Re: New Version Notification for draft-petrescu-6man-ll-prefix-len-17.txt

[Alexandre Petrescu <[e-mail]>]

Le 10/05/2019 à 12:38, tom petch a écrit :
> ----- Original Message -----
> From: <[e-mail]>
> Sent: Friday, May 10, 2019 9:30 AM
> 
>>>> Why OSPF(v3) for IPv6 does not use an IPv6 address in the neighbor
> ID,
>>>> and why does it use an IPv4 address instead?
>>> both OSPFv2 and OSPFv3 use a 32-bit instance ID, which is
>>> traditionally represented as an ipv4 address. It's not actually an
>>> ipv4 address.
>>
>> The 32bit instance ID responds to ping?
> 
> No.
> 
>> The 32bit instance ID is present in routing tables as next-hop?
> 
> No.
> 
> Think of it as a 32-bit *number*, not as an IPv4 address (because it
> *is not* an IPv4 address).

In that case, the problem is in the textual representation of a 32bit 
value.  An uint32 cast into a struct in_addr and further converted by 
inet_ntoa to be displayed as a '1.1.1.1' instead of the intuitive 
'16843009' is misleading.

Probably the operating system of that router could be updated to no 
longer display the instance ID as an IP address.

In terms of spec, it seems Router ID of RFC2328 is the culprit, because 
RFC5340 says:
>  When receiving an Hello Packet from a neighbor on a broadcast,
>         Point-to-MultiPoint or NBMA network, set the neighbor
>         structure's Neighbor ID equal to the Router ID found in the
>         packet's OSPF header.
[...]
>     Router ID
>         A 32-bit number that uniquely identifies this router in the AS.
>         One possible implementation strategy would be to use the
>         smallest IP interface address belonging to the router.

RFC5340 OSPFv3 IPv6:
> o  OSPF Router IDs, Area IDs, and LSA Link State IDs remain at the
>       IPv4 size of 32 bits.  They can no longer be assigned as (IPv6)
>       addresses.

It sounds as if it was possible to use IPv6 addresses as Router IDs 
before, but no longer.

Alex

> 
> <tp>
> Absolutely, it is a 32 bit identifier with no semantics at all.
> 
> Trouble is, an all too popular manufacturer decided that one way to
> specify it was to make it an IPv4 address as well - both 32 bit, what
> could go wrong? - sothere are many who now think that it is, even must
> be, an address that can be reached over the network.  If you have some
> of this kit, and an IPv4 network, then you may indeed be able to use
> those 32 bits as a source IP address or a destination IP address but
> that would be a case of a manufacturer deciding to go over and beyond
> the specification and making future development more complex.
> 
> Thank heavens the OSPF WG carried the right concept across to OSPF v3.
> 
> Tom Petch
> 
> Steinar Haug, AS2116
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
> 
>