Some recent discussion about Freenet on phones (and routers, see my other message on tech). Clearly running a regular broadband based Freenet node on a cracked iPhone is unlikely because of cost and power issues (although it has 256MB of RAM and at least 8GB of storage). However, a smartphone with some fast short range networking would be ideal for 1) Haggle, and 2) Freenet darknet rendezvous.
Haggle is, roughly speaking, opennet sneakernet. It was originally conceived for phones and is more or less ideal, on certain provisos. The basic operating principle is this: - Shout at the top of your voice "Does anyone have a copy of [ censored ]" wherever you happen to be - on a bus, in a crowd - and if they have it they'll send you it. - Hope that nobody has tracing equipment. - I believe there is some level of opportunistic relaying of requests, but it's not really routed in a scalable sense we would recognise. - One worry is that phones might not have untraceable wifi. Another is that relatively cheap infrastructure or patrols with tracing equipment could bust a lot of users, if the law is sufficiently harsh. But really it *is* an interesting system, it's just not Freenet. This may be an out of date view of Haggle, see here: http://www.haggleproject.org/index.php/Main_Page Freenet darknet rendezvous is essentially darknet sneakernet, but without having to pass around USB keys. Your phone detects when your registered friends are nearby, and does a burst transfer over UWB (wireless USB, 480Mbps over very short range) with them. Modern phones are built to do the detection phase cheaply, but don't yet have UWB; it is likely they will soon however. Host-side USB with a networking cable would be another possibility. Thus any time you go for a pint with friends you automatically propagate data, requests, etc, without having to do anything. When you go home your phone will again automatically sync with your fixed node (PC/router), which then forwards stuff over your fixed connections - regular Freenet connections (possibly with the same friends, which gives some nice optimisations), stego connections over the internet, fixed wifi links, etc. Unlike Haggle, it is strictly limited to known friends (hence safer), and has a more freenet-like (and therefore probably slower, because less broadcasty) routing system. Publish/subscribe is essential, but much of it is organised around global (and anonymous) streams. Routing is also possible: although it may take days or weeks depending on the transports, popularity, etc, it should be possible to obtain a huge variety of content. So what we are talking about is a robust, anonymous, often slow meta-internet capable of using whatever connectivity is available. This is a natural evolution of current Freenet IMHO: Messaging and other applications require good publish/subscribe for efficiency in the medium term. Even now, Freenet's expectation that nodes will be online 24x7 is unreasonable. Darknet (connecting only to friends) was introduced to make it possible to run Freenet in places where opennet may be harvested and blocked. Darknet networks composed of low-uptime systems are unlikely to have full end-to-end connectivity most of the time, which means we will need some form of persistent requests. And any competent state trying to eliminate an underground Freenet darknet (note Iran's recent communications crackdown) will soon realise that looking for long-lived peer to peer UDP connections will bust most nodes. Steganography can only go so far, traffic flow analysis will ultimately find all nodes - but even good stego will need to not exchange data continually, making the uptime issues even worse. Ultimately sneakernet and rendezvous based transports become very attractive. And they can have pretty good bandwidth too, although latency is poor. All that routing really requires (provided we can figure out a way to assign locations), is that there be many short links (e.g. meeting up in a pub after work, fixed wireless links), and a few long links (LUG/2600 meetings, mailing a box of disks etc). So I expect *before 1.0*, and largely on the basis of our present network behaviour (poor node uptime, messaging) we will have to implement: - Good publish/subscribe (aka passive requests) - Bloom filter sharing (awareness of data on friends' nodes, speeds up routing considerably but also has some nice impacts when you reconnect) - Long-term requests (meaning they persist on the network, pick up data as nodes come online, and forward it back to the originator) Both of these tie in reasonably well with existing UI and APIs IMHO. After 1.0, we should look very seriously into sneakernet, and non-real-time steganographic transports (e.g. faking VoIP calls). With the above feature set, it becomes quite plausible. Obviously fetching rare content could take weeks - but popular content should be faster, and popular publish/subscribe streams should be reasonably responsive. Such a system might need some adjustments such as larger block sizes, but IMHO it would likely reuse a lot of Freenet technology, even if it was not Freenet itself. And I'd argue that even if it is incompatible with Freenet 1.0, it would be worthy of support from FPI, given its mission statement. Thoughts?
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ chat mailing list chat@freenetproject.org Archived: http://news.gmane.org/gmane.network.freenet.general Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/chat Or mailto:chat-requ...@freenetproject.org?subject=unsubscribe
