[Openswan dev] Error building klips-ipv6 (missing include?)

Harald Jenny harald at a-little-linux-box.at
Tue Oct 12 11:02:08 EDT 2010


On Tue, Oct 12, 2010 at 10:46:23AM -0400, Paul Wouters wrote:
> On Mon, 11 Oct 2010, D. Hugh Redelmeier wrote:
> 
> > | I have been working a bit on the listen code, and I was considering an
> > | option to just listen to ANY. I am not sure what the history was of not
> > | listening on all IP's whenever they become available to the system.
> > |
> > | Perhaps Hugh or Hugh can sched some light on that?
> >
> > Off the top of my head, I can vaguely recollect two issues:
> >
> > - we want control of which interfaces are used.  The only way was to
> >  control the IP addresses
> >
> > - we want to know the inbound IP address on each packet.  recvfrom
> >  doesn't give you that, so we nail down which IP by only accepting
> >  a specified IP address on each socket.  So we have to know the IP
> >  address.
> >
> > Note: there seems to be something newish called recvfromto that is now
> > half supported in the code, controlled by HAVE_UDPFROMTO.  But that
> > code currently ignores to "to" that recvfromto returns.
> 
> Yes, we should use recvfromto instead of recvfrom in all cases.

Won't this break backward compatibility?

> 
> The problem these days is that interfaces and IP's are quite dynamic, and
> our old-style interface handling is not very appropriate anymore these
> days.

Maybe we should take a peek at strongswan as it seems they are not facing this
problem (anymore).

> 
> Paul

Harald

> _______________________________________________
> Dev mailing list
> Dev at openswan.org
> http://lists.openswan.org/mailman/listinfo/dev


More information about the Dev mailing list