The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2004-Aug> msg00030



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

[mpls] Requesting your feedback - issues/errors/clarifications

  • From: Luca Martini <lmartini@cisco.com>
  • Date: Wed, 25 Aug 2004 09:33:51 -0600
  • Cc: mpls@ietf.org
  • Organization: Cisco Systems
  • User-Agent: Mozilla/5.0 (X11; U; SunOS i86pc; en-US; rv:1.7) Gecko/20040620
  • X-from-outside-Cisco: [10.32.224.186]
  • X-PMX-Version: 4.6.0.99824



Vach Kompella wrote:

>I agree we should get rid of the host FEC type.
>
>Issue: it would be nice to be able to shut down an adjacency without
>having to time it out.  E.g., two interfaces between the same pair of
>nodes running LDP.  If one is shut down, there is no way to notify the
>peer endpoint, e.g., notification message to the multicast address over
>UDP.
>
>Issue: do we really have to send all FECs in our database whenever we
>have an LDP session between two peers?  E.g., A is a PE which does LDP
>for address FECs as well as PW FECs.  If it sets up a targeted adjacency
>clear across the network to another PE, does it really have to send all
>its address FECs there?  In our implementation, we do not send address
>FECs across a session that was set up with a targeted adjacency, since
>we only use targeted adjacencies for PW FECs.
>
>Issue: with all the combinations of Downstream Unsolicited, Downstream
>on Demand, Ordered Control, Independent Control, etc., it makes sense to
>define a mandatory combination.  DU/OC seems to be the favored one.
>E.g., IC was useful back when MPLS LSPs replaced IP forwarding.  Now
>that that isn't the case anymore, if there is a P router doing
>independent control in the middle of the network, it propagates a FEC
>for a tunnel that may be used by 2547, BGP-free core routing, or PWs,
>falsely giving the impression that the PSN tunnel is end to end.  It
>would be far better to identify, at ingress, that the tunnel is not up.
>In either case, we blackhole packets, but in one case, we clearly know
>the service is not up, the route is not resolvable, etc., while in the
>other, you have to figure that there is a problem somewhere in the
>middle of the network.
>
>  
>
Actually I disagree. If anyting we should have DU/IC , the ordered 
control caused lot's of operational problems , and it took almost a year 
to get it working properly. There are a number of situations that cause 
problems in ordered control mode , but not in independent control. Most 
of these issues are due to timing.

>Issue: Minor optimization: why do you send a FEC back to the owner of
>the FEC?  E.g., A sends its loopback to B.  Why should B send it right
>back to A (as described in LMp.21)?
>
>Issue: Minor optimization: if A is a stub node, i.e., only one LDP
>session, does it really have to send a label mapping for every FEC that
>it has, or can it do it lazily, e.g., when it has a second LDP session?
>
>  
>
No sure I understand what you mean here.  LDP cannot know what path is 
selected, so it must send all the FECs.

Luca

>-Vach 
>
>  
>
>>-----Original Message-----
>>From: mpls-bounces@lists.ietf.org 
>>[mailto:mpls-bounces@lists.ietf.org] On Behalf Of Markus Jork
>>Sent: Tuesday, August 24, 2004 1:17 PM
>>To: Ina Minei
>>Cc: mpls@ietf.org
>>Subject: Re: [mpls] Requesting your feedback - 
>>issues/errors/clarifications
>>
>>
>>    
>>
>>>	RFC3036 (LDP specification) is advancing to draft standard.
>>>
>>>	As part of this process, it is necessary to compile a list of 
>>>changes/clarifications to the rfc, based on the experience 
>>>      
>>>
>>gained with 
>>    
>>
>>>the protocol.
>>>
>>>	Please send me errors/changes/clarifications that you 
>>>      
>>>
>>know about, so 
>>    
>>
>>>that they can be taken into account in the revised document. Please 
>>>send your comments by September 6, either directly to me or to the 
>>>list.
>>>      
>>>
>>Here is one issue:
>>The spec defines two types of FECs: "address prefix" and 
>>"host address". The address prefix FEC is what's in 
>>widespread use. We have also implemented support for the host 
>>address FEC but I have never seen it used. It would be 
>>interesting to know whether there is any known actual use of 
>>it. If not, I suggest to remove it from the spec.
>>
>>Markus
>>
>>
>>    
>>
>>>	I will post the complete list of the issues a couple of 
>>>      
>>>
>>weeks later.
>>    
>>
>>>		Thank you,
>>>
>>>			Ina
>>>      
>>>
>>
>>_______________________________________________
>>mpls mailing list
>>mpls@lists.ietf.org
>>https://www1.ietf.org/mailman/listinfo/mpls
>>
>>    
>>
>
>
>_______________________________________________
>mpls mailing list
>mpls@lists.ietf.org
>https://www1.ietf.org/mailman/listinfo/mpls
>
>
>  
>

_______________________________________________
mpls mailing list
mpls@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/mpls