Re: [dhcwg] Re: PD lifetimes

[JINMEI Tatuya / 神明達哉 <[e-mail]>]

>>>>> On Thu, 23 Jan 2003 20:15:10 -0500, 
>>>>> Ralph Droms <[e-mail]> said:

> 1. In response to your first comment, suppose we change the paragraph in 
> question to:

>       In a message sent by a delegating router the preferred and valid
>       lifetimes for each prefix should be set to the values
>       AdvValidLifetime and AdvPreferredLifetime for that prefix, as
>       specified in section "Router Configuration Variables" of RFC2461
>       [3], unless administratively configured.

> If I understand your second comment correctly, following the PD spec will 
> result in the valid lifetimes in the RAs on the downstream links from the 
> RR grow smaller over time, so as not to exceed the valid lifetime for the 
> prefix in the IA_PD, until the RR sends a Renew to extend the valid 
> lifetime on the prefix.  Did I get that right?

Yes.

> Is this situation a problem 
> for the hosts on the downstream links and should we try to fix the problem?

Honestly, this won't be a "problem" in practice, since
AdvValidLifetime and AdvPreferredLifetime are large enough.  In
theory, however, the current default lifetimes specified in the PD
draft changes the condition about stability of prefix lifetimes for
downstream hosts.  

According to the default values of RFC 2461, an autoconfigured address
for a host is valid unless the host misses router advertisements for
AdvValidLifetime (I ignore the advertisement interval for simplicity).
According to the current PD spec, however, a host can receive a prefix
with the valid lifetime being L = (AdvValidLifetime - .5 *
AdvPreferredLifetime).  Thus, if the host misses succeeding router
advertisements for L seconds, the corresponding address will be
invalidated, even if the renew/reply exchange between RR and DR
succeeded (and continued to succeeding).

Perhaps I'm a stickler, but I'd like to ensure the same stability from
the downstream hosts' point of view, as long as the renew/reply
exchange succeeds at the PD side.

> 2. Proposed clarification for the sentence in question:

>     The requesting router MUST select the value of AdvValidLifetime
>     assigned to any prefixes subnetted from a delegated prefix to be
>     less than the valid lifetime in the prefix option for that
>     delegated prefix.  The requesting router MAY use the preferred
>     lifetime in the prefix option for a delegated prefix as the value
>     of AdvPreferredLifetime for a prefix subnetted from that delegated
>     prefix, if the requesting router chooses to allow the delegating
>     router to control that value.  Otherwise, the requesting router
>     chooses a value of AdvPreferredLifetime for the subnetted prefix
>     that is less than the value of AdvValidLifetime for that prefix.

(I corrected a typo in the citation above)

Then are you suggesting to keep the constant value of AdvValidLifetime
after receiving a Reply (to Request) and before trying a Renew/Reply
exchange?  If so, I don't think this is a good idea.  We must ensure
the invariant that the "site" valid lifetime given by PD must be equal
to or larger than the valid lifetime (in router advertisements) for
the prefix of each downstream link.

I'll switch to Erik's message for further discussion.

					JINMEI, Tatuya
					Communication Platform Lab.
					Corporate R&D Center, Toshiba Corp.
					[e-mail]
_______________________________________________
dhcwg mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/dhcwg