Re: [Help-gnunet] connecting to startpoint?, BM

[Christian Grothoff]

On 06/01/2015 12:52 AM, demos wrote:
> Hi GNU folks,
> 
> Does GNUnet work without connecting to: 131.159.74.67 and
> 2001:4ca0:2001:42:225:90ff:fe6b:d60 ?
> What is this connection for?

There are three possible connections you might be refering to.

When running 'gnunet-setup', we connect to gnunet.org for the NAT test.
gnunet.org runs 'gnunet-nat-server' and allows the connecting client to
check if it can receive inbound TCP/UDP packets on a certain port.
Basically a minimal STUN protocol to test hole punching.  Anyone else
(with a reasonably unrestricted global IP) could run
'gnunet-nat-server', and you can trivially reconfigure your peer to use
that other system instead.  Also, this only happens when you explicitly
click on 'test'.

Second, there is a "HELLO" hard-coded into the typical install (in
share/gnunet/hellos/).  This is simply to make sure that your peer has
ONE peer to start with and bootstrap.  We'd like more, if you're running
a peer at a stable address provide us your hello and we can ship those
as well.  If you don't like it, you can safely delete this file and the
peer will attempt other bootstrapping methods.

Third, there is a HTTP request to http://v10.gnunet.org/hostlist, where
a peer with insufficient HELLOs will try to download some via HTTP.
There are other hostlists which you could use instead, and if you want
to run one long-term please let us know and we'll be happy to add yours
to the list.

> Does GNUnet support: ping and iperf?

You should (theoretically) be able to run those over the GNUnet VPN, but
I have not tried (I guess I should). The GNUnet VPN does (theoretically)
support ICMP and UDP traffic.

> The Battlemesh benchmarks
> https://github.com/battlemesh/wibed-battlemesh-experiment/blob/master/packages/wibed-battlemesh-experiment/files/usr/bin/wbm-test
>  would need it if one could find those nasty bugs that result in
> "nodeinfo keeps crashing, nse keeps eating 100% cpu"
> and there comes nothing else in between from having GNUnet
> participating in the Battlemesh in two months.

The NSE CPU load is not a bug:

https://gnunet.org/faq-page#t10n2119
https://gnunet.org/bugs/view.php?id=3803

For testing in a testbed, you can get rid of it by changing

[nse]
WORKBITS = 0

But please do not do this in a production system connected to the rest
of the network.

> http://battlemesh.org/BattleMeshV8

Looking forward to it ;-).

-Christian

signature.asc
Description: OpenPGP digital signature