Re: [Idr] [Bier] FW: New Version Notification for draft-xu-idr-bier-extensions-00.txt

[Antoni Przygienda <[e-mail]>]

      Length:5

      Value: contains a two-octet BFR-ID.



I am missing here the indication of subdomain-id since without that, the BFR-ID will easily collide within the BIER domain advertising the BGP info @ the edge.



[Xiaohu] Thanks for point it out. Will fix it in the next version.



[Tony said] ack



The BSL TLV is encoded as follows:



      Type:TBD3



      Length:4



      Value: contains a one-octet BSL which indicates the length of the

      Bitstring in 4-octets.



Is this encoded equivalent to MPLS draft ? Indicate so and indicate WHICH bits are used and which are reserved. If not, indicate how e.g.  33 as bitstring length should be treated if e.g. the ensuing encaps cannot deal with it ?



[Xiaohu] IMHO, it'd better to specify the legal BSL values in the BIER arch draft which should be independent of specific BIER encapsulations.



[Tony said]  agreed, that's a heads-up for Ice to put it into the architecture draft me thinks. It's hanging in the MPLS draft right now. Given how widely it's referenced, it should be moved up me thinks.



   The BIER MPLS Encapsualtion TLV is encoded as follows:



      Type:TBD4



      Length:7



      Value: contains a one-octet Label Range Size field indicating the

      size of the label range, and a 3-octect Label Rang Base field

      where the 20 rightmost bits represent the first label in the label

      range.



indicate that the other bits are reserved/must be ignored/send 0



Generally, it is considered de rigeur to make nice pictures of those TLVs to make the documents more readable. I suggest doing so.



[Xiaohu] Thanks for the above suggestions. Will fix them in the next version.



[Tony said] ack



Another question here is what happens when multiple eBPG edges of an AD advertise this attribute. Is the preference driven by MED [or another attribute?] based e.g. on the IGP distance within AD to the BFR ?



[Xiaohu] I think so.



[Tony said] ack, I suggest to put it into lead-in section (Special considerations, requireements, assumptions, something ?)



Is the expectation that each BGP router advertising this is BFR itself ? Or has capabilities to unicast tunnel into the BIER domain ?



[Xiaohu]Since the BIER attribute is an optional, transitive BGP path attribute, it should support the scenario where a BGP speaker is allowed to tunnel a BIER packet or the payload of a BIER packet to the BFER directly if the BGP next-hop is a non-BFR. Furthermore, it should support as well the scenario where a BGP speaker is allowed to tunnel a BIER packet to the BGP next-hop if these two BGP peers are not directly connected (e.g., multi-hop EBGP) .



[Tony said] [Tony said] Yes, I suggest a small section stating this and explaining this.



Overall, this seems to imply somewhat that BIER domain is aligned with the whole IGP domain (and not areas) and e.g. two OSPF areas cannot be two BIER domains.  This is an interesting architectural question that wasn't brought up yet ?





[Xiaohu] In this doc, it's assumed that a BIER domain is aligned with the whole network consist of multiple ASes.

[Tony said] Let's spell it out and have a discussion @ next BOF/WG [whatever we become when we've grown up to have a charter ;-) ]  I think it will be a very relevant architectural decision at this point in time and possibly influence all IGP drafts. What speaks against BIER domain aligned with AD is scalabilty, what speaks for it is simplicity.

--- tony