GNUnet | Hacker News Reader

GNUnet

gnunet.org

193 pointscdwhite46 comments

I wish one of these networks would take off, be it GNUnet, OneSwarm, I2P, freenet or any other, I don't care.

The thing is, performance must be at least decent if the authors of these protocols want people to start using them.

i2p is quite decent. torrents often exceed 100 kilobytes/s which to me IS fast for lazy downloads.

ajax probably won't ever be fun but I could not care less about that. :D

I find that I like I2p quite a bit, but I'm not as fond of the reference implementation. It's in Java, which makes running it on small systems more difficult, and it routes traffic by default.

Routing by default GREATLY improves the network, but may be undesirable/unrequested behavior by people.

Yeah. I want to host an i2p site on my AWS free tier account but can't because it will exceed the 15 GB bandwidth limit due to all the volunteer routing.

You can disable/lower routing. There is also bandwidth limiting. I don't recommend disabling the routing entirely, however.

"small systems"; I get that it can be hell-like on very small systems but the Raspberry Pi works great.

There's a thread on zzz.i2p about it.

It's ENTIRELY possible that that I didn't tune it correctly, but when I was looking at it as a routing underlayer for Tavern.com, it didn't leave me a whole lot of resources for running my apps.

If you're interested in running networking applications on small hardware, you may be interested in the Beaglebone. It's more expensive, but vastly outperforms the Pi with regards to network I/O. The Pi is awesome for video, however.

For you I2P users, What are you using for P2P software, I2PSnark, iMule, or I2Phex? I tried it a few years ago, but didn't have much success in finding anything. Any tips for setting up, searching, performance tweaks?

Last I read about it the performance was directly linked to the size of the network. In Freenet, if I remember correctly, larger number of nodes with the same content is proportional to the speed of the content download. And the content spreads to other nodes automatically, so if the network is big and demand for the content is big the content becomes very quickly downloadable (it's both "near" in terms of hops between nodes and there are more nodes with better bandwidth).

So I think that such networks will be niche until enough governments try to censor internet at once. This will make enough people interested in networks like Freenet to make their performance ok. I would like for Freenet to be adopted much faster and on a greater scale without some cataclysm for free speech, but it doesn't look like it's possible :(

So, for the uninformed like me, this is a more anonymous form of torrents?

Not really. The emphasis in freenet is on crypto and anonymous distribution / plausible deniability, not moving huge amounts of data. What parent says is that content gets easier to reach with more nodes compared to average FN content, not that it will (likely ever) be faster to download than with BT or similar.

Right, but in the best case (the content is stored on all neighboring nodes in FN) the only speed difference would be in crypto algorithms used and protocol peculiarities (handshakes, key exchange or something like this). I think? I mean, if the content is not further than one hop, the Freenet functions very similarly to torrent. Or am I mistaken?

Those "protocol peculiarities" are massive performance hogs.

More like torrent network where everyone uses TOR, I think.

I wish there were an easy, drop-in lib that would give me swarm/peerless networking like this, for Mobile platforms.

Even a Lua implementation would be enough for me.

WebRTC http://www.webrtc.org/

While currently only available in-browser (Chrome and Firefox Nightly), it is actually a C++ Library and you can build native apps as well.

The links you're posting are broken, ending with "/".

Ah, very nice hint, thanks for that! I see here:

https://code.google.com/p/libjingle/source/browse/trunk/talk...

That in fact your idea has a fair bit of merit. I will add "investigate adding libjingle to MOAI" to my MOAI to-do list, because .. it seems do'able.

Agreed. These systems have been around for a while, but as we see, haven't been widely adopted. So why? What's a good theory as to why these ones haven't taken off? If we can identify why these haven't taken off we can build the next one better.

  What's a good theory as to why these ones haven't taken off?

Bittorrent got there first.

Could someone explain to me how these protocols are superior to Bittorrent?

Your definition of "superior" may imply vastly different trade-offs. If you want files fast, then they are quite inferior to Bittorrent. The aforementioned protocols can vary a bit in what they do well, but in general the goal is to protect the identity of participants from hostile peers on the network. With Bittorrent, the goal is to transfer files efficiently with no regard for privacy, and it's very good at it.

It's harder to sue someone transfering a file with these protocols, however it's almost certainly slower to get a file with these protocols.

Then there's the network effect. BitTorrent is more popular, ergo there will be more files available

More that bittorrent was significantly faster and more convenient for users. In general, the more anonymous you want to make a network, the worse the tradeoffs involved will favor performance. Tor, for example, is designed to be low-latency, such that you can actually use instant messaging etc... but if an entity has the ability to monitor enough onion routers simultaneously it is actually trivial to pinpoint the origin of a message. This can be countered with traffic mixing, random delays, fake traffic, etc... all of which significantly increase latency and reduce bandwidth.

GNUnet is IME the only one whose performance actually is decent.

Key is efficient caching. So there are more seeds than leechecrs. This is also true with Freenet and one of the points which I really like Freenet & GNUnet versus other protocols. Older versions of GNUnet didn't have cache eviction at all which I found to be major longterm issue, but they fixed it after discussing the topic. Btw. Bittorrent could be also modified to provide caching, but currently it doesn't and that wastes a lot of bandwidth. Tor doesn't use caching either, and wastes even more bandwidth.

How would you implement caching in BT? I mean, who would cache what?

I like Retroshare.

http://retroshare.sourceforge.net/

Fixed link: http://retroshare.sourceforge.net

There is a new bug in HN, posting a trailing slash on an URL gives /.

Remove the trailing slash and your link works.

Disregard this, i'm just curious

http://www.example.com

http://www.example.com/

http://www.example.com//

You can add # which seems to make it ignore the trailing slash:

http://www.example.com/#

What I thought was bad about freenet was that it was/is every-imaginable-goody-of-encryption all in one application blob. It seems like what you would want instead would be peer-to-peer software with series of optional encryption layers. Getting adoption with faster unencrypted layers would allow user to start using encryption easily when "things get rough".

Perhaps some of the other things are closer to this.

Encryption is cheap nowadays; I doubt making it optional would do anything but reduce the effectiveness of the system.

Would performance increase with average internet speed (upload), right now it's ~5.34 MBit/s[1] which probably gets worse when hopping through several nodes. It's a shame Ookla doesn't provide latency statistics. I'm sure ping is a big factor too.

Might be interesting to see if a few computers strictly on a LAN could communicate quickly over networks like these or if the actual software is slow.

[1] http://netindex.com/upload/#

Now would be a good time to give variuous good organisations cash.

EFF is a good place to start. The Internet Archive also has helped document some of the lies made by government officials .

The EFF's view on exploit sales and forced disclosure is questionable at best.

https://lists.immunityinc.com/pipermail/dailydave/2012-Augus...

Their stance is limited to opposing 'green hats' who are selling to customers that don't intend on fixing the flaws (buying a hacker's silence) and selling to governments whom intend on using the exploits for clandestine operations.

Now would be a good time to give variuous good organisations cash.

I just sent the EFF $100.00 earlier this evening. Needless to say, I concur wholeheartedly with this sentiment. Slacktivism and clicktivisim or whatever are nice, but if you aren't in a position to do anything else, sending cash to advocacy / activism groups like the EFF is probably the most straightforward way to advance causes you support.

For my money, so far this year, that means supporting the EFF, NRA, Gun Owners of America, Libertarian Party, Jews for the Preservation of Firearms Ownership, Second Amendment Foundation, Grassroots NC, and the International Mountain Bicycling Association. I think the FSF will be getting a donation from me soon as well. Probably the ACLU as well.

It sounds like you support quite a few causes, but I think the EFF and cyclists need your money more than the NRA does.

Well, we all have to make choices, and that's the great thing about living in a free and open society: We get to actually make those choices. I can understand why you might feel the way you do, but I consider defense of the 2A and gun rights to be one of the most critical issues of all, so a slightly greater portion of my disposable "support causes" dosh goes to gun rights groups. But that's OK... I'm just one man, and no one man by himself can do it all. I just do my part as best as I can. There are other people (perhaps yourself) who will do it differently and maybe send the EFF the bigger check.

> consider defense of the 2A and gun rights to be one of the most critical issues of all

Would you mind explaining your position on this?

Ya know, honestly, I don't think this is really the place to get any deeper into that. I mean, I don't mind talking about it, but it's not really on-topic here at HN. And as soon as I give you an answer, somebody is going to fire back telling me I'm wrong, and why... and then somebody else is going to chime in, and we're going to be rehashing a debate about guns that's been argued over and over and over and over and over and over and over and over and over and over and over and over again...

If you're really interested though, shoot me an email and I'll be happy to discuss it "out of band" w/ respect to Hacker News.

The list of publications is really impressive:

https://gnunet.org/bibliography?f%5Bkeyword%5D=2

Reply on news.ycombinator.com