On Friday 21 August 2009 20:28:53 vive wrote: > On Fri, Aug 21, 2009 at 06:35:16PM +0100, Matthew Toseland wrote: > > #5 Write a killer file-sharing application (88 votes) > > > > IMHO distributed search (based on infinity0's work), better data > > preservation, and *maybe* some element of secure insert on demand would go > > a long way to improving Freenet's usability for filesharing... > > IMHO this is currently unrealistic and should be clarified for the prospective > users: Freenet is not likely to work as a good file-sharing tool in the near > future. > > I recently completed work with two students where we made Freenet compatible > with > BitTorrent (insertion and search for torrent information/data in the overlay, > using > a standard BT client) The performance for file-sharing turned out to be 1-3 > kB/s > consistenntly ... even with a number of interested opennet peers querying or > inserting the data for some time.
I have seen much higher transfer rates on very popular files - search indexes for example. But 3KB/sec has been the average for many years. :| Nonetheless I believe we will be able to significantly improve it. Bloom filter sharing should help significantly, and so should scaling the number of peers with the bandwidth usage. There are various other options we have yet to explore also. > > I realize that I should back up these claims with facts, or graphs and > numbers. > But this is not published yet, probably to come.. How old were these nodes? > > I think that with current performance of Freenet it is really important to > control > and manage expectations. Having users with too high expectations will make > them > perceive something qualititative negative when faced with the actual > performance. > A lot of help in allowing free communication is possible despite not > providing heavy > services such as file sharing, interactive video streams, et cetera. I'm not sure. A good search system is useful even on the web, where big downloads are rare. And 3KB/sec allows you to fetch an ISO in 3 days, so it is usable if it's your only option. We may never be able to search, have an answer in 10 seconds, and then stream video, but that doesn't mean a good file search system isn't useful. As regards helping free communication, people in hostile environments (who don't often give to freenet or contribute code, but interested organisations might) need: - Security. Both network and local. We have done a good deal of work on local recently, and while our network security isn't great, it may be better than many of the alternatives. - Survivability. Darknet works where many other apps won't work. - Basic chat or microblogging. - More involved blogging probably. Right now we link to Thingamablog for this... - Sharing relatively small video files (20MB or so). - To work adequately in fairly poor network conditions. A lot of hostile environment users are on dial-up or have severe bandwidth limits. Often bandwidth to the outside world is severely limited, or external connections may be blocked altogether. An important part of dealing with this is to improve Freenet's general performance. IMHO broadly speaking improving the retrievability of rare data is important - because it is fairly bad right now, because in nasty places there may be difficult topologies, because Freenet ought to be for rare stuff... Some of the earlier items on uservoice do relate to performance. We are working on most of these things ... anything else we need? > > Cheers, > vive -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 835 bytes Desc: This is a digitally signed message part. URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20090821/a26d871c/attachment.pgp>
