#dev 2021-01-19

2021-01-19 UTC
ShadowKyogre, [tantek] and [fluffy] joined the channel
#
[fluffy]
Yeah, that’d make sense.
#
[fluffy]
My very basic understanding of how magnets work is they rely on a huge distributed hash table which is seeded by a few initial endpoints which are somehow made known to the torrent clients (presumably by having an address hardcoded in them), and the “awaiting magnet metadata” phase is basically just your client sending out pubsub requests for anyone who knows anything about the hash.
#
[fluffy]
Someday I’ll have to look at the source for both Transmission and Syncthing to see how they do their things.
#
[fluffy]
I was briefly using Syncthing to sync my things but I found that adding/removing machines from the quorum was too annoying and I went back to using my Nextcloud instance. But I can still totally see some use cases for Syncthing.
#
Ruxton_
there is DHT nodes, they share and exchange hashes+seeds. So you create a torrent, share it to a node, that node exchanges with other nodes it knows and so on and so on infinite
#
Ruxton_
clients can also tell the nodes about all the hashes+clients they know about
ShadowKyogre and deathrow1 joined the channel; ShadowKyogre left the channel
#
nekr0z
The Syncthing developers have been discussing DHT for at least 4 years, but haven't done it yet. The general consensus seems to be it would be nice to have it, but the existing implementations don't quite fit, and writing a new one is too big a hassle. The thing with Syncthing is that an average Syncthing folder has way less peers than an average torrent, Kademlia file or some such other thing that DHT has been
#
nekr0z
successfully used for.
btrem left the channel