[afnog] /127 ??

[Mark Tinka]

On Tuesday 17 November 2009 12:02:41 pm Randy Bush wrote:

> draft-kohno-ipv6-prefixlen-p2p-00.txt

I've gone through the draft - the points are noteworthy in 
consideration of a /127 on point-to-point links. Just some 
thoughts on the outstanding issues re: RFC 4443:

> 1) A rule described in ICMPv6 [RFC4443] indicates that a
>    Destination Unreachable (Code 3) message should be sent
>    by a router rather than forwarding packets back onto
>    point-to-point links from which they were received if
>    their destination address belongs to the link itself. 
>    Checking all traffic for this condition is likely to
>    affect performance.

Since ICMP Destination Unreachable messages would be 
generated by the router's control plane, one can implement 
ICMP rate limiting to restrict the control plane's exposure 
to handling these kinds of response messages to the extent 
that control plane functions could be affected.

Some service providers may outright drop Destination 
Unreachable messages (probably not a good idea as this 
breaks things), but perhaps rate limiting provides some 
middle-ground between care-free processing and outright 
dropping of said messages.

Of course, rate limiting ICMP messages opens up issues with 
customers who may relate high ICMP packet latency or packet 
loss to data plane performance as well, particularly in 
distributed routing platforms where one may have nothing to 
do with the other.

> 2) There could be a case that a packet needs to be sent
>    back onto point-to-point links from which they were
>    received. For example, LER (Label Edge Router) could
>    just forward the packet solely based on its label
>    without IP resolution.  In this case, if the
>    destination was the LER's egress interface, then the
>    downstream router would do an IP lookup and sent back
>    to the interface.

Cases where service providers would run MPLS with their 
customers are very few & far between (and should be close to 
nil since alternative solutions now exist, to address cases 
where such scenarios would have been required).

Typical cases for running MPLS on point-to-point links would 
be for core (PoP-to-PoP) circuits, where more control of 
traffic forwarded over the link would be possible.

But yes, it is understood issues affecting this scenario 
could be broader than just MPLS topologies.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <http://afnog.org/pipermail/afnog/attachments/20091119/96295393/attachment.pgp>