[3] in SIPB IPv6
setting up the 6bone router..
daemon@ATHENA.MIT.EDU (Bill Sommerfeld)
Wed Sep 29 19:20:11 1999
Message-Id: <199909292319.XAA06012@orchard.arlington.ma.us>
To: sipbv6@MIT.EDU
Date: Wed, 29 Sep 1999 19:19:44 -0400
From: Bill Sommerfeld <sommerfeld@orchard.arlington.ma.us>
Ok, here's what needs to happen, at a relatively high level..
Comments/questions/etc/ to me.
i'm prepared to do most of the preliminary steps myself in the next
little while; if you want to watch/help, let me know..
(there may be "too many cooks" issues for some of these, but i'll try
to document what i've done ...)
"Please explain this buzzword" is an appropriate question here..
1) set up hardware
2) set up software
3) bring up ipv6
4) get 6bone address space from an upstream tunnel provider.
5) set up "upstream" tunnels.
6) set up "downstream" tunnels
7) set up dns inverse-map zones.
8) set up more v6 systems.
9) start playing with apps over v6..
1) set up hardware:
- moving the machine to the right place in the rack
- connect power, ethernet, and console.
i have parts for the console cable right now, but want to double-check
things with an RS232 breakout box before building it..
2) set up software/bring up v6:
- boot and install a recent NetBSD-current/sparc snapshot.
- copy in -current netbsd sources
- build a -current kernel with ipv6 and with lots of gif*
interfaces (these are used for tunnels).
- boot the new kernel
- rebuild -current userland to match the new kernel.
- build & install software packages from pkgsrc:
- sshd (with inet6 support)
- emacs
- zsh, tcsh
- zebra (multiprotocol routing package).
- and probably a few others
4) get 6bone address space.
I'm working on this right now; i've got a couple queries sent
out to various 6bone nodes.
some background reading:
http://www.ietf.org/rfc/rfc2373.txt (ipv6 addressing architecture)
http://www.ietf.org/rfc/rfc2374.txt (aggregateable unicast address format)
http://www.ietf.org/rfc/rfc2471.txt (ipv6 testing address allocation)
once we have address space, we can then start assigning it to
subnets.
i'll send more stuff on this in a separate note.
5) set up "upstream" tunnels.
i'm working on this as well.
one issue here is dynamic routing and AS numbers.
if we just have upstream tunnel, it's trivial to do routing: just
point a default upstream, and have subnet routes for each subnet into
the right tunnel.
However, some sites will want to use BGP over the tunnel to learn
routes to us. we need to figure out which autonomous system number to
use here..
6) set up downstream tunnels.
This is generally simpler than the upstream tunnel, since "we" control
both ends. i'll write up specific procedures for this later.
7) DNS inverse-map entries.
(for address->name translation)
ip6.int is sort of like in-addr.arpa; the main difference is that the
address->dns name conversion is done at hex digit rather than decimal
byte boundaries. we'll eventually want to set up dns servers for the
inverse-map zone..
8) set up some other systems with v6 client code
9) start playing with apps over v6.
so far, i've only really used ping, traceroute, and ssh.. there are a
relatively small number of apps which run over v6 so far, but it's
relatively straightforward to port most things.
- Bill
