>>>>> Miles Fidelman <[email protected]> writes:
>>>>> Ivan Shmakov wrote:
> Since nobody else has responded, I'll weigh in.
Actually, Christian Grothoff has kindly pointed me to the
GNUnet's own DHT privately. Indeed, it seemed generic enough
for my purposes (and it accounts for restricted connectivity
networks!), so I've scanned through the Nathan Evans' thesis
[1], and deployed three GNUnet nodes on my hosts (at two
distinct locations.) Hopefully, I could take a closer look at
the DHT implementation shortly.
[1] http://gnunet.org/sites/default/files/NET-2011-08-1.pdf
> First off, you're asking two different questions:
>> Abstract
>> There're (several?) BitTorrent-specific DHT's, and the P2P anonymity
>> protocols (such as GNUnet) seem to (effectively) implement their own
>> DHT's. But is there any kind of a “universal” (i. e., transport
>> protocol-independent) DHT? And if not, why?
> This is really a non-issue. Some DHT's run directly over IP (UDP),
> some over TCP, some over some over something higher in the stack (say
> SSL, HTTP, or HTTPS) — but regardless of the transport protocol, the
> DHT algorithm maintains its own routing and overlay protocol that is
> inherent to the specific algorithm.
Actually, my point was that, as it seems, certain DHT's (such as
BEP 5 [2]) are tied to their respective /applications/, which
are file transfer protocols (such as BitTorrent; which I was
careless enough to call /transport/ protocols.)
I'm also interested in a file transfer protocol that accounts
for restricted connectivity scenarios, but without inherent
anonymization (which is unnecessary for me.) Unless I could
find such a protocol (and a decent implementation thereof), I'd
probably take an opportunity to design one myself, using a
suitable (preferably generic and deployed) DHT for peer and data
discovery.
[2] http://bittorrent.org/beps/bep_0005.html
[…]
>> What seems to be missing, however, is a “generic” DHT network that
>> could be used to search both the relevant metadata (such as .torrent
>> or Metalink files), and the peers participating in a particular data
>> exchange (and the respective protocols they support), using one or
>> more of an extensible set of identifiers (including BitTorrent
>> infohashes, GNUnet URI's, and the plain SHA-1, SHA-2, or SHA-3
>> values.)
> A "generic" DHT is a completly different issue, and seems to further
> break into two issues:
> 1. A common algorithm: Given that DHTs are still very much a research
> area, it seems unlikely that we're going to see one DHT to rule them
> all anytime soon. You pick one of the more mature ones (e. g.,
> Chord, Pastry, Kademlia, ...) for an application and go from there.
The point is that, if possible, I'd like to stick with an
already existing (and having reasonably wide deployment)
implementation. GNUnet seems like a good fit, but I haven't
tested it much so far.
> 2. A common interface, say an algorithm-independent URI scheme: This
> seems feasible, as the functionality of DHT's is very basic
> (store/retrieve/delete key-value pairs, maybe with a search function
> added). Such a scheme could be as simple as using the basic HTTP
> operations (GET/PUT/DELETE/?for searching).
ACK, thanks.
--
FSF associate member #7257
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers
