[6bone] RFC2772 rewrite

[Janos Mohacsi]

On Wed, 13 Nov 2002, Nicolas DEFFAYET wrote:

> On Wed, 2002-11-13 at 16:23, Rik van Riel wrote:
> > On Tue, 12 Nov 2002, Pim van Pelt wrote:
> >
> > > * Should we seperate the 6BONE cloud from the production IPv6 cloud ? This
> > > is not really an RFC2772 issue, but does come to mind.
> >
> > > ... I have heard sounds of operators filtering out 3ffe::/16 due to its
> > > impact on general availablility of IPv6 to their customers. This
> > > deserves discussion!
>
> http://www.noc.easynet.net/network/public/peering-ipv6.html
>
> ---
> Route Filtering
>
> We apply the following filters to announcements:
>
> Block Minimal prefixlength Maximal prefixlength
> 2001::/16 29 40
> 2002::/16 16 16
> 3FFE::/16 not announced
> ---
>
>
> What's the best ?
>
> - have a tunnel peer with a ISP who use production address
> - have a native peer with a ISP who use 6bone address

Your examples are not the dividing points:
Dividing point is, that you know (from your routing policy or the routing
policy of your peering partners), that routes/prefixes you get is
controlled in certain manner:

1. They can guarentee some kind of reachability (does not matter tunnel or
peer). The tunnel reachability is much harder if you cannot control the
IPv4 infrastructure...

2. The prefixes are aggregated accordingly.

3. You know how to control acceptance of the announced prefixes/routes.

In one word you have a decent peering policy.


>
> I prefer have a native peer with a ISP who use 6bone address !
>
> The address type (production or 6bone) should NOT be a peer criteria.
> Only allocation size (sTLA, pTLA, NLA) can be a peer criteria.
>
> You can have a bad peering with a peer who use production address and
> have a good peering with a peer who use 6bone address.
>
> > That doesn't make nearly as much sense as filtering out routes
> > that come via 3ffe::/16 sites, or simply giving these routes a
> > much lower preference so traffic always goes via production
> > sites, if there is a route via production sites.
> >
> > I'm all for some kind of separation between the experimental
> > and the production side of the ipv6 universe, especially if it
> > means that I can keep ipv6 connectivity in the near future,
> > when my ISP doesn't have ipv6 yet, but my own applications do
> > rely on it.
>
> What's a production site ?
>
> - a ISP who have RIR address (2001::/16)
> - a ISP who have native peering
> - a ISP who do commercial activities
>
> Many ISP who use production address (2001::/16) claim to do experimental
> stuff...

See my comments above.
	Janos Mohacsi