On Mon, Jan 11, 2010 at 10:56 PM, steve oliver <mrstevem...@gmail.com> wrote: > I was thinking out loud in the chatroom earlier, and thought that it might > help attract new users and bring enhanced usability if we had a way to > support arbitrary but secure URLs in a similar way to the public DNS system. > Perhaps this has been proposed and ridiculed before, i don't know. I've > thought through some of the potential problems and ways to resolve them, and > i think long term it is a good idea :) > > > > ------------ > Proposal > ------------ > > Our current URLs are implemented by requesting specific key types, SSKs, > CHKs, KSKs, etc. However they are completely unfriendly and confusing to new > users, difficult to tell apart at a glance, and extremely long. > > So I'm proposing a Freenet system analogous to the public DNS system, but > customized to retain anonymity, decentralization, and to fit the Freenet > model of requesting data directly rather than requesting data from a specific > computer/node. > > We've got Freemail to replace E-mail, Freetalk to replace Usenet and forums, > this would be FreeNS to replace DNS :) > > While these "domain" URLs would not be directly analogous to DNS domains, > functionally they would be similar, and the concept is already familiar to > internet users so it would be quickly understood; we're requesting a specific > Freesite. > > It is partly cosmetic of course but so is public DNS, which also makes things > a bit easier on the users. > > Each FreeNS name would identify a specific "site" the user would like to > visit, and FProxy could accept arbitrary site "domains" like this: > > http://127.0.0.1:8888/myfreesite.freenet/6/history.html > > Or, with a new key type: > > http://127.0.0.1:8888/f...@myfreesite.freenet/5/us-constitution.txt > > --------------------- > Implementation > --------------------- > > Instead of linking to a specific node/computer like the DNS system does > (which we don't want to do even if we could), it would be a redirect to an > existing site inserted using an SSK, etc. It would also "mask" the real SSK > URL much like the DNS system masks the IP of a public website. > > To enable this to work without building in central points of failure or > control over the whole thing (or censorship), the system would not have a > true "root" as the DNS system does, instead the node would support an > arbitrary number of FreeNS lookup and mapping spaces, which would be SSK URLs > containing a specific filename in a specific format (XML?), that would > include key pairs of FreeNS names mapped to specific SSK URLs. These tables > would be updated automatically in the background by the node, much like we > update new editions of activelinks periodically, or check for new Freenet > builds. > > New installs could include a default FreeNS lookup space much like we include > default activelinks, including mappings for common freesites to enable > out-of-the-box use. Users could then add one or more community maintained > FreeNS lists after installation. Of course this may introduce the possibility > (near guarantee, actually) of collisions, but perhaps the system could > include a relative trust system using WoT, where FreeNS domains returned by > one list are trusted more than another, or perhaps we just warn the user that > the freesite or filename they clicked points to two different SSKs, and let > them pick if a clearly trusted mapping does not exist. > > --------------------------------- > Possible enhancements > --------------------------------- > > The second part of the "domain" could potentially serve as a namespace for > the backend implementation of the name lookup somehow > > Individual nodes could maintain a list of "known bad" and "known good" FreeNS > mappings, and publish them alongside WoT identity trust lists, this may > enable the entire thing to function decentralized with no true "lists" at all. > > If we are making use of WoT, the creation of a FreeNS name for a newly > inserted Freesite could be announced simply by adding it to whatever message > digests we're already inserting from Freetalk for the day, others would pick > it up and propagate it in their own decentralized FreeNS lists. > > ------------- > Problems > ------------- > > One potential problem could be that a user follows a badly formed link > somewhere in freenet, or tries to just stick > "myfreesite.freenet/6/index.html" in a browser box, which may cause the > browser to issue a real DNS request, potentially compromising anonymity. > Perhaps we can subtly alter the string format such that common browsers just > fail to make the lookup, but at least Firefox is known to forward malformed > URL strings to Google if they can't be resolved. >
I only skimmed this, but I don't see an answer to the obvious question. What happens when two different users start using the same "domain"? How do you resolve who gets it, when there are already two existing sites using it? Solve that, and it becomes usable. Alternately, ignore me and other nay-sayers and start coding. Evan Daniel _______________________________________________ Devl mailing list Devl@freenetproject.org http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
