On Thu, Sep 15, 2005 at 04:21:03PM +0200, lorenzo wrote: > Greetings. I've joined the freenet mailing list moved by curiosity and > by some will of helping. Please forgive me if I'm talking about > already-discussed issues, and ignore me in such case. > In the last few years I've been hearing about freenet every once in a > while. I've tried it but yet I have no use for it nor could give my > spare bandwidth and disk space to the community so my support is > limited to thinking and suggesting features.
:) > My doubts about the spreading of freenet are about the searching > issues and the (im)possibility of using a google-like search engine. I > understand the implications of a search engine that could reveal what > kind of documents are on your machine; but what about metadata as a > form of searching tool? The current preferred options for searching are: - Spiders (or manual indexers) generate indexes, which are inserted into freenet, which are then fetched on demand by clients when they want to search. - Some sort of TCP-like tunnel from the client to a central server. > Let me give an example: a user pushes some data into the freenet, > being it the darknet or the lightnet. Along with the data a user > should be able to push metadata, such as any kind of information about > the content it has been posted. Information and metadata should go > along together but should be two separate entities; metadata should > also be "signed" or tied to the data by way of crypto signatures and > such. They can't be tied together for 2 reasons: 1. It must not be easily possible to go from data to metadata, for the same reason that data is encrypted. 2. It must be possible to search the metadata scalably. This means not a broadcast search! Hence the suggestion of simply having a central spider, which then extracts metadata (and contents) and sticks them in a database, a la google. Then either it provides a service through a tunnel, or it publishes the indexes. > A freenet-search engine should search only into the metadata, and give > the results under the form of hashes or document keys. Once a user has > a key he/she can ask for the appropriate document via the freenet. Google searches full text (but not the content of images etc). > > Let me give an example: if I publish sensitive information about how > to circumvent the Great Firewall of China, and I live in China, > probably I won't be able to tell the key of the document directly to > friends or to post it on the web. True. So you tell it anonymously to friends. Via Frost, via freenet hosted IRC, freenet hosted mailing lists, TFE's submission queue, etc. > I'd be traced and killed very soon. > But if I can publish the data anonymously and tie metadata such as > "firewalling/China/circumvention" to the document, any search engine > can see it and make it findable by anyone. In theory yes, however there is the huge problem of spamming. If [bad guys] can get onto the network, they can insert a zillion files with obvious metadata which point to garbage. > Even more: the metadata > should be encrypted with one or more search engines public keys. > If my publishing net is connected with the search engine, my job is > done. Nobody can be sure that the networking I'm publishing in is > really connected to the "main" freenet network, but that's another > issue. > Encrypting metadata and giving it more priority into the net can make > also impossible for a user to understand what's on their machine by > looking at the metadata that is going along. Why not just include it in the content of the website, then announce the website, and have spidering search engines watch for updates, new links, and new announced sites? > > Problems could be: if a user searches for the key they can find what's > the document about. I don't see a solution to this, so I think > metadata should not be mandatory. > > this is just an idea; as I said before, please forgive me if such > issues have already been discussed or are already implemented in > freenet; the only thing I've been reading are the faqs and some > archives of this mailing list. > > best regards, > > lorenzo g. -- Matthew J Toseland - toad at amphibian.dyndns.org Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20050915/b7109b45/attachment.pgp>
