On Saturday 30 Mar 2013 13:33:42 Matthew Toseland wrote: > On Saturday 30 Mar 2013 12:21:51 Michael Grube wrote: > > You may want to consider that the traffic from freenet may significantly > > wear down any mobile device's battery. That being said, a "lightweight" > > android client might be amazing. You may also want to consider working to > > put Tahrir(http://github.com/sanity/tahrir) on Android, which may be easier > > on a mobile device's battery. > > > > Just some thoughts. Go for it!! > > Right, the main problems are: > - The amount of traffic freenet generates will get you a huge bill or get > your data package blocked. > - Mobile networks are usually severely restricted anyway, e.g. blocking UDP. > - Battery life, as above. > > However there are a number of things that might be interesting, possibly in > collaboration with a fixed node: > - Exchanging friend connections. If you can just bring your phones close > together to set up a friend/darknet connection between your "main" nodes, > that would be very interesting. QR codes are another possibility. > - Remote access to your main node, without using centralised systems to look > up its IP address, possibly maintaining connections to some of your friends' > nodes, but with the main node doing most of the work. > - A lightweight node as Michael suggests, probably request only (no routing > of incoming requests); darknet only, since it wouldn't work well on opennet. > - Sneakernet: Transfer large quantities of data (you have a big XD card > installed, right?), via short-range high-bandwidth wireless (e.g. UWB, > 802.11n etc), when you and your friends' phones are close together. Once you > get home your phone downloads to your fixed node in the same manner. This is > particularly useful for subscriptions of various kinds (which we don't yet > support, but are planned), but also for long-term downloads (e.g. of big > files, video files, rare files), and also opportunistically filling up empty > datastores. The bandwidth is quite interesting too: Lets suppose you meet up > for an average of 10 minutes to share a pint, 5 days a week, and make an > average 200Mbps. That's enough time to exchange 14GB, which averages > 130KB/sec over the week, which by Freenet standards is very respectable. And > it's purely local communication; The Man never sees it. It can complement > your regular node to node communications quite efficiently: they can > determine what to transfer ahead of time, for example. > > Obviously there may be other architectures that work better for sneakernet. > One "opennet" way to do this is Haggle, which is the electronic equivalent of > asking around on a train for whether somebody has a book you want. We're only > interested in "darknet" approaches here. > > In principle sneakernet could be done purely on cellphone nodes - but bear in > mind the limited storage, problematic networking if you want some real-time > comms, and very high latency if requests have to be routed; routing would be > a major challenge with *no* realtime comms, since we can't swap locations > etc, and also long-term requests don't exist yet, though IMHO we could > implement *some* functionality for sneakernet tomorrow (opportunistically > filling datastores eg), and some more once we have bloom filter sharing. IMHO > the best way forward long-term for freenet is pi-like home-server boxes > combined with thin cell clients. > > Anyway ... before we can do the advanced stuff, the easy stuff is likely to > be useful.
IMHO we'd probably want to base the mobile client on the main fred source code so it can use the same transports, but have a special mobile-compatible mode. Apart from some difficulties with databases, porting Freenet to android should be straightforward as we only use the main APIs. If we require that each mobile client is tied to a "real" node, even if we're only using it for reference exchanges and to do requests through the "real" node, we will still need to maintain a connection to it; the easiest way to do this is to use the standard Freenet Node Protocol, so we can use Freenet to do this. Plus, if we have friend-of-a-friend connections, we can use them for connectivity; mobile connections are usually firewalled etc, so getting the latest IP address of the "home" node from its friends is a good thing.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Devl mailing list [email protected] https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
