Re: [v6ops] A common problem with SLAAC in "renumbering" scenarios

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

On 13/02/2019 18:43, STARK, BARBARA H wrote:
>>> Not necessarily. There is no requirement that the ISP leases the
>>> same prefix if the client just asks a new prefix.
>> 
>> Seriously? Same DUID should get the same PD inside the lifetime
>> cycle (whatever lifetime is set, 1h or 30 years).
>> 
>> No requirement? Well, then we need one, do we?
> 
> Maybe it would be helpful to place such a requirement on the network
> equipment implementations, rather than the ISPs? It's really hard for
> an operator to meet "requirements" when the equipment either doesn't
> make it possible, makes it difficult, or increases cost (e.g., if
> this capability is enabled, this equipment will only be able to
> handle half as many customers). I'm not saying that is or isn't the
> case for any particular implementations for prefix delegation. But in
> my experience, the capabilities of deployed equipment play a *huge*
> role in an operator's deployment decisions and configurations. 
> Barbara
> 

Dear Barbara,

Exactly right, you hit the nail on the head.

Operators would use whatever vendors implement and make available in 
configurations. If we make such a requirement a must in standard 
document, then I believe we place this requirement on vendors first, 
right? If vendors would build stuff that doesn't allow breaking the 
protocol, then life would be better.

However... I've seen many operators (also the ones that I worked for or 
with) placing feature requirements towards vendors and some of them 
were... well... strange, to be polite.

At this point if a request to do weird things with PDs comes from 
operator to vendor - a vendor should be in position to say "well, we 
like your request, but we can't do it because it's against the standard 
and would break the protocol".

Cheers, Jan