Re: draft-bourbaki-6man-classless-ipv6-00

[Fernando Gont <[e-mail]>]

On 06/09/2017 08:51 PM, David Farmer wrote:
>     >
>     >
>     >     That's not true at all. As an example: see if you can build a
>     >     network that uses a /99 and satisfies the recommendations in
>     section
>     >     8 or BCP 204.
>     >
>     >
>     > Yes, to meet the specific RECOMMENDATIONS in BCP204 /64 is necessary.
> 
>     I don't see anything in BCP204 where /64 is a MUST or even RECOMMENDED.
> 
> 
> It's not in the normative language, but it is part of the RFCs
> recommendations section, see below.

That comment, in non-RFC2119, is like "one possible way to do it".

That is, we don't need to comply with non-normative language in a
document that does employ normative language.



>     > It seems to me that extremely long prefixes like /120 or longer are
>     > likely unable to achieve the intent of BCP204 for any significant
>     number
>     > of hosts. However, anything in the range /112 or shorter should have
>     > plenty of address to achieve the intent of BCP204 for a quite
>     reasonable
>     > numbers of hosts, at least many hundreds of hosts.
>     >
>     > So, of the examples you provided, /123 is unlikely to achieve the
>     intent
>     > of BCP204, but /81 or /99, while they don't meet the specific
>     > RECOMMENDATIONS of BCP204 there doesn't seem to be any reason they
>     > couldn't achieve the fundamental intent of BCP204, granted with
>     without
>     > much sparseness or entropy in the assignments, but that's not
>     > specifically part of BCP204.
> 
>     Could you copy&paste the "specific recommendations" you're referring to?
> 
> 
> Second paragraph, second and third sentences of section 8;
> 
>    This can be achieved either by allowing the host
>    to form new addresses autonomously (e.g., via SLAAC) or by providing
>    the host with a dedicated /64 prefix.  The prefix MAY be provided
>    using DHCPv6 PD, SLAAC with per-device VLANs, or any other means.
>  
> Furthermore, I would concur that those are the only current standards
> track ways to achieve the desired result and they require /64 either per
> host or per subnet.

Exactly: the document from the subject line is exactly about forcing the
hardcoded /64 everywhere.



>     I haven't found any. And I see no reason for which the authors of bcp204
>     should have cared about a specific prefix length.

If the specific length was important, they should have cared. However,
the aforementioned bcp talks about being able to get multiple addresses.
The pool size is, for the most part, irrelevant.

You can get such multiple addresses with /64, /60, /56/ or /48, or even /32.

Pushing for the specific /64 value is not about a technical agument, but
rather about trying to convert an historical actifact into a technical one.

-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492