p2psp team mailing list archive

[Max Mertens <max.mail@xxxxxxxxxx>]

Hi Vicente,

On 08.06.2015 10:49, Vicente Gonzalez wrote:
> A remark. We can suppose that the splitter (I think, your server) will
> run in a public host, with all the port open.
In this test, the server can not be compared to the splitter, the name
'server' is misleading here. It is assumed that each other peer's public
endpoint is already known as I typed in the addresses as a parameter, so
the splitter part is not needed for this test.
The only difference between the server and client is that the server
sends UDP packets every 5 seconds, and the client every 1 second.
It is a bit surprising that e.g. RCN->PRCN connection does not work but
PRCN->RCN does, and sometimes one peer can see the other peer connecting
but cannot reply. I think that this asymmetric behaviour is due to the
chownat implementation (as server and client are not exactly equivalent)
and should not occur in a NTS implementation.

>     Is there already some kind of case differentiation implemented,
>     where the peers determine the type of NAT they are behind (using
>     the splitter), and automatically choose the Set of rules they need?
>
>
> No, but we can think that all peers will implement the overhead of the
> NAT Traversal Set of rules (I hope, with a negligible protocol
> overhead ... I mind, with a small number of small messages).
Ok, so all peers will use peer_nts.py?
How is TCP streaming handled by P2PSP, is this also planned to be
handled by NTS?

>     When using UDP to transmit data, where should lost
>     packages/datagrams be recovered? When using software like
>     udptunnel, the P2PSP classes could expect a reliable stream; but
>     as stated in the Lost chunks Recovery Set (LRS) of rules, P2PSP
>     uses UDP for communication, so if LRS is used, no extra package
>     recovery is necessary?
>
>
> No, LRS resends only video chunks that were lost from the splitter to
> the peer. LRS will not provide any kind of recovery for other UDP
> messages.
Ok. So the NTS implementation will have to recover lost UDP packages, as
not only video chunks are sent and all UDP messages are needed to
maintain the connection?

I wrote a UDP communication test program to test NAT traversal outside
the P2PSP code (to simplify testing and debugging), and will integrate
it into a NTS implementation in P2PSP as soon as it works correctly. You
can have a look at it here:
https://github.com/jellysheep/p2psp/blob/nts_tests/nts_tests/nts_tests.py
Currently it simulates two peers connecting using a splitter and then
communicating directly. I will now test if this also works with
different NAT types between them, and then extend the code to correctly
implement the NTS algorithm as specified in the paper.

And thanks for sending the paper, it is great to read it in "printed" form.

@Cristóbal Medina López: Thanks for the guide, I forked the repo and
added the NTS branch, and will work in this branch.

Regards,
Max