Re: [DNA] Issue 2: What if no prefix is advertised?

[Sathya Narayanan <sathya@research.panasonic.com>]

Hello Greg -

I agree with what you have said.

But, the open issue is different: the case we are talking about with 
this issue is when stateful address configuration is used on a 
particular link. We agreed during the last meeting to mandate the router 
on such a link to advertise atleast one prefix (the smallest prefix) 
with A=0 and L=0, so that the hosts can still use the RS/RA exchange to 
find out about network attachement. At the moment, there is text in the 
draft for using DHCP CONFIRM - we thought we could remove this text 
after mandating that routers advertise atleast one prefix.

My questions is: how will the routers know about a prefix if stateful 
address configuration is used - do we understand/agree that the routers 
MUST be configured with this information (atleast one prefix) even when 
stateful address configuration is used. Is that acceptable?

Sathya

>Hi Sathya,
>
>We have an issue in that routers may already be available which do
>not advertise a prefix (as it is not precluded).
>
>I would suggest that these routers should try to advertise a prefix, 
>be it a learned or native (PIO) prefix.
>
>That way routers can participate if they learn of other prefixes.
>Otherwise, they will still advertise if there are no prefixes known anyway.
>This will provide hosts with the A= and M= flags sufficient to initiate DHCP
>(which is not a DNA process but can provide the change detection required).
>
>Greg
>
>----- Original Message -----
>From: Sathya Narayanan <sathya@research.panasonic.com>
>Date: Friday, January 19, 2007 8:10 am
>Subject: Re: [DNA] Issue 2: What if no prefix is advertised?
>To: Dna <dna@eng.monash.edu.au>
>
>  
>
>>If we are going to mandate that the router advertise one prefix even
>>when stateful autoconfiguration is used and that it must be the 
>>smallestprefix, we are implicitly requiring the adminstrator to 
>>configure each
>>of these routers with this information, right? Or, are we talking 
>>aboutthe routers sniffing DHCP messages - What does it mean?
>>
>>I am leaving this open for now, until I hear comments on this 
>>question.
>>- Sathya
>>
>>JinHyeock Choi wrote:
>>
>>    
>>
>>>>Issue: If stateful autoconfiguration is used, the routers may not
>>>>advertise any prefix in its RA message. DNA will not work 
>>>>        
>>>>
>>without prefix
>>    
>>
>>>>advertised. Current text relies on DHCP confirm message exchange 
>>>>        
>>>>
>>for DNA
>>    
>>
>>>>in that case.
>>>>
>>>>Agreement at the meeting: Lets mandate the inclusion of one 
>>>>        
>>>>
>>prefix in
>>    
>>
>>>>all RA message, with L= 0 and A=0. But, atleast one prefix is a 
>>>>        
>>>>
>>MUST.>> The router MUST include the smallest prefix either in PIO 
>>or LPO of all
>>    
>>
>>>>RA messages.
>>>>Note: I am not refering to the smallest prefix as "LinkID" 
>>>>        
>>>>
>>because in
>>    
>>
>>>>"Issue 6", I will be proposing the removal of the concept of 
>>>>        
>>>>
>>LinkIDs.>> The forward compatibility advantage of the LinkIDs can 
>>be easily
>>    
>>
>>>>supported and refer to the email on Issue 6 is further details.
>>>>
>>>>If you have any objections to the proposed change please post your
>>>>objections on the list.
>>>>        
>>>>
>>>I agree and, IMO, the concept of 'LinkID' is of help. It would be
>>>simpler to mandate to include LinkID prefix.
>>>
>>>Best Regards
>>>
>>>JinHyeock
>>>      
>>>
>>
>>    
>>