[6bone] separating IPv6 experimental from production traffic

[Rik van Riel]

On Wed, 21 Aug 2002, Jeff Barrow wrote:

> Even better, change 3) and 4) to:
>
> 3) experimental pTLAs announce whatever they want to, but the production
> peers all filter out incoming 2001::/16 prefixes from them.  This way,
> even if an experimental pTLA messes up and starts announcing all prefixes
> with their own ASN, or gets hit by that dreaded withdrawal bug, the
> 2001::/16 network will still be operational to itself--just the 3ffe://16
> prefix may be bad.
>
> Production pTLAs announce all prefixes to everyone, and listen from
> other production pTLAs without the 2001::/16 filter.
>
> This will, of course, require all production networks to be interconnected
> by way of production networks only.  However, this method should be much
> easier to accomplish, as only production pTLAs need to make configuration
> changes to set this in place by adding a simple filter, instead of relying
> on the experimental pTLAs to filter outgoing announcements (and trying to
> contact them all); we just have to deal with the production-level pTLAs.

This sounds like a _much_ better idea, since the ones having
to implement this idea (the production sTLAs) are the exact
same ones who will benefit from it.

A proposal like this barely needs peer pressure, pure self
interest is enough to get this thing implemented by everybody.

kind regards,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".

http://www.surriel.com/		http://distro.conectiva.com/