Re: [CCAMP] [Technical Errata Reported] RFC4872 (3304)

["Ong, Lyndon" <[e-mail]>]

Hi Adrian,

Thank you for the fast evaluation of the errata.  It sounds like the correction that I suggested has ended up overspecifying the method to do reversion with full rerouting when it is very possible to support a form of reversion that doesn't involve maintaining the old LSP.

>From your response I believe that you do agree that it was not the intent of the original specification text to imply that reversion with full rerouting is not allowed (or to require that the old LSP always be torn down in full rerouting), so hopefully with some more discussion we can determine if there is anything that could be done to make that clearer.

Best regards,

Lyndon

-----Original Message-----
From: Adrian Farrel [mailto:[e-mail]] 
Sent: Tuesday, July 31, 2012 6:28 PM
To: [e-mail]
Cc: [e-mail]; [e-mail]; [e-mail]; [e-mail]; [e-mail]; [e-mail]; Ong, Lyndon
Subject: RE: [Technical Errata Reported] RFC4872 (3304)

Hi CCAMP,

I find that this erratum is raised against two sections one of which I supplied text for. If this get contentious, I will call on Stewart to call consensus and handle the Erratum in the system.

In my opinion, this proposal goes further than the intention of the authors/WG in publishing 4872.

With regard to the proposed addition to section 11...
The use of mb4b is already in scope. The existing text says "The new LSP resources can be established using the make-before-break mechanism," so there is no need to re-state "The new LSP can be established without tearing down the old LSP".

I think your concern here is whether the old LSP is ever torn down. I think that you are worried that if the old LSP is torn down, it might be impossible to perform reversion because, after repair, an attempt to revert (also using mb4b) might find that key resources have been "stolen" by some other LSP. I don't see this as at all different from the issue of the protection LSP itself. That is, it is of the nature of LSP Rerouting as a protection mechanism that:
a. protection may fail because of lack of resources b. reversion may fail because of lack of resources

*If* reversion is so important, I don't quite see why protection is not important. If protection is important then you should be using a proper protection mechanism and not waiting for post facto rerouting. Furthermore, if you require that the LSP be retained for restoration, why are you not using a protection mechanism? 

But the general paradigm here is that you are willing to use the best available LSP when it is set up in the first place, the best available LSP when you re-route after failure, and the best available LSP when you "revert".

Lastly, it *does* remain an _option_ to retain the failed LSP in order to switch back. Nothing in the old text precludes that, although I understand that there is an implication that it might be expected to be torn down.

So I conclude that the proposed addition to section 12 is not what the authors/WG intended.

We should discuss further.

Adrian

> -----Original Message-----
> From: RFC Errata System [mailto:[e-mail]]
> Sent: 31 July 2012 17:39
> To: [e-mail]; [e-mail]; dimitri.papadimitriou@alcatel- 
> lucent.be; [e-mail]; [e-mail]; [e-mail]; 
> [e-mail]
> Cc: [e-mail]; [e-mail]; [e-mail]
> Subject: [Technical Errata Reported] RFC4872 (3304)
> 
> 
> The following errata report has been submitted for RFC4872, "RSVP-TE 
> Extensions in Support of End-to-End Generalized Multi-Protocol Label 
> Switching (GMPLS) Recovery".
> 
> --------------------------------------
> You may review the report below and at:
> http://www.rfc-editor.org/errata_search.php?rfc=4872&eid=3304
> 
> --------------------------------------
> Type: Technical
> Reported by: Lyndon Ong <[e-mail]>
> 
> Section: 11 & 12
> 
> Original Text
> -------------
> Section 11 says:
> 
> 
>    (Full) LSP rerouting will be initiated by the head-end node that has
>    either detected the LSP failure or received a Notify message and/or a
>    PathErr message with the new error code/sub-code "Notify Error/LSP
>    Locally Failed" for this LSP.  The new LSP resources can be
>    established using the make-before-break mechanism, where the new LSP
>    is set up before the old LSP is torn down.  This is done by using the
>    mechanisms of the SESSION_ATTRIBUTE object and the Shared-Explicit
>    (SE) reservation style (see [RFC3209]).  Both the new and old LSPs
>    can share resources at common nodes.
> 
> Section 12 says:
> 
>    [No text on reversion for (full) LSP Rerouting.]
> 
> Corrected Text
> --------------
> Section 11 should say:
> 
> 
>    (Full) LSP rerouting will be initiated by the head-end node that has
>    either detected the LSP failure or received a Notify message and/or a
>    PathErr message with the new error code/sub-code "Notify Error/LSP
>    Locally Failed" for this LSP.  The new LSP resources can be
>    established using the make-before-break mechanism, where the new LSP
>    is set up before the old LSP is torn down.  This is done by using the
>    mechanisms of the SESSION_ATTRIBUTE object and the Shared-Explicit
>    (SE) reservation style (see [RFC3209]).  Both the new and old LSPs
>    can share resources at common nodes.  The new LSP can be established
>    without tearing down the old LSP in case of reversion (see section 12).
> 
> Section 12 should say:
> 
>    For "(full) LSP Rerouting", reversion implies that the old LSP is not
>    torn down by the head-end node after the new LSP is established. For
>    reversion, the head-end node re-activates the old LSP after this has
>    recovered.
> 
> 
> 
> Notes
> -----
> Current text in RFC 4872 describes reversion in the cases of 1+1 
> bidirectional Protection, 1:N Protection with Extra Traffic and 
> Rerouting Without Extra
Traffic,
> however it has no description of reversion with (Full) LSP Rerouting.
> For (full) LSP Rerouting, the description in Section 11 instead 
> implies that
the old
> LSP is torn down. This has led to some confusion as to whether 
> reversion with
> (full) LSP Rerouting is allowed or not allowed by the RFC. We believe 
> this was
not
> intentional. The additions would make it clear that reversion can be 
> supported with (Full) LSP Rerouting.
> 
> Instructions:
> -------------
> This errata is currently posted as "Reported". If necessary, please 
> use "Reply All" to discuss whether it should be verified or rejected. 
> When a decision is reached, the verifying party (IESG) can log in to 
> change the status and edit the report, if necessary.
> 
> --------------------------------------
> RFC4872 (draft-ietf-ccamp-gmpls-recovery-e2e-signaling-04)
> --------------------------------------
> Title               : RSVP-TE Extensions in Support of End-to-End Generalized
Multi-
> Protocol Label Switching (GMPLS) Recovery
> Publication Date    : May 2007
> Author(s)           : J.P. Lang, Ed., Y. Rekhter, Ed., D. Papadimitriou, Ed.
> Category            : PROPOSED STANDARD
> Source              : Common Control and Measurement Plane
> Area                : Routing
> Stream              : IETF
> Verifying Party     : IESG