Speaking as WG Member:

Hi Chris, 


> On Sep 27, 2025, at 6:42 AM, Christian Hopps <[email protected]> wrote:
> 
> Acee Lindem <[email protected]> writes:
>>> On Sep 18, 2025, at 11:05 AM, Peter Psenak <[email protected]> wrote:
>>> 
>>> Hi Acee,
>>> 
>>> On 18/09/2025 16:32, Acee Lindem wrote:
> 
>>>>>> The AC-flag simply indicates that the prefix has been configured as 
>>>>>> anycast - i.e. originated by multiple routers.
> 
>>>> [...] where and how this is configured for a prefix since
>>>> we've had the IS-IS Anycast prefix flag for some time. [...]
>>> 
>>> Example from one implementation:
>>> 
>>> router isis 1
>>> interface Loopback0
>>>  prefix-attributes anycast
>> 
>> Ok - than this would support keeping the config in the OSPF YANG model as 
>> Yingzhen has proposed. Authors, let’s keep it there.
> 
> Its not a great solution IMO. I think I agree with the direction of your 
> earlier questions, (e.g., the property can also be advertised in IS-IS). 
> Anycast prefix is an IGP agnostic concept -- an anycast prefix is a prefix is 
> associated with multiple nodes in the network.
> 
> This also isn't an interface wide property it's a prefix property. Someone 
> earlier said, "most of the time the user doesn't configure multiple prefixes 
> on their loopback", I'm not sure that's true, especially if we consider 
> supporting (multiple) anycast application uses along with non-anycast 
> applications.
> 
> Anyway, to have the only configuration option only cover "most of the time" 
> cases doesn't seem great.
> 
> A more precise config is to augment the interface prefix configuration 
> (ietf-ip) as suggested by Acee earlier (I think).

I'm afraid we are letting "Prefect be the enemy of good" here. 

I think if RFC 8344 is to augmented, it needs to be done in a separate document 
as adding flags to the IP and IPv6  addresses will undoubtedly bring in a 
larger group of opinions into play and stall this document indefinitely. 

I understand the reasons for putting it in the IGP model (as Cisco did with 
their IS-IS configuration). 

  1. The fact that the prefix is any cast is needed solely by the IGPs. 
  2. Given #1, it is expedient to add it to the IGP configuration. 

So, I think the question should be whether or not to leave it as OSPF interface 
config or do nothing in this draft. I don't feel that strongly on this but am 
leaning towards leaving it since Cisco already did it that way in their IS-IS 
configuration. 


Thanks,
Acee



> 
> When ietf-ip was suggested, Yingzhen (I believe), asked how to configure the 
> protocol to distribute the property then. Why not just distribute it if it 
> has been set. What is more likely, that the same loopback is configured under 
> IS-IS and OSPF but you only want to advertise it's anycast property in one of 
> those IGPs *or* you have anycast and non-anycast prefixes on your loopback? I 
> think multple prefixes on your loopback is much more likely between those 
> two. And if you've used multiple loopbacks to support multiple prefixes, well 
> that solution still works with the ietf-ip augmentation as well.
> 
> Thanks,
> Chris.

_______________________________________________
Lsr mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to