Re: [v6ops] A proposal for draft-ietf-6man-rfc4291bis-07 - FE80::/10 and IID length

[[e-mail]]

Alex,

>>>>> It would be good too if RFC4291bis said in a small place that
>>>>> "FEBF::1 is a Link-local address too".
>>>> 
>>>> No. It says the link-local prefix is FE80/10 + zeroes +
>>>> interface-id. If you have a 118 bit long IID you can create the
>>>> above link-local address, but that really does not have to be
>>>> stated.
>>> 
>>> But to let that happen, it should also refrain from picturing or
>>> writing that IIDs must be of 64bit length.  If it needs to give
>>> examples it should rather use "e.g. /118".
>> 
>> Write an IPv6 over foo document with 118 bit IIDs.
> 
> Invitation considered.
> 
>> There is a good reason the prefix-length of the link-local address
>> space is explicitly specified.
> 
> Well, I think there is no such reason.
> 
>> If all nodes on a link didn't agree on this value, you would get into
>> problems with protocols using the link-local address.
> 
> Well, I dont think so.
> 
>> E.g. address resolution for the default router,
> 
> Well, address resolution: the default router sends NA including its source address.  It does not include a plen.
> 
> If the Host does not want to wait for the NA it will send an NS; this NS
> does not include a plen either.
> 
> I dont know where do you see the need to agree on a plen for the LL
> addresses?  Can I know?

https://tools.ietf.org/html/rfc4861#section-5.2

Has a good introduction to the topic.

>> or sending packets from DHCP server to host...
> 
> Well... it's the same: NS and NA are needed, and these dont use a plen.
> 
> Besides, LL addresses are only valid on a link; so there is no need to perform some matching in the tables to identify whether should go to rt table or to neighbor cache (which would need plen, or otherwise exact match).

on-link determination also applies to link-local addresses.

Are you by any chance talking about how you _would_ have liked the IPv6 specifications to be, as opposed to how they actually are?

Ole