Re: [Roll] RPLInstanceID parameter from Source Routing Header is missing in RFC6554

It doesn't have to be all.  As mentioned in my previous mail, you can target the specific RPL Instance by inspecting the original packet header in the ICMP error payload.

--
Jonathan Hui

On Jun 29, 2012, at 9:33 AM, "Dario Tedeschi" <dat@exegin.com> wrote:

> Ok, I think I've got it. That's similar to what Mukul suggested: "all the instances that have Y in the source route to D"
>  
> However, I still think it's a bit unfortunate from an implementation perspective.
> 
> Thank you for the explanation.
> 
> Dario
> 
> On 12-06-29 08:56 AM, Jonathan Hui wrote:
>> 
>> 
>> Hi Dario,
>> 
>> The ICMPv6 Destination Unreachable error will come from Y and indicate Y -> D is broken.  Any RPL Path that utilizes Y -> D (again, regardless of the RPL Instance) will require repair.  The DAG Root can choose to issue a DTSN update for any RPL Instance that utilizes link Y -> D in a downward path.  It can even choose the RPL Instance that it uses to route the original packet.  Because the ICMPv6 Destination Unreachable message encapsulates the error-generating packet (both the SRH and original packet), it has all the information it needs.
>> 
>> --
>> Jonathan Hui
>> 
>> On Jun 29, 2012, at 8:48 AM, Dario Tedeschi wrote:
>> 
>>> Hi Jonathan
>>> 
>>> I would agree if link A -> B existed in more than one instance. In my example however, the source route for the first RPL instance has link Y -> D while the second instance has link Z -> D. How does R inform D to resend its DAO for only the first instance if only Y -> D is broken?
>>> 
>>> Dario
>>> 
>>> 
>>> On 12-06-21 10:54 PM, Jonathan Hui wrote:
>>>> 
>>>> 
>>>> Hi Dario,
>>>> 
>>>> The ICMP Error message will contain the IPv6 header of the error-generating packet.  If link A -> B is broken, the IPv6 Destination Address will indicate B and the SRH will indicate A.
>>>> 
>>>> My point is that if you want to clean up routes, you need to clean up all routes that use A -> B, not just the particular Instance that you used to generate the route and discover the error.
>>>> 
>>>> --
>>>> Jonathan Hui
>>>> 
>>>> On Jun 21, 2012, at 9:05 AM, Dario Tedeschi wrote:
>>>> 
>>>>> Hi Jonathan
>>>>> 
>>>>> Sorry, for late response to your email. 
>>>>> 
>>>>> Consider the scenario where a DODAG root node supports two or more RPL instances under one prefix. Each instance providing different routing objectives (i.e. different objective functions). Potentially, the source-route to a destination under one instance can be different to the source-route to the same destination under a different instance. For example: Instance 1 could have a source-route R-->X-->Y-->D, while instance 2 could have a source-route R-->X-->Z-->D. Lets say node Y sent a Destination Unreachable ("Error in source-route") back to R, because D was unreachable from Y. How would R know which source-route needed repair and, consequently, which instance needed a DTSN update?
>>>>> 
>>>>> Regards
>>>>> Dario
>>>>> 
>>>>> On 01/06/2012 7:46 PM, Jonathan Hui wrote:
>>>>>> 
>>>>>> 
>>>>>> I don't understand the need for knowing the RPLInstanceID when handling an ICMP Destination Unreachable error.  The SRH does not follow any RPL Instance, just the IPv6 addresses listed in the SRH.  Receiving an ICMP Destination                         Unreachable error from a node S simply means that S could not forward to the next IPv6 address in the SRH.  The inability of S to forward to the specified next hop would be true regardless of what RPL Instance the SRH was constructed from.
>>>>>> 
>>>>>> --
>>>>>> Jonathan Hui
>>>>>> 
>>>>>> On Jun 1, 2012, at 7:21 PM, Dario Tedeschi wrote:
>>>>>> 
>>>>>>> Yes, I noticed this as well. It is unfortunate because as you say, there is now no way to identify which instance a Error in SRH is for, and subsequently the root node can't just remove that routing entry in a specific instance. Instead it must either remove that entry from all instances or do nothing and wait for the next DAO to update the entry.
>>>>>>> 
>>>>>>> Dario
>>>>>>> 
>>>>>>> On 30/05/2012 1:58 AM, Tecuceanu Andreea-Dana-B10623 wrote:
>>>>>>>> 
>>>>>>>> Hello,
>>>>>>>>  
>>>>>>>> I have observed that the RPLInstanceID field from Source Routing header (present in  draft-ietf-6man-rpl-routing-header-07) was removed in RFC6554.
>>>>>>>> Are there particular reasons for this? This field is necessary at the DODAG Root when it receives a  Destination Unreachable with code "Error in Source Routing Header" to identify the instance with the problem (only if there are two instances with the same prefix and if the node is Root in both of them).
>>>>>>>>  
>>>>>>>>  
>>>>>>>> Andreea Tecuceanu
>>>>>>>> (Freescale Semiconductor)
>>>>>>>> 
>>>>>>>> 
>>>>>>>> _______________________________________________
>>>>>>>> Roll mailing list
>>>>>>>> Roll@ietf.org
>>>>>>>> https://www.ietf.org/mailman/listinfo/roll
>>>>>>> 
>>>>>>> _______________________________________________
>>>>>>> Roll mailing list
>>>>>>> Roll@ietf.org
>>>>>>> https://www.ietf.org/mailman/listinfo/roll
>>>>>> 
>>>>> 
>>>> 
>>> 
>> 
> 
> 