonymity" approach), and comparable to competing information distribution mechanisms in terms of reliability and efficiency. > > FreeHaven addresses a different problem than Freenet, which is how to > > create an anonymous, distributed, *permanent* data haven. Data > > distribution is actually incredibly slow. It can take days to fetch a > > document (but it will always be there). > > Yes -- one way of thinking about this is data "storage" vs. > "availability." A related design choice we made is that our system > is what we call "content-neutral" -- we try to keep everything until a > publisher-determined expiration date (which could be 'forever'), > regardless of how often it is accessed. Ouch, the real problem with this is that it means your system could fill up. Basically you give priority to old data, where as Freenet gives priority to new or popular data. Both approaches have their shortcomings but I think in practical terms Freenet's is preferable for most circumstances. If you want, view Freenet more like a radio-station than a library. > One of the real-life examples we found of why this is important is the > fact that old phone books from Kosovo are now being used to document > residency for people claiming refugee status there. The old phone books > don't need to be available quickly, and they wouldn't last long if they > had to be requested often. Perhaps, but if you kept every old phonebook your system would rapidly fill up making it nothing more than a read-only archive of old data. > > I've talked with them and they like the idea of somehow integrating > > Freenet and FreeHaven so that we can have the best of both worlds. > > In particular, data can be stored on a Free Haven or similar service which > has the strongest resistance to adversaries we know how to build. Then > when the data needs to be distributed, it can be extracted and placed on a > Freenet node or web page. This is an option, but who does the extraction? Ian.