[bittorrent] Friend-machines-assisted approach to boosting one's BT download

Nikitas Liogkas nikitas at CS.UCLA.EDU
Mon Jun 12 01:12:15 EDT 2006


----- Original Message ----- 
From: "Elliott Mitchell" <ehem at m5p.com>
To: "Nikitas Liogkas" <nikitas at CS.UCLA.EDU>
Cc: <bittorrent at lists.ibiblio.org>
Sent: Sunday, June 11, 2006 9:14 PM
Subject: Re: [bittorrent] Friend-machines-assisted approach to 
boosting one's BT download


> >From: Nikitas Liogkas <nikitas at CS.UCLA.EDU>
>> ----- Original Message ----- 
>> From: "Elliott Mitchell" <ehem at m5p.com>
>>
>> > >From: Nikitas Liogkas <nikitas at CS.UCLA.EDU>
>> The objective of selfish peers is irrelevant to the matter at hand.
>> What the paper shows is that the attack you described is 
>> infeasible.
>
> No it doesn't. You were talking about a peer trying to get extra
> download. I'm suggesting a peer that doesn't care about downloading,
> merely destroying the swarm.
>
> The difference in objective is crucial.

I was trying to make the point that, no matter what the objective of 
the malicious peer is, the proposed attack is infeasible. That is to 
say that the difference in objective does not affect the 
(unsuccessful) outcome of the attack.

>> The main reason for the distinction between pieces and blocks is 
>> the
>> number of new piece HAVE messages that would be required if the 
>> blocks
>> were verifiable. The increase in the size of the .torrent file 
>> would
>> not be a problem for hosting servers.
>
> That would of been one of the original reasons, but the size of the
> .torrent file was also a reason. I will agree that advertising 
> posession
> of every 32KB block would be a bad idea due to HAVE overhead, but 
> this
> does not mean you cannot have hashes for every 32KB. You simply have 
> the
> granularity that you can advertise at be distinct from the hash
> granularity. There is no reason a protocol redesign couldn't make 
> those
> two distinct.

True. I have always been curious whether there have been any studies 
showing the impact on performance when varying the piece and block 
sizes. Are you people aware of any such studies?

nikitas 




More information about the BitTorrent mailing list