* WGs marked with an * asterisk has had at least one new draft made available during the last 5 days

Mpls Status Pages

Multiprotocol Label Switching (Active WG)
Rtg Area: Alia Atlas, Alvaro Retana, Deborah Brungard | 1997-Mar-03 —  

IETF-100 mpls minutes

Session 2017-11-14 1330-1530: Sophia - Audio stream - mpls chatroom


minutes-100-mpls-00 minutes

          Welcome to Etherpad!
          This pad text is synchronized as you type, so that everyone viewing
          this page sees the same text. This allows you to collaborate seamlessly
          on documents!
          Get involved with Etherpad at http://etherpad.org
          General updates:
          LDP and MD5
          MD5 is deprecated today.. PALS and MPLS WG writing a draft for new
          security requirements.
          New RFCs published 8223,8227,8234,8256,8277
          Not all authors getting request for status update. Few of the drafts
          were missing status updates.
          Presenter: Stephane Litkowski
          Draft first version Feb 2014.
          Resolved issues from May 2016 WG LC.
          Kireeti: You don't have to read all the labels. Just need to read the EL.
          Stephane: Agree.
          Jeff: Binding SID disapeared from the draft.
          Stephane: Still there for LDP. Can still use it for LDP and LDP to SR.
          Jeff: Introduce a separate draft?
          Acee: Take if offline.
          Greg Mirsky: handing for multiple ELI case. global policies. might be
          useful to have strict policies. use the first one?
          Stephane: From implementation perspective it does not change. Not changing
          the data plane here.
          Loa: When issue a WG LC. Take comments to the WG LC.
          Stephane: multiple possibilities
          Kireeti: all EL should be the same.. hashing on different EL can cause
          Stephane: Should not make is mandatory
          Kireeti: May be hashed differently with multiple ELs.
          Presenter: Stewart Bryant
          Kireeti: Best not to call it SR capable on Transit SR processing slide.
          Stewart: Send you suggestions
          Stewart: Presenting this in spring this week.
          Himanshu: WG related question. There is another identical draft. process
          both drafts or combine them?
          Stewart: Authors are talking.
          Himanshu: Combine two drafts?
          Stewart: Still talking.
          ??: This can support interworking with SR and MPLS native trasport?
          Stewart: Yes
          ??: How to map entropy label?
          Stewart: There is already text on it. Anything missisng please tell us.
          ??: Different SR IGP areas, how does it know which IGP area for the
          first hop label.
          Stewart: How do we know the labels in two diff domains?
          Kireeti: Two SR capable nodes
          Adrian: You do not target SR capable node, you target the node from
          SR stack.
          Ahmed: new behaviour for MPLS, why in this WG? Spring is more
          appropriate. MPLS over GRE and UDP RFCs?
          Adrian: 7510 was here, this is how to use 7510, good simple use case
          of it.
          Ahmed: use-case for SPRING, should be informational. Why standards track?
          Adrian: Yes, make a decision in future.
          Jeff: Entropy label could be good. No difference - describe how mapped,
          additional h/w resources.
          Ahmed: mapping entropy. Nothing to do with SR.
          Ahmed: SR should be in SRPING.
          Ali: difference between this draft and last one?
          Xiaohu: Two are describing the same tech with different styles. Need to
          close some details.
          Loa: Take it to the list. Discuss in spring.
          Martin: We have 10 minutes discussions in spring.
          Greg Mirsky
          Kireeti: LSP ping should be sent with a reply and you don't get a reply.
          Greg: Purpose of bootstrapping of sending with BFD discr is to bootstrap
          BFD session. I want this BFD session and don't care about reply. Else
          you send a reply with error code
          Kireeti: Sender asked for a reply, then reply. Agree with the second
          part but not the first one.
          Greg: Using SHOULD, if you want to reply you can.
          Kireeti: Make sure ping is finished and then start BFD session.
          Greg: this is just a proposal.
          Kireeti: Agree, it doesn't matter the order.
          Greg: Current text is explicit about it. Makes the implementation easy
          this new way.
          Greg Mirsky
          Greg Mirsky
          Zafar Ali: No metnion for seamless BFD. Do not need boot-strapping there.
          Greg: why not seamless BFD, many reasons. Less control on return path.
          Loa: Take it to the list
          Adrian Farrel
          Xiaohu: Proposed two MPLS labels, another containers SPI and SR values?
          Adrian: Two label stack values, yes.
          Xiaohu: Does not use existing MPLS forwarding table.
          Adrian: It will work on existing hardware
          Xiaohu: Currently MPLS forwarding behaviour will change.
          Adrian: Can't speak for your hardware
          Xiaohu: Need stateful service function chaining?
          Adrian: Yes
          Loa: Take it to the mailing list.
          Zafar Ali: before defining the solution, is this the right approach in
          SFC? Starting in MPLS WG is wrong thing to do.
          Adrian: This was already presented in SFC WG today.
          Stephane: It's straight forward. Existing hw. worried about the label
          stack size. We have issues with SRTE with label stack size.
          Adrian: Agree, Long SF paths will have challenges. You can indicate the
          SFF in the label stack. Swapping approach is what I like.
          Stephane: Next generation of hardware?
          Adrian: agree there are limits.
          Nokia ??: Metadata multiple part
          Adrian: Stacking approach. Originally multiple times multiple places. Then
          looked at NSH, metatadata only once.
          Nokia ??:
          Adrian: SFC WG problem, exists in NSH. What to do with metadata you
          don't know.
          Loa: Take it to the mailing list.
          Shaowen Ma
          Jim Guichard: MPLS can't mimick everything NSH is doing. Why we want to
          build Service Chaining in SR?
          Shaowen Ma: Understood
          Jim Guichard: Integrate NSH with SR. Which SR stack to push.
          Shaowen Ma: Good suggession.
          Sam Aldrin: Dataplane into the control plane. What if the SF do not
          support network stack?
          Shaowen Ma: We can remove the SF forward and get back.
          Sam Aldrin: why do you have to push this from dataplane?
          Loa: take it to the list.
          MPLS YANG model status update
          Xufeng Liu
          Greg Mirsky: MPLS OAM Yang model will be ??
          Kamran Raza
          Loa: WG LC befofe the next IETF?
          Kamran: Will decide at that time.
          Loa: Requested yang dr review.
          Stuart: Globally unique?
          Shraddha: SR domian unique.
          Greg: ??
          Shraddha: looking at the solution space
          Greg: MPLS source label draft couple of years back.
          Shraddha: Yes, had looked at it. it was just the source label, which SR
          path from that source.
          Greg: transit node has ECMP, so you still losing the node travered.
          Shraddha: Figure out how much traffic for an SR path not to discover
          the path.
          Jeff: special label and counters.
          Shraddha: forwarding plane will look for the SPI label and below that
          is an identifier.
          Jeff: Generic MPLS label. SR is a special label anyways.
          Shraddha: That could be one of the solutions. allocate a special global
          block and provision on each node. Don't think SRLB can be used. We need
          a new global block.
          Jeff: you don't need two labels, just need one.
          Shraddha: you really need a special label. don't exist today. one of the
          solutions. you can achieve with one label. you really need to dedicate
          a global label block. 100K labels may not be sufficient for example.
          Stewart: never use a special purpose label. Couple of 1000 nodes and
          use that many labels to indicate them.
          Mach Chen: every solution one or more labels has pros and cons.
          Jeff: every hw vendor, huge cost associated 10+ label.
          Shraddha: Label stack depth is a concern, agree
          Jeff: Binding SID to create a sub-path. Need to extend BGP.
          Stewart: IP can do this lot of it already. Use existing technology.
          Shraddha: using IP source/dest?
          Stewart: using netflow.
          ?? : global label block - good suggestion
          ?? : node may carry thousands of labels, need counters
          Shraddha: read draft - describes solutions
          MPLS ECMP Issues
          Stewart Bryant

Generated from PyHt script /wg/mpls/minutes.pyht Latest update: 24 Oct 2012 16:51 GMT -