Re: RFC4291bis

On 08/08/2018 20:51, Mark Smith wrote:
> On Wed., 8 Aug. 2018, 09:39 Brian E Carpenter, <brian.e.carpenter@gmail.com>
> wrote:
> 
>> Hi,
>>
>> Assuming that we adopt draft-farmer-6man-exceptions-64
>> for the standards track or BCP, I suggest that we should
>> also update draft-ietf-6man-rfc4291bis as proposed below,
>> and plan for the two documents to be published as RFCs
>> simultaneously:
>>
>> OLD:
>>    Interface Identifiers are 64 bit long except if the first three bits
>>    of the address are 000, or when the addresses are manually
>>    configured, or by exceptions defined in standards track documents.
>>    The rationale for using 64 bit Interface Identifiers can be found in
>>    [RFC7421].  An example of a standards track exception is [RFC6164]
>>    that standardises 127 bit prefixes on inter-router point-to-point
>>    links.
>>
>>       Note: In the case of manual configuration, the Prefix and
>>       Interface Identifier can be any length as long as they add up to
>>       128.
>>
>> NEW:
>>    Interface Identifiers for stateless address autoconfiguration
>>    are 64 bits long except if the first three bits
>>    of the address are 000,
> 
> 
> 
> 
> or when the addresses are manually
>>    configured,
> 
> 
> So this reads as though manually configured addresses will never have 64
> bit IIDs, as it is stating it is an exception to the 64 bit IID part of the
> sentence.
> 
> Actually, reference to manually configured addresses doesn't seem relevant
> at all, because presumably "stateless address autoconfiguration" is
> referring to SLAAC, and SLAAC is an automated address configuration method,
> not a manual one.

But there are potentially many automated methods, such as DHCPv6,
where the concept of a fixed IID length is unnecessary. However,
people previously objected to wording in which the standard length
was formally only applicable to SLAAC.

> There seems to be a bit of conflating address configuration methods and
> address properties.

Yes.

> In some cases the address properties are specified by the address
> configuration method so that an automated address configuration method has
> default address properties it can assume out of the box, and will work for
> common and ideally all cases without requiring a network operator to
> bootstrap the network i.e. SLAAC.

Yes. 

> (With a large enough IID space, the constraint on the number if nodes that
> can be attached to a link is shifted from being the available and
> affordable layer 3 addresses to the underlying link layer constraints.)
> 
> In the case of stateful DHCPv6, an assumed default prefix size/IID size is
> also useful to minimise the operator effort to configure it.

Useful, but not required.

> The operator
> however has decided to accept the expense of setting up a stateful DHCPv6
> infrastructure. The cost of overriding defaults is incremental.

Sure. But if you ask: "Could we invent a method of autonomously assigning
addresses that does not depend on a standardised IID length?" of
course we could. So maybe the correct formulation is:

 Interface Identifiers are 64 bits long for any method of
 address assignment that depends upon a standardised length.

I'll stop there, because in my mind that is a significant
change that might just allow us to stop digging this hole.

    Brian

> 
> Manual addresses are totally manual, so the operator is required to decide
> an address structure. There is no opportunity to reduce operator effort
> through reasonable defaults, and going without the benefits of reasonable
> defaults is the choice the operator has made by going with manual
> configuration.
> 
> 
> Regards,
> Mark.
> 
> 
> 
> 
> 
> or by exceptions defined in standards track documents.
>>    The rationale for using 64 bit Interface Identifiers can be found in
>>    [RFC7421].  An example of a standards track exception is [RFC6164]
>>    that standardises 127 bit prefixes on inter-router point-to-point
>>    links. The relationship between prefix length and Interface Identifier
>>    length is discussed in [I-D.farmer-6man-exceptions-64].
>>
>>       Note: In the case of manual configuration, the Prefix and
>>       Interface Identifier can be any length as long as they add up to
>>       128. In all cases, routing and forwarding processes must be
>>       designed to process prefixes of any length up to /128 [BCP198].
>>
>> Comments:
>>
>> 1. This change makes it clear that the 64-bit IID length is for SLAAC.
>> 2. It specifically send the reader to draft-farmer for details.
>> 3. It underlines that routing must work for any length prefix.
>>
>>     Brian
>>
>> --------------------------------------------------------------------
>> IETF IPv6 working group mailing list
>> ipv6@ietf.org
>> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>> --------------------------------------------------------------------
>>
> 