Re: A common problem with SLAAC in "renumbering" scenarios

[Jan Zorz - Go6 <[e-mail]>]

Dear Barbara,

On 04/02/2019 16:41, STARK, BARBARA H wrote:
>> However, RFC7084 is Informational and not Standard. I think this
>> should make it into Standard. I've seen many CPE vendors saying "we
>> found relevant standard documents and implemented whatever was a
>> MUST and shipped the product."
> 
> RFC7084 has no business being labeled an "Internet Standard", and its
> ability to impact CE router vendor development decisions is
> completely unrelated to how it is labeled.
> 
> It has no business being an "Internet Standard" because CE router
> requirements are derived/designed based on access architecture. CE
> router requirements do not, cannot, and will not ever drive access
> architecture. Whenever access architecture changes, CE router
> requirements change. Some of the changes that we saw from RFC6204 to
> RFC7084 were because ISPs decided they wanted to do different things
> in their access networks than what they originally thought (DS-Lite,
> 6rd, and some other items). So the requirements had to adjust. Now we
> have ISPs not agreeing on "IPv6aas" architecture. So the CE router is
> being asked to do all of them. What if 5G/wireline convergence
> becomes a reality? CE router requirements will change. Trying to
> standardize something that is outside IETF's control is unrealistic.

Agree, to some degree. BRAS/BNG behaviour and CPE behaviour when 
assigning IPv6 PDs is the same and should be predictable wherever you go 
and whichever transport technique you use. Mobile/3gp is outside of the 
scope of this conversation. For others we could standardize how things 
should look like. After that - it doesn't matter how access changes, the 
principle is always the same.

> 
> CE router vendors (and ISPs) really don't care whether something is
> labeled as "Standard" or "Informational". If it's well-defined and
> useful, that's all that's needed. L2TP is "Standards Track". PPPoE is
> "Informational". Which one do ISPs use in the access network, and
> which is implemented in CE routers? PPPoE. Why? It was a good tool
> for what was needed, and the RFC is easy to read and implement. L2TP
> was not the right tool.

True. However what I see with many CE vendors (and not just CE vendors) 
is that they implement just MUSTs from Standards tracks and ship the 
product. Some of them actually sees RFCs as something that they must do 
to satisfy that grumpy customers that senselessly demands standards and 
interoperability. Yes, some of them do that, and the cheaper the CE 
gets, more they see RFCs and MUSTs as a burden to their creativity and 
imagination in solving access problems.

That's where the desire for Standards document comes from. Otherwise 
it's just a good wish. There are exceptions, but very rare (like PPPoE, 
as you mentioned).

> 
> Are there any major CE router vendors who haven't tried to implement
> all the "MUST" requirements of RFC7084? Not that I'm aware of. Some
> may not have done it right, but that's not the same as not trying.

RFC7084 is a great document and could easily be a Standards track.

> 
> I'm opposed to any attempt to turn CE router requirements into a
> "Standard". It's unrealistic and the status would have no impact on
> whether it's implemented. Barbara

What's a difference for you?

Cheers, Jan