Hello Pim
Followup please to swissix@swissix.ch.
Because this is of general interest, I'll keep swinog looped in;
But the pollux.swissix.ch machine doesn't seem to have a route back to me (AS12859 for example).
As I explained in another mail already we are probably hit by a kernel-bug which is corrupting the ipv6 kernel-routingtable altough the bgpd-routingtable has a correct view of the world. We are working on this issue and will keep you posted.
I noticed that you are using an IXP prefix for your webserver.
Yes, that's correct.
Looking at SixXS' distributed looking glass[1], I see that still many ISPs are not accepting this /48 into their routing tables (and in my opinion, rightfully so). The IXP prefixes were not meant to host services in, rather only to make sure switch fabric is using a globally unique prefix (which comes in handy with stuff like DNS and ICMP unreachables and the like). I think that RIPE NCC has mentioned something to that effect when 2001:7f8:24::/48 was allocated.
RIPE's policy to this topic is clear: http://www.ripe.net/ripe/docs/ripe-256.html
--snip 4.0 Warning Networks assigned under this policy may not be globally routable. --snap
So, yes, you're right, there might be some networks that will not route our ipv6-prefix. Thank you for pointing this out. OTOH we had, have and most probably will operate some services for our members running in that ip-space (also reachable via ipv4).
Running these services in another ip-space would afford some entity to sponsor the service and hardware or housing while, at the moment, some peers are generously sponsoring some upstream and we can take care of the rest. Since the later option is a very short amount of work for the sponsors I guess it will stay this way. I think the board maybe will have to consider to stop running the services on IPv6 for the sake of availability. OTOH I'd be interested to hear if any of our peers is blocking our prefix since they are the main audience for our services...