Re: [mpls] [Teas] FW: New Version Notification for draft-ietf-mpls-residence-time-05.txt

[Lou Berger <[e-mail]>]

Greg,

On 3/25/2016 12:52 PM, Gregory Mirsky wrote:
> Hi Lou,
> Greatly appreciate you've found time during these extremely busy days before the IETF meeting.
>
> I agree, that there could be different ways a node can handle the RRO limitation. I addition to those you've suggested node may change RTM Capability state advertised in IGP-TE. I think that the choice can be left to a developer.

I think the RSVP change should be in the draft as it impacts
interoperability and expected behavior.

Once this change is made I think the draft will be ready to progress.

Lou

> I've published the -06 version with all updates noted in my mail before the cut-off and if you agree with the changes perhaps Loa can start the WGLC at the most suitable time.
>
> 	Regards,
> 		Greg
>
>
> -----Original Message-----
> From: Lou Berger [mailto:[e-mail]] 
> Sent: Friday, March 25, 2016 5:51 AM
> To: Gregory Mirsky; Loa Andersson; [e-mail]
> Cc: [e-mail]; TEAS WG
> Subject: Re: [Teas] FW: New Version Notification for draft-ietf-mpls-residence-time-05.txt
>
> Greg,
>     Sorry for not responding before the ID cutoff. 
>
> I think this is fine, but why not go a step further and have the nodes that have the limitation indicate that RTM can't be used, e.g., clear the RTM_SET  Attribute Flag and/or remove the RTM_SET TLV to notify the ingress of the issue.  This would allow the automatic detection of the
> (corner) cases were there are actual issues and let RTM operate properly in the common case.
>
> Lou
>
> On 3/16/2016 3:11 PM, Gregory Mirsky wrote:
>> Hi Lou,
>> following the discussion we propose the following new paragraph before Section 4.7.1:
>>    There are scenarios when some information is removed from an RRO due
>>    to policy processing (e.g., as may happen between providers) or RRO
>>    is limited due to size constraints .  Such changes affect the core
>>    assumption of the method to control processing of RTM packets.  RTM
>>    SHOULD NOT be used if it is not guaranteed that RRO contains complete
>>    information.
>>
>> Hope that would address your concern with potential blackholed RTM packets.
>>
>> 	Regards,
>> 		Greg
>>
>> -----Original Message-----
>> From: Lou Berger [mailto:[e-mail]]
>> Sent: Wednesday, March 16, 2016 7:19 AM
>> To: Gregory Mirsky; Loa Andersson; [e-mail]
>> Cc: [e-mail]; TEAS WG
>> Subject: Re: [Teas] FW: New Version Notification for 
>> draft-ietf-mpls-residence-time-05.txt
>>
>> Greg,
>>     What happens when some information is removed from an RRO due to policy processing (e.g., as may happen between providers).  Couldn't the TTL calculation some up too small if the wrong SOs are removed, and wouldn't this result in a black hole of actual RTM messages?  How do you cover this case?
>>
>> Thanks,
>> Lou
>>
>> On 3/16/2016 10:15 AM, Gregory Mirsky wrote:
>>> Hi Lou,
>>> thank you for the detailed explanation of the case. I agree that there could be other ways to handle it from what been described in the document. If a node ID of the first node in the RTM_SET TLV is not found in the RRO list, then the TTL will be set to 255 and packet will not be processed by intermediate RTM-capable nodes but arrive at LSP egress LSR. Thus not all RTM capable nodes will be accounted for in the Scratch Pad but still some nodes may. I'll add some text to highlight this case.
>>>
>>> 	Regards,
>>> 		Greg
>>>
>>> -----Original Message-----
>>> From: Lou Berger [mailto:[e-mail]]
>>> Sent: Wednesday, March 16, 2016 5:30 AM
>>> To: Gregory Mirsky; Loa Andersson; [e-mail]
>>> Cc: [e-mail]; TEAS WG
>>> Subject: Re: [Teas] FW: New Version Notification for 
>>> draft-ietf-mpls-residence-time-05.txt
>>>
>>> Greg,
>>>
>>> Thanks for the response.  See below.
>>>
>>> On March 15, 2016 6:56:34 PM Gregory Mirsky <[e-mail]> wrote:
>>>
>>>> Hi Lou,
>>>>
>>>> thank you for the most detailed comments. Please find my notes 
>>>> in-line tagged GIM>>.
>>>>
>>>>  
>>>>
>>>>                 Regards,
>>>>
>>>>                                 Greg
>>>>
>>>>  
>>>>
>>>> -----Original Message-----
>>>> From: Lou Berger [mailto:[e-mail]]
>>>> Sent: Tuesday, March 15, 2016 3:41 PM
>>>> To: Gregory Mirsky; Loa Andersson; [e-mail]
>>>> Cc: [e-mail]; TEAS WG
>>>> Subject: Re: [Teas] FW: New Version Notification for 
>>>> draft-ietf-mpls-residence-time-05.txt
>>>>
>>>>  
>>>>
>>>>  
>>>>
>>>> Greg,
>>>>
>>>>  
>>>>
>>>> On 3/15/2016 5:13 PM, Gregory Mirsky wrote:
>>>>
>>>>> Hi,
>>>>> this version primarily addresses comments related to the proposed
>>>> extensions to RSVP-TE and operation of the control plane in support 
>>>> of Residence Time Measurement. Would appreciate the confirmation from Lou.
>>>>
>>>>  
>>>>
>>>> I don't think I saw a response to:
>>>>
>>>>> I think this doesn't always work, e.g., when RRO is modified due to 
>>>>> policy or limited due to size constraints.  This has to be addressed.
>>>>> I can see how to easily detect this situation, but not how to get 
>>>>> the number of non-supporting hops.  I guess it's always possible to 
>>>>> play ttl comparison games, but that's pretty ugly too.  Any 
>>>>> thoughts?
>>>> GIM>> The following text in Section 4.7 intended to handle the
>>>> situation you've described:
>>>>
>>>>    If the node cannot find matching ID in RRO,
>>>>
>>>>    then it MUST try to use ID of the next node in the RTM_SET until 
>>>> it
>>>>
>>>>    finds the match or reaches the end of RTM_SET TLV.  If match have
>>>>
>>>>    been found, then the calculated value is used by the node as TTL
>>>>
>>>>    value in outgoing label to reach the next RTM capable node on the
>>>>
>>>>    LSP.  Otherwise, the TTL value MUST be set to 255.
>>>>
>>>> In essence, the node tries to locate the closest to it downstream 
>>>> RTM-capable node. Otherwise, it sets TTL to 255 so that the RTM 
>>>> packets arrive at egress LER.
>>>>
>>> So this doesn't cover the case where there are holes in the RRO -- which I think then translates to RTM black holes.  Also not covered is what happens during Resv processing if there is no room (or a policy) against adding local information.
>>>
>>> Perhaps it would just be better to add a section that explicitly states that the defined mechanism only works  in cases where full RRO is supported (i.e., where RRO is not limited due to size constraints or otherwise impacted due to policy)? -- not saying I like this, I just think it fills the open gaps in the spec - but at the price of limiting applicability.  I haven't spent too much time thinking about this so there may be a better solution out there.
>>>
>>> Lou
>>>
>>>>  
>>>>
>>>> Parts of Sections 4.6 and 4.7 are a bit rough and could stand a 
>>>> reread and editorial pass.
>>>>
>>>> GIM>> I'm trying.
>>>>
>>>>  
>>>>
>>>> Section 4.7 says PathErr messages are sent in response to Resv 
>>>> processing error.  This is clearly wrong.
>>>>
>>>> GIM>> Great catch! I'm ready to update to ResvErr in -06.
>>>>
>>>>  
>>>>
>>>> Lou
>>>>
>>>>  
>>>>
>>>>> Authors always welcome comments and questions about RTM.
>>>>>             Regards,
>>>>>                             Greg
>>>>> -----Original Message-----
>>>>> From: [e-mail] <mailto:[e-mail]>
>>>> [mailto:[e-mail]]
>>>>
>>>>> Sent: Tuesday, March 15, 2016 2:08 PM
>>>>> To: Alexander Vainshtein; Gregory Mirsky; Stefano Ruffini; John 
>>>>> Drake; Sasha Vainshtein; Eric Gray; Stewart Bryant; Eric Gray
>>>>> Subject: New Version Notification for 
>>>>> draft-ietf-mpls-residence-time-05.txt
>>>>> A new version of I-D, draft-ietf-mpls-residence-time-05.txt
>>>>> has been successfully submitted by Greg Mirsky and posted to the
>>>> IETF repository.
>>>>
>>>>> Name:                               draft-ietf-mpls-residence-time
>>>>> Revision:          05
>>>>> Title:                  Residence Time Measurement in MPLS network
>>>>> Document date:           2016-03-15
>>>>> Group:                              mpls
>>>>> Pages:                               26
>>>>> URL:           
>>>> https://www.ietf.org/internet-drafts/draft-ietf-mpls-residence-time-
>>>> 05
>>>> .txt
>>>>
>>>>> Status:        
>>>> https://datatracker.ietf.org/doc/draft-ietf-mpls-residence-time/
>>>>
>>>>> Htmlized:      
>>>> https://tools.ietf.org/html/draft-ietf-mpls-residence-time-05
>>>>
>>>>> Diff:          
>>>> https://www.ietf.org/rfcdiff?url2=draft-ietf-mpls-residence-time-05
>>>>
>>>>> Abstract:
>>>>>    This document specifies G-ACh based Residence Time Measurement 
>>>>> and
>>>>>    how it can be used by time synchronization protocols being
>>>>>    transported over MPLS domain.
>>>>>    Residence time is the variable part of propagation delay of 
>>>>> timing
>>>>>    and synchronization messages and knowing what this delay is for 
>>>>> each
>>>>>    message allows for a more accurate determination of the delay to 
>>>>> be
>>>>>    taken into account in applying the value included in a PTP event
>>>>>    message.
>>>>>                                                                                  
>>>>> Please note that it may take a couple of minutes from the time of
>>>> submission until the htmlized version and diff are available at 
>>>> tools.ietf.org.
>>>>
>>>>> The IETF Secretariat
>>>>> _______________________________________________
>>>>> Teas mailing list
>>>>> [e-mail] <mailto:[e-mail]> 
>>>>> https://www.ietf.org/mailman/listinfo/teas
>>>>  
>>>>
>>>>  
>>>>
>>>> _______________________________________________
>>>> Teas mailing list
>>>> [e-mail] <mailto:Teas%40ietf.org> 
>>>> https://www.ietf.org/mailman/listinfo/teas
>>>>
>>
>
> _______________________________________________
> Teas mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/teas
>