Hacker News Reader

Comment by jacquesm | Hacker News Reader

How hard would it be to make a torrent:// protocol extension that allows web like content to be stored in torrents? Obviously that would only work for static content (unless the creation of such torrents could be used to preserve state) but this seems like an interesting development.

Decentralized hosting of content that can easily be browsed seems like a useful thing to have.

View on HN

In terms of accessing small resources (HTML pages), a client-server setup is by far faster than a P2P system. The Web is client-server specifically for this reason. Not to mention that the whole power of a client-server setup is the ability to have clients alter the state of the server. With a P2P system, propagating changes and making sure each peer has the correct copy, requires more time.

Check out this paper to learn more about the Web applied to P2P networks [1].

Li, Ruixuan, et al. "WebPeer: A P2P-based System for Publishing and Discovering Web Services." http://www-inf.it-sudparis.eu/~tata/papiers/sellami/P2P_Disc...

Client-server is faster if the server's upload speed can match the client's download speed, for all clients. Which implies a large, central server.

There are other parameters that are useful when evaluating a configuration besides just speed. Resilience is one of those, as is long term availability. Torrents seem to do quite well in the resilience department, I'm not sure how well they do when it comes to long term availability. Maybe someone has already studied that, it would seem to be the torrent equivalent of 'link rot', and quite possibly it is very high.

Sorry, what I meant by speed is: the ability to quickly access small sized resources such as HTML pages. The initial handshake required for P2P systems makes this process slower. Obviously for larger files where there are lots of peers serving the content, you can get better speeds.

Latency is also a factor.

But a hybrid server/P2P system has multiple potential benefits (P2P acting as a CDN for common libraries, sharing live video or simply uploading 'unhostable' content).

If P2P were a "first-class citizen", the initial handshake wouldn't take longer--it would be faster!

Suppose that my city's newspaper (or TV station) monitored the (socially accepted, widely adopted) P2P networks for reliable mirrors of their own content, then sent each user's browser hrefs to the mirrors, rather than their own site.

The more popular the content, the more likely that your next door neighbor's P2P system can stream it to you.

Since running such software is socially acceptable in this hypothetical future, your P2P software doesn't have to go through the 'discovery' step. One of your hostnames is 1234.main.st.ci.lincoln.ne.us (see http://owen.sj.ca.us/~rk/howto/usdomainname.html) and one of them is mirror5.journalstar.org.ci.lincoln.ne.us. Everybody on Main St already knows your IP address, and if you mirror a lot of content, they might even have an existing active connection... you see where this is going? :)

Check this out: https://github.com/jbenet/ipfs

Something like this already exists in the form of SyncNet.

http://jack.minardi.org/software/syncnet-a-decentralized-web...

This seems like an attempt: http://kyle.graehl.org/coding/2012/12/07/torrent-in-your-bro...

A few years back, I saw a proof of concept on PlanetKDE using magnet links in Konqueror to browse. I was unable to find it again.

https://web.archive.org/web/20120213102248/http://whilos.blo...

Unfortunately, I couldn’t find the video. It worked really well though. You could just open the magnet and drag-and-drop copy the files you wanted, or just open them and they’d start streaming.

I'd be concerned of this becoming an additional security risk. You just know someones gonna find a way to inject malware into it. On a home network it's not a big problem, but on a corp it'll be open season.

Surely this isn't a problem with checksums/hashes to verify the integrity of the files?

Which in fact Bittorrent already does as a core part of the protocol.

Reply on news.ycombinator.com