Hi Thomas,

On Sun, 07 Nov 2010 14:29:16 -0500
Thomas Narten <nar...@us.ibm.com> wrote:

> The document currently says:
> 
>    Routers MUST support the assignment of /127 prefixes on point-to-
>    point inter-router links.
> 
> I fully support this.
> 
> However, I believe that as far as routing is concerned, IPv6 continues
> to be based on CIDR. There is nothing special about the  64 boundary
> from a routing perspective. This, I believe the above should be
> changed to the following:
> 
>    Routers MUST support the assignment of arbitrary length prefixes
>    (including but not limited to /127s) on point-to-point inter-router
>    links.
> 
> I do not see any reason to restrict implementations to only supporting
> /127s prefixes.
> 
> Thoughts? In particular, I'd like to hear from operators as to whether
> they want the functionality of being able to assign subnets of
> arbitrary length, or whether it would be sufficient to only support /127s.
> 

if the goal of this draft is to fundamentally mitigate the ND cache
exhaustion issue, by preventing the ping-pong issue when ND NS/NAs are
switched off (as implementations are commonly doing on point-to-point
links), then anything other than a /127 on a point-to-point link won't
prevent the pong-point issue.

> Note: I am quite aware that stateless addr autoconfiguration uses
> 64-bit Interface Identifiers (IIDs) and that the addressing
> architecture says that addresses need to be formed using IIDs. However
> these requirements relate to the formation of addresses. I am aware of
> no architectural requirement (or justification) that says routing
> should only be done on the first 64 bits of an address, regardless of
> how an address was formed.
> 
> Note: This does not mean I am in favor of or am suggesting any changes
> to stateless addr config, or the IPv6 addressing architecture. I
> believe the above can be made without changing those documents.
> 

I think the Addressing Architecture RFC would need to be changed, as it
stipulates 64 bit interface ids, which I think implies a maximum of 64
bit prefix lengths, unless these /127s were only permitted within 0::/3
- the impacted text is in 2.5.1 of RFC4291 -

"For all unicast addresses, except those that start with the binary
 value 000, Interface IDs are required to be 64 bits long and to be
 constructed in Modified EUI-64 format."


Regards,
Mark.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to