Re: Linux & draft-gont-6man-slaac-dns-config-issues

[Fernando Gont <[e-mail]>]

Hi, Jen,

On 03/24/2015 12:24 PM, Jen Linkova wrote:
> On Tue, Mar 24, 2015 at 5:22 PM, Fernando Gont <[e-mail]> wrote:
>>> I agree that "<= 2x MaxRtrAdvInterval" limitation is way too strict
>>> and we SHOULD fix this.
>>> However I do not like the idea of sanitizing the received values and
>>> limiting them to 10xMaxRtrAdvInterval.
>>
>> The thing here is that that's the only way (along with the change of the
>> semantics of the Lifetime parameter) in which you can fix the problem on
>> the client side.
> 
> Did I get it right that you are trying to solve one particular case
> when people are shooting themselves in the foot?  I do not believe
> that every single possible network misconfiguration case should be
> solved on the client side, especially at the price of making life
> harder for people who know what they are doing.

It is not a network misconfiguration. It's a bug in the spec. And until
the spec is updated and the routers are updated accordingly, the only
thing for a client to do is to sanitize the Lifetime value to a more
sane value. -- that's what Linux is doing.



>>> It really sounds like a
>>> micromanaging network administrators in how they run their networks.
>>> There might be valid reasons to have RDNSS/DNSSL Lifetime to be close
>>> enough to how often RAs are sent (what if a router is a DNS server
>>> itself? So if a router does down I don't want a client to use it).
>>
>> If that's de goal, then the Lifetime value should be set according to
>> the "Valid Lifetime" of the RA, rather than according to MaxRtrAdvInterval
> 
> You mean Router Lifetime, don't you?

Yes (sorry).


> And yes, it might be equal to Router Lifetime, which might be
> significantly lower than 6000 sec (1 hr 40 mins) you are proposing.

I think that the specific value that we use is the low-order bit (I can
certainly live with 6000 sec). The high order bit is that there is a
problem to be fixed.



>>> Actually, the wording of Sanitizing section is not accurate. Clients
>>> do not know MaxRtrAdvInterval value as it's something configured on a
>>> router. I assume it should be "the default MaxRtrAdvInterval value of
>>> 600 sec"?
>>
>> Yes.
> 
> It would be nice to update the text accordingly.

We certainly will.

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