Re: [GNUnet-developers] On applications using GNUnet

[Christian Grothoff]

On 8/6/19 5:18 PM, Alessio Vanni wrote:
>> I don't know which documentation sentence in what manual you are
>> referencing here, but I think that documentation is simply wrong. CADET
>> is port-scan resistant, in that the peer will simply not send anything
>> back if the port is not open. In fact, CADET will accept your incoming
>> request into an internal table in anticipation that maybe in the future
>> a local application will open that port, and then establish the
>> connection (as the client might have just been a bit faster than the
>> service opening the port). So as a client connecting to a closed port
>> will just seem to take "a long time" (= forever), until and unless some
>> application opens the port, at which point the session is acknowledged
>> and properly opened. But by design you cannot distinguish between a
>> closed port and CADET actually failing to reach the target peer -- or
>> things just being slow.
> 
> In gnunet_cadet_service.h, the documentation comment for
> GNUNET_CADET_channel_create says:
> 
> /**
>  * Create a new channel towards a remote peer.
>  *
>  * If the destination port is not open by any peer or the destination peer
>  * does not accept the channel, @a disconnects will be called
>  * for this channel.
> 
> According to this comment, if the destination peer does not have the
> specified port opened, then at some point in time I should be notified
> about it by the `disconnects' callback.  However, if CADET works like
> you says, then this comment is wrong (or outdated.)

Yes, indeed it is wrong. I have corrected the comment in
1b0e7d3f5..6b23a8225

Thanks for pointing this out!

-Christian

signature.asc
Description: OpenPGP digital signature