Re: [7] [Libbt-devel] added header files that will be installed

[Kevin S. <ke...@an...>]

Peter Stuge wrote:

>On Tue, Mar 08, 2005 at 11:35:31PM -0800, Kevin Smathers wrote:
>  
>
>>Peter Stuge wrote:
>>    
>>
>>>No, listenport is not required, if we're going to do the networking
>>>in libbt I strongly suggest that we only use a single port per
>>>application, and by default one in the dynamic port range as defined
>>>by IANA, 49152 through 65536. Start from the bottom and try
>>>allocating upwards.
>>>      
>>>
>>People who operate firewalls generally prefer to know what it is
>>that they are opening up a hole for.
>>    
>>
>
>We know that we can't tell anything about the traffic we're opening
>holes for unless we actually inspect it. A simple stateful packet
>filter will not and can not know what the application is doing and
>assuming that a certain port only carries certain traffic is silly.
>
>  
>
The way that the protocol specification reads, it is difficult to 
imagine an effective proxy for BitTorrent.  It really is designed solely 
as a peer to peer protocol.  Traversing a NAT gateway and firewall by 
opening up the appropriate port is the most widely supported way to 
connect on a peer to peer level.  A specialized proxy or NAT plug-in, if 
an appropriate one is constructed, would improve security by adding a 
boundary layer, but experience with SIP and other gateway protocols 
tells me that even hard pushing by well monied champions can fail to 
penetrate deeply into the router market, and BitTorrent is grass-roots.  
If you want to pursue it then more power to you, but I have different 
priorities.

>>>And most importantly a dynamic port fits the application. This is
>>>not a server that peers will try to connect blindly to as a result
>>>of a user command, instead peer lists come from the tracker and so
>>>ports do not matter.
>>>      
>>>
>>Bram has a lot of experience with actual deployments.  Pretend that
>>he knows what he is doing.
>>    
>>
>
>You can't seriously mean that arbitrary allocation of ports in the
>IANA reserved range by individual software manufacturers is a good
>thing?  Further, someone mentioned a post by Bram in which he
>recognized that this particular part of the protocol documentation
>was unfortunate. It's a non-issue AFAICS, since all clients always
>tell the tracker which port they're using.
>
>  
>
Actually I do think it is a good thing that software developers allocate 
ports in the absence of IANA's blessing, but that is a different and 
very philosophical discussion. 

>As for firewalls, that's another issue, I think the best solution
>would be for Bram to get a port allocated by IANA and then all BT
>software could use that as the default. Until then, nothing we do
>in libbt will be of much help for firewall operators..
>
>Plus; users should be able to pick the port if they want.
>
>  
>
On this we agree; without a doubt users should be able to pick their own 
ports.