[pisa-src] Re: r1008 - trunk/include

On Wed, Sep 30, 2009 at 05:12:09PM +0200, Thomas Jansen wrote:
> 
> Log:
> MTU tweaking part 2. Even bigger buffers for testing.

Hmmmm...

> --- trunk/include/global.h    Wed Sep 30 16:59:05 2009        (r1007)
> +++ trunk/include/global.h    Wed Sep 30 17:12:09 2009        (r1008)
> @@ -34,16 +34,20 @@
>  
>  /**
>   * Size of the MTU for our tunnel device
> - * The MTU of 1000 was found out by trial and error - 1050 didn't work for me
> - * - neither via hipd nor via std-ipv6 connection (PMTU in both cases was
> - *   1500)
> + * 1370 (MTU of dummy0) - 20 (IPv4) - 8 (UDP) - 6 (MAC) = 1336 (payload).
> + *
> + * Since the kernel fragments PISA packets that are too large for the dummy0
> + * interface, we could actually trade off processing of PISA packets against
> + * fragmentation. For further experiments, you can use
> + * "ifconfig pisaTunnel0 mtu $MTU" on the command line after pisa{c,s}d is
> + * started.

This is something else entirely.

Mixing unrelated changes is not good.  But mismatches between commit
content and log message are supremely evil.  Please don't do it and
please fix the commit message.

Diego

Other related posts: