On 08/08/2018 12:02, Toerless Eckert wrote:
> Brian:
> 
> As explained in reply to Alissa, this was leftover text, so
> i could simply remove it, except that i figured that it was wrong
> to allow these addresses into the Certs, which lead to the mayor fixes
> for Alissas discuss.
> 
> The ACP connect LANs are just normal IPv6 subnets, managed via
> existing manual methods, SLAAC etc. This is written already for
> several revisions of the doc now into 8.1 and specifically 8.1.3.
> 
> I did not fancy to bother about recommending any specific SLAAC
> enhancements in that section, because i wanted to concentrate
> on the important methods only, and thats primarily the route
> autoconfiguration (RFC4191) because those NOC nodes will be multi-homed,
> one interface on ACP connect subnet, one on Data-Plane, and then
> you want to make sure they correctly route the ACP destinations
> across the ACP interface only.
> 
> I looked through RFC8064 and didn't find it particularily beneficial
> for the use-case. It seems to be very much focussed on user-devices
> with a browser connected to the Internet and the resulting security
> and privacy issues. Not really relevant IMHO for Network Management
> servers in a NOC that likely announce their addresses via GRASP
> into the ACP anyhow, so those addresses do not need to be stable,
> and they certainly can't be private/hidden, and i wouldn't see
> how any of the rfc8064/rfc7217 considerations would increase security
> for these servers. 

True. My thinking was more that since the current general-purpose
recommendation is RFC8064, ACP implementers don't need to do
anything different (e.g. continue using old-fashioned Modified EUI-64).
But saying nothing is fine.

> 
> Besides, the rfc7217 algorithm recommended by rfc8064 is IMHO
> not very good to achieve stable addresses. A common issue
> i have with those servers is that ethernet HW fails, and then
> you have to reconnect your subnet to another physcial port
> (often servers have more than enough spare ports) and i bet that
> the Net_Iface parameter will cause the stable address to change.

Yes, that has been noticed. But requiring a node to re-identify
itself after a MAC address change seems like a good thing anyway.

   Brian

> 
> This whole disussion was important though, because it had me finally
> realize that i can not put normally managed addresses, even if
> they claim to be stable into certificates, because they never
> are stable, and a registrar can't determine them upfront
> when you want to leverage existing IPv6 protocols to manage
> them.
> 
> Cheers
>     Toerless
> 
> On Thu, Aug 02, 2018 at 08:44:53AM +1200, Brian E Carpenter wrote:
>> On 02/08/2018 03:52, Alissa Cooper wrote:
>> ....
>>> ----------------------------------------------------------------------
>>> DISCUSS:
>>> ----------------------------------------------------------------------
>>>
>>> Section 6.10.4:
>>>
>>> "The value of the Interface Identifier is left for future definitions."
>>>
>>> I don't understand how this is usable without some definition. I.e., if I
>>> assign every interface ID in my ACP to be 0, it's not going to work. Could 
>>> you
>>> elaborate about what is expected in this field?
>>>
>>
>> Why not:
>>
>> The value of the Interface Identifier SHOULD follow [RFC8064].
>>
>> That would work fine and is a SHOULD to allow for alternatives
>> if there's a good reason.
>>
>>    Brian
>>
>> _______________________________________________
>> Anima mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/anima
> 

_______________________________________________
Anima mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/anima

Reply via email to