Re: [Roll] Error flows, which ICMP errors and to which root

["Li Zhao (liz3)" <[e-mail]>]

Hello Pascal,

One concern for section 4 “Extending RFC 6553”.
Are other fields “'O', 'R', 'F' flags, SenderRank ” necessary to be zero if ‘P’ flag is set?
For storing-PDAO, maybe we can leverage these fields for error-detection if need. So do we need replace “MUST” to “SHOULD” or “MAY” to keep the flexibility?


Best regards,
Li

From: Roll <[e-mail]>
Date: Monday, November 30, 2020 at 17:28
To: Routing Over Low power and Lossy networks <[e-mail]>
Subject: Re: [Roll] Error flows, which ICMP errors and to which root
Yes Huimin

And this is the normal behavior.

With PDAO we created the unusual behavior to send the error in p-route to the root. This was to make things faster since the communication through the Track is directional so you cannot talk back. Also in non storing talking to the source is via the Root and ultimately the Root needs to update the P DAO.

With RAW we may want to to something faster but unclear for now.

Please note that after the discussion with you and Li I published a revision that has the flag in the RPI that indicates that the packet is forwarded on a P-route.

Do you want to change the error flow?

Pascal

> Le 30 nov. 2020 à 09:15, Huimin She (hushe) <hushe=[e-mail]> a écrit :
>
> Hi Pascal,
>
> RFC 6550 section 11.1 says the ICMPv6 error in SRH should be sent to the source of the packet.
>
> Best regards,
> Huimin
>
>    Message: 2
>    Date: Fri, 27 Nov 2020 07:47:22 +0000
>    From: "Pascal Thubert (pthubert)" <[e-mail]>
>    To: Routing Over Low power and Lossy networks <[e-mail]>
>    Subject: [Roll] Error flows, which ICMP errors and to which root?:
>        [extends] IETF 109 open Questions on P-DAO
>    Message-ID:
>        <[e-mail]>
>
>    Content-Type: text/plain; charset="us-ascii"
>
>    Hello Li
>
>    This is the wrong thread. I created a new one.
>
>> Section 7.9 of pdao-draft defines a new code for  ICMPv6 error message "Error in Projected Route". Does it only work for ICMP errors sent to the main Root?
>
>    Section 5 says "
>
>
>       In case of a forwarding error along a Projected Route, an ICMP error
>
>       is sent to the Root with a new Code "Error in Projected Route" (See
>
>       Section 7.9<https://tools.ietf.org/html/draft-ietf-roll-dao-projection-14#section-7.9>).  The Root can then modify or remove the Projected
>
>       Route.  The "Error in Projected Route" message has the same format as
>
>       the "Destination Unreachable Message", as specified in RFC 4443<https://tools.ietf.org/html/rfc4443>
>
>       [RFC4443<https://tools.ietf.org/html/rfc4443>].
>
>    "
>
>    So yes the intention was to send the ICMP to the main Root. But as you point out the packet does not indicate it is following a P-route. This was related to storing mode P DAO. In non-storing the node does not know it's a P-route.
>
>
>> In non-storing PDAO, forwarder can't recognize whether data packet is in PDAO instance. Forwarder should send ICMP Destination Unreachable error to root (the source of the packet), then root generates ICMPv6 error message with "Error in Projected Route" to main Root.  Is it correct?
>
>    That would work. Seems neat. The alternate would be to signal it is a P route in the RPI. That's item 2) in the list in this thread. If we do that the current text works. What makes more sense to you?
>
>> In storing PDAO, forwarder can recognize the PDAO instance from the RPI. It can send "Error in Projected Route" or  "Destination Unreachable error" to root. Maybe we need more claims for which code forwarder should use.
>
>    We have to decide if we send it to the main root as written in the current draft or to the Track Root. If the P route is reversible could be done that way. But that's added complexity. I'm not very convinced either way. The Okham razor could be to do the simplest.
>
>    Keep safe!
>
>    Pascal
>
>
>
>
>
> _______________________________________________
> Roll mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/roll
_______________________________________________
Roll mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/roll