[bittorrent] Questions related to mainline DHT specs...

Patrick Moor pmoor at student.ethz.ch
Tue Aug 22 05:36:44 EDT 2006


> Hey Patrick, I ran across one other things I thought I would see if
> you ran into either - over the course of running my DHT for several
> hours (let's say 5), I get about 60 or so announce_peer messages -
> now here's the weird thing, not a single one looks valid.  Either
> they have a token set with a length of 0 (obviously invalid), or with
> a length of 20 (correct), but the contents are nothing but 0's - have
> you seen that before?  I doublechecked, and I'm setting the token
> correctly in response to the get_peers, but it doesn't appear to be
> getting returned - any thoughts?
Hm, this is weird. I usually see _way_more_ announces. I once let my dht
node run for 4 or 5 days and I had well over 100,000 announces from
distinct IP addresses stored. However, I _do_ see these invalid
announces as you mentioned with all zero tokens or empty ones as well.
Broken clients again, I guess...

Maybe you should use ethereal/wireshark once more and capture dht
packets the mainline client sends and receives from the network to
compare them bit for bit to the get_peer responses your own client sends

> Also, I've had peers that disclose different node ID's - for
> instance, I get a response from a get_peers with an ID of 0x6A46...
> but when I ping it the response ID is 0xD916...  So there's no
> consistency in the node ID.
maybe trying to fill up your routing table with bogus entries...
mpaa/riaa again ;-)

> I'm also experimenting with not adding peers to the list that respond
> on a different ip/port than what the request was made on, since these
> seem to always be the million node ID machines.
yep, I guess that's a good idea.


More information about the BitTorrent mailing list