Re: IID length text [was Re: Review of draft-ietf-6man-rfc4291bis-06]

[Brian E Carpenter <[e-mail]>]

In line...
On 17/01/2017 08:59, Fernando Gont wrote:
> On 01/14/2017 04:49 PM, Brian E Carpenter wrote:
>> A modest suggestion:
>>
>> OLD
>>    For all unicast addresses, except those that start with the binary
>>    value 000, Interface IDs are required to be 64 bits long.  Background
>>    on the 64 bit boundary in IPv6 addresses can be found in [RFC7421].
>>
>> NEW
>>    IPv6 routing is based on prefixes of any valid length up to 128 [BCP198].
>>    For example, [RFC6164] standardises 127 bit  prefixes on point-to-point
>>    links. However, consistent use of Stateless Address Autoconfiguration
>>    (SLAAC)[RFC4862] requires that all interfaces on a link use the same length
>>    of Interface ID. In practice, this means that to guarantee interoperability
>>    of SLAAC, a fixed length of Interface ID is necessary.
> 
> fixed as "all nodes on the link use the same IID length" or as in "a
> hardcoded IID length, as the current 64 value"?
> 
> If the former, I agree. If the later, I don't. 

You're right, that choice of word could be contradictory.
I suggest s/fixed/consistent/.

    Brian

> The 64-bit length seems
> to have a lot to do with embedding MAC addresses (IIRC, the IID length
> was something like 48, but then changed to 64 in response to EUI-64).
> Certainly, if you generate IIDs by embedding some number which has
> constraints (as a MAC address), you need a fixed length. OTOH, if you
> select the IIDs from a stream of bits (as RFC7217 does), then you don't
> care about the length (other than "as many bits as necessary such that
> the host density is low, and hence collisions of IIDs are reduced).
> 
> Thanks!
> 
> Cheers,
>