[6bone] Re: routing concern

[John Fraizer]

On Wed, 31 Jul 2002, Ronald van der Pol wrote:

> On Wed, Jul 31, 2002 at 10:27:49 -0400, John Fraizer wrote:
> 
> > Until that happens, if a peer is
> > unresponsive and crappy, depeer.  Again, just like v4.  If transit to site
> > X is better via A_B_C_X than via D_X, set up a route-map to pref the
> > A_B_C_X route.  Again, just like v4.
> 
> This is not the problem. The problem is A_B_C_X is looping between B and C
> and D_E_F_B_C_X is also looping between B and C. Changing peers from A to D
> does not help.
> 

Ia a routing loop exists, it's because someone is misconfigured or has a
broken BGP implementation.  It is NOT because the peering session is over
a tunnel vs a native V6 connection.  If B and C are looping, contact B and
C.  If they don't respond, call them out on the 6bone list.


> > Just because a connection between two peers is in a v4-in-v4 tunnel does
> > NOT make that connection a "6bone" connection vs "production" connection
> > and native v6 peering does not mean that that particular connection is not
> > a "6bone" connection.
> 
> True. Did I say otherwise?

You are trying to make distinction between the 6bone and production v6 at
the NETWORK layer.  There *is* no distinction and as Bill pointed out,
there probably won't be.  The distinction is in the network PREFIXES.

> > > I want an IPv6 network which is *at least* as reliable as the IPv4
> > > network is *today*.
> > 
> > Choose your peering partners wisely then.  No different than IPv4.
> > 
> > Rob from Sprint hit on something that I don't quite agree with.  He wants
> > to limit transit. 
> 
> As indicated above the problem is not with the peering. It has a lot to do
> with tunnels all over the world and transit by everyone.

Nobody is forcing anyone to accept or provide transit to anyone else.  The
fact of the matter is that right now, we have connectivity because it DOES
exist.

> 
> > Unless we're all going to establish direct peering between
> > each other (that scales just wonderful, doesn't it?) it is wise for us to
> > provide transit to each other (pTLA to pTLA).
> 
> How does that compare to v4?
> 
> 	rvdp

TRANSIT-AS's peer with TRANSIT-AS's.  They exchange customer routes.  Some
TRANSIT-AS's are CUSTOMERS of other TRANSIT-AS's.  Granted, the world
would be a better place if we ALL peered with EVERYONE else over native
layer2 (4v + v6).  Until that happens, transit is a fact of life.  Since
all pTLAs don't peer with all other pTLAs (native or otherwise) there must
be transit involved to establish connectivity.  If you don't want to
provide transit to other pTLAs, you don't have to.  If you don't want to
accept transit from other pTLAs, prefix-list filter them based on their
pTLA assignment.  If you don't like the connectivity between you and
site-X because it goes over an unreliable transit path, establish your own
native/tunnel peering with that site or their pTLA.


---
John Fraizer              | High-Security Datacenter Services |
EnterZone, Inc            | Dedicated circuits 64k - 155M OC3 |
http://www.enterzone.net/ | Virtual, Dedicated, Colocation    |