You are not logged in.
Pages: 1
Offline
I like it.
It might make sense to clarify that "seed"=1, "scrape"=1 and "noseed"=1 are supposed to be added to the "a" dictionary. And that "BFsd" and "BFpe" are supposed to be added to the "r" dictionary.
Offline
I updated the BEP, it now includes test vectors, some legacy handling, arvid's changes and some changes to the BF pseudocode definition:
Same url as before: http://infinite-source.de/az/BEP_33.rst
Az dev
Offline
I like it, too.
Offline
Sorry for the delay. I finally put up your bep on the site. It should be propagated and available soon.
Offline
It's still not there.
Az dev
Offline
http://www.bittorrent.org/beps/bep_0033.html
I like what I see so far.
Offline
http://azureus.sourceforge.net/plugin_d … ugin=mlDHT
My implementation is publicly available now.
Az dev
Offline
I'm very much undecided on this.
On the one hand -- it's certainly a very sexy piece of work. Merkle trees [edit: typo, I meant Bloom filters] are cool, and 8472 makes a rather elegant application of them. I feel that I've learned something new.
On the other hand, I'm not convinced that this extension is needed, and I'm also not convinced that we want to trust BitTorrent implementors to get this right -- I expect a lot of buggy impementations of this to surface.
On the gripping hand, this extension only makes sense if it is deployed by a significant number of implementations. 8472, what are the µTorrent fowks saying?
Last edited by jch (2010-04-05 14:05:18)
Offline
On the one hand -- it's certainly a very sexy piece of work. Merkle trees are cool, and 8472 makes a rather elegant application of them.
You either mentioned merkle trees by mistake or posted in the wrong thread.
Anyway, bloom filters are well known and easily implementable.
It is probably the best solution for distributed swarm size estimation.
Nice work The 8472 !
Offline
On the gripping hand, this extension only makes sense if it is deployed by a significant number of implementations. 8472, what are the µTorrent fowks saying?
As far as I've understood it they're interested but it's a low priority thingy to implement.
On the other hand, I'm not convinced that this extension is needed, and I'm also not convinced that we want to trust BitTorrent implementors to get this right -- I expect a lot of buggy impementations of this to surface.
I agree that this is a serious concern, considering the sloppyness of various DHT implementations.
That's why the provided bloom filter implementation is as simple as possible and i've also provided a test vector.
All we can do now is see how it'll work out in practice, if it actually gets picked up.
Az dev
Offline
@The 8472: The use of the terms "seed" and "peer" is a bit confusing in this BEP. I would imagine that both downloaders and seeds would be considered peers, but it sounds like in the BEP, "peers" are limited to just downloaders. I think it would be clearer if you would use the term "downloader" instead.
Offline
Yes, Peers is intended to mean downloaders in that BEP. To express "members of the network" i usually say Node. A bit of historic nomenclature.
I could change/clarify the prose if you like, but i'd prefer not to change the protocol parameters themselves since i've been implementing it for a while.
Az dev
Offline
Pages: 1