The problem is that one problem is FAR more likely to happen than the other.

I shutdown my machine every night and power it on again in the morning when I 
come to work.
Therefore, every night of every workday I experience the type of outage 
described in our draft.
Furthermore, I occasionally go on vacations too - so the outage may last more 
than a day.

What this means for an administrator is that he has to predict, in advance, how 
long I may 
be on vacation so that the RA deprecating the old prefix can last long enough.  
That puts
an unreasonable expectation on the network administrator.  Furthermore, I don't 
want to
have to get permission from my network administrator in order to go on vacation.

On the other hand, switch/router outages are typically noticed right away by 
many angry engineers
and are usually fixed in a matter of hours by the same administrators who 
control the RA.

This is not a minor implementation detail.  Fortunately, the DNA people also 
realized that this
was a potential problem and they have a solution specified - first deprecate 
the cached information
and verify before using it.  This solution will fix the blind cache reuse 
problem.  We propose this 
same solution in our new text (last paragraph of section 2) that Hemant just 
sent out.

- Wes

-----Original Message-----
From: JINMEI Tatuya / 神明達哉 [mailto:[EMAIL PROTECTED] 
Sent: Thursday, July 10, 2008 3:54 PM
To: Hemant Singh (shemant)
Cc: Suresh Krishnan; Wes Beebee (wbeebee); Thomas Narten; Brian Haberman; Bob 
Hinden; [email protected]
Subject: Re: 6MAN WG Last Call:draft-ietf-6man-ipv6-subnet-model-00.txt

At Thu, 10 Jul 2008 12:09:01 -0400,
"Hemant Singh (shemant)" <[EMAIL PROTECTED]> wrote:

> Since you don't want any new rules added by our draft, we changed 
> bullet
> 3 related to caching on-link determination. The new bullet text does 
> not add any normative requirements but clearly says why it is a bad 
> idea to cache on-link determination.  Also, our draft is about on-link 
> determination - we are not adding anything related to IPv6 address 
> caching - we have said repeatedly, save it for another day.

The new text does not make sense to me.

In this scenario, the same problem can occur when a host (that just keeps 
working, without a reboot) happens to fail to receive the RAs containing 
0-lifetime prefixes (such a failure can happen for various reasons: there may 
be a temporary failure in an intermediate switch; the host may have been just 
too busy and cannot handle the RAs, etc).  So, what's wrong in this scenario is 
that the router doesn't keep advertising 0-lifetime-prefixes sufficiently long. 
 This scenario itself doesn't explain 'why caching on-link prefix is a bad 
idea'.

This story also explains why I previously said "such caching is a minor 
implementation detail".  In terms of external behavior, a node that caches 
configured address/on-link prefix and reuses it after a reboot is often 
indistinguishable from a node that happens to fail receiving some updates from 
RA for some period.  Killing the former (while forgetting the latter) can just 
miss the more fundamental problem.

Aside from this essential point, the new bullet does also not make sense in the 
context of 'A correctly implemented IPv6 host MUST adhere to the following 
rules'.  If we find any valid 'justification' like this, it should be described 
outside this listing, somewhere more appropriate in the entire context.

---
JINMEI, Tatuya
Internet Systems Consortium, Inc.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to