> Wiktionary reuses OED content? I'm surprised the OED allows that.
Facts (etymologies) are not copyrightable. The exact same text can't be used of course, but it can be rephrased, or to some extend quoted under "Fair use".
Why is not reliable? The same has been said about Wikipedia for long time, but for some reason Wiktionary still does not get the same level of trust as Wikipedia. A lot of the etymologies on Wiktionary come from reputable sources such as the mentioned OED. In some cases there might be multiple conflicting sources and theories, and such complex cases are likely misrepresented by the automatic extraction tools.
Ah, I had a look at goios, I can imagine that it is a lot of work, the integration looks quite complicated and far from being finished, good luck.
About sharing bandwidth, was thinking about only enabling when the user is on wifi, but I think I'll use torrent more as a distributed mirror network than a "real" P2P system, which is difficult in a mobile context. People can easily "donate" bandwidth by seeding the data files needed for the app (using a separate torrent client though, not the app itself).
It's just libtorrent packaged up as iOS8 framework + a convenience API, available here: https://bitbucket.org/jberkel/torrada. Still WIP and not used in a production app, but hopefully soon.
It works well if you're on a non-throttled network and has support for local peer discovery which is great for testing (install from a local torrent seed, or another device). Apart from bandwidth saving it's also very convenient to let libtorrent handle consistency checks and resuming of transfers.
ipfs looks interesting, have you published your port?
I'm currently using this to build a decentralized distribution system to ship relatively big database files to iOS clients without having to go through S3 all the time. It works quite well but some providers throttle the bittorrent traffic (or worse, throw in RST packets) so I'm not sure how well this will work in practice. Also note that the S3 default seeders only upload at around 80kb/s, so you'll always need at least one external seed to get good performance.
i've replaced it with a new version which scales the node size according to the pagerank. i agree that the graph is a bit messy with all these connections floating around but it was also the intention to show the density of references between films.
it's explained in the blog post. it's the result of running pagerank on connection/references between movies, and graphing the connections between the top 100. you can see the pagerank when you hover over a title.
maybe the size of a node should indicate the rank to make it clearer.
That's definitely possible, GC takes up a lot of time (and stops execution). I haven't done any formal benchmarking, but everything seems to point to the missing JIT compilation at the moment. I'll do some more testing.
Facts (etymologies) are not copyrightable. The exact same text can't be used of course, but it can be rephrased, or to some extend quoted under "Fair use".