Does Gnutella peers need liveness checking?
 

Hi there,

I wanna know whether Gnutella peers need liveness probing to check their neighbors online or not. If yes, how often is this done? If not, how does a Gnutella peer know that its neighbors are gone?

Thanks.

Gnutella Clients are connected via TCP connections, which are being kept online, and they Ping each otehr from time to time, but only, when they receive no data from their neighbours for some time (which doesn't happen so often :-) )

thanks. According to the protocol, Ping messages are only used for finding new neighbors. By liveness probling, I mean maybe the peers need to be knowledged if one or all of its neighbors are gone (for example, leaves or fails). Does such liveness checking incur a lot of system overhead?

That is something I can't tell you, as I don't program the code myself. But what I know is that there is a BYE message, which gets used most times, when you close connection.

I don't know what lost connections drain... I know that finding a new host creates quite some overhead, though.

But I assume, you have something in mind, don't you?