Re: [netlmm] Review of LMA Discovery document

Behcet:

Right, this document does not describe a DNS-based LMA discovery method.  But it analyzes some potential issues with such a method, and my comments relate to this analysis.  I felt that it be noteworthy that some of the potential issues can be avoided by proper implementation and administration.

Does this make sense?

- Christian

On Apr 30, 2010, at 21:37, Behcet Sarikaya wrote:

> Hi Christian,
> 
>   The way DNS is used in the draft is how an ordinary host makes a DNS query.
> 
>   This is no DNS based LMA discovery. I think you (and Julien) had already indicated this in your mails to the list earlier. RFC 5026-like DNS discovery has been described in our draft as we had indicated many times in the list.  In fact DNS based LMA discovery in PMIPv6 is much more secure and efficient that MIPv6 DNS discovery of HAs.
> 
> My suggestion is to drop totally Section 4 as this section is artificial without a real DNS discovery solution and instead merge draft-sarikaya-netlmm-lma-dnsdiscovery-01
> 
> 
> with this draft.
> 
> Regards,
> 
> Behcet
> ----- Original Message ----
>> From: Christian Vogt <christian.vogt@ericsson.com>
>> To: NETLMM Mailing List <netlmm@ietf.org>
>> Cc: jouni korhonen <jounikor@gmail.com>
>> Sent: Thu, April 29, 2010 4:53:40 PM
>> Subject: [netlmm] Review of LMA Discovery document
>> 
>> Folks,
> 
> Jouni asked me to review the LMA Discovery document.  I did 
>> this, and
> found that the document is mature and well written, and should 
>> be
> progressed further as soon as possible.  I have just one comment 
>> on
> section 4:
> 
> Section 4 gives two classes of reasons not to use the 
>> DNS for LMA
> discovery.  The first class is related to update propagation 
>> latencies
> caused by caching.  The second class is related to update 
>> latencies
> caused by primary-to-secondary server synchronization.  My 
>> comment
> relates to the first class of reasons:
> 
> Can't the issues 
>> described be well avoided?  In deployments where the
> DNS is to be used 
>> for LMA discovery, make sure you have DNS
> implementations that accept low 
>> TTLs, and set your TTLs low -- that
> seems to be it.  I agree that the 
>> issues you are describing do need to
> be considered, and hence it is very 
>> important to document them.  But I
> personally don't see them as 
>> show-stoppers.
> 
> This said, your second class of reasons may be more 
>> intractable.
> 
> Hope this helps.  Best regards,
> 
> - 
>> Christian
> 
> _______________________________________________
> netlmm 
>> mailing list
>> href="mailto:netlmm@ietf.org">netlmm@ietf.org
>> href="https://www.ietf.org/mailman/listinfo/netlmm" target=_blank 
>>> https://www.ietf.org/mailman/listinfo/netlmm
> 
> 
> 