Re: [mpls] MPLS WG adoption call for draft-mirsky-mpls-p2mp-bfd

["Carlos Pignataro (cpignata)" <[e-mail]>]

Hi Greg,

On Nov 21, 2018, at 7:00 AM, Greg Mirsky <[e-mail]<mailto:[e-mail]>> wrote:

Hi Carlos,
apologies for the prolonged silence. Thank you for your consideration of the proposed new text and the acknowledgment that we're converging.

I do not recall (nor can I find below) any text from me with any acknowledgement of convergence. That is a misrepresentation.

I do see you (not I) wrote below again (“Glad that we're converging.”) I do not see the basis for that.

I did tell Loa I saw progress (not convergence) but also that my main question remained, in my humble opinion, unsatisfactorily answered.


Please find the new comments in-line tagged GIM3>>.

Regards,
Greg

On Wed, Nov 7, 2018 at 8:52 PM Carlos Pignataro (cpignata) <[e-mail]<mailto:[e-mail]>> wrote:
[Greg, Loa, responding to both on this single email reply]

Hi, Loa,

On Nov 6, 2018, at 1:49 PM, Loa Andersson <[e-mail]<mailto:[e-mail]>> wrote:

Carlos,

Since the a wg adoption poll I read your comments as that we are doing
progress, and that we can address the rest during the wg process,
correct?


I agree we are making progress, thank you. Most questions can be addressed later, but only the very first question goes to the heart of an adoption poll. If we can close on that, the rest can be addressed later (note the same question is related also to the last question.)

/Loa

Hi, Greg,

Thank you very much for your responses — please see inline.

On Nov 6, 2018, at 7:18 PM, Greg Mirsky <[e-mail]<mailto:[e-mail]>> wrote:

Hi Carlos,
thank you for your consideration of my responses. Glad that we're converging. Please find additional notes in-line tagged GIM2>>.

Regards,
Greg

On Tue, Nov 6, 2018 at 12:11 AM Carlos Pignataro (cpignata) <[e-mail]<mailto:[e-mail]>> wrote:
Hi Greg,

Many thanks for your response and suggestions! Please see inline.

On Nov 2, 2018, at 6:13 AM, Greg Mirsky <[e-mail]<mailto:[e-mail]>> wrote:

Hi Carlos,
thank you for your comments. Please find my notes, answers in-line tagged GIM>>.

Regards,
Greg

On Thu, Oct 25, 2018 at 8:47 PM Carlos Pignataro (cpignata) <[e-mail]<mailto:[e-mail]>> wrote:
Hi,

Cc BFD WG

It would be useful to understand the use case motivation or applicability of this draft, other than it can be done.
GIM>>  The motivation can be seen in the following (from another draft that discusses OAM over G-ACh:
  In some
   environments, the overhead of extra IP/UDP encapsulations may be
   considered as overburden and make using more compact G-ACh
   encapsulation attractive.
Will add text in the draft.

CMP: Thank you very much. This is a good start, although it would be useful to add precision into which environments specifically, and the burden comparison between IP/UDP and G-ACh.
GIM2>> Thank you for agreeing to this, and I've added the text in the working verion. Will work on improving the text in the meantime.

CMP: Sorry if I was not clear. Like I said, this is a good start and probably necessary (but not sufficient) text.

CMP: Which environments specifically? At this point, the scope and target of the work is not clear to me. That was my question. Is this for MPLS-TP P2MP? If so, the underlying seems to have stalled:
https://datatracker.ietf.org/doc/rfc7167/referencedby/
CMP: I think these two questions should be answered: 1. What specific environments? 2. How current solutions do not solve it (i.e., what is and can we quantify the overburden)?
GIM3>> Andy Malis has pointed to the requirements for proactive OAM, particularly monitoring path continuity, listed in Section 4.1 RFC 4687.

Just to understand — the basis of this work is Requirements from RFC 4687 from the year 2006?

These are not specific to MPLS-TP but to OAM over p2mp MPLS LSP. The following text has been added to the Security Considerations section in the recenly uploaded -04 version of the draft:

Adding scope and rationale for some work in the Security Considerations does not seem like the right sequentiality to set the stage.

   Also, BFD for p2mp MPLS LSP MUST follow the requirements listed in
   section 4.1 [RFC4687] to avoid congestion in the control plane or the
   data plane caused by the rate of generating BFD control packets.  An
   operator SHOULD consider the amount of extra traffic generated by
   p2mp BFD when selecting the interval at which the MultipointHead will
   transmit BFD control packets.  Also, the operator MAY consider the
   size of the packet the MultipointHead transmits periodically as using
   IP/UDP encapsulation adds up to 28 octets, which is more than 50% of
   BFD control packet length, comparing to G-ACh encapsulation.



I’m also increasingly concerned by confusing scope and definition of specifications.

For example:

https://tools.ietf.org/html/draft-mirsky-mpls-p2mp-bfd-04#section-3.2

3.2.  Non-IP Encapsulation of Multipoint BFD

   Non-IP encapsulation for multipoint BFD over p2mp MPLS LSP MUST use
   Generic Associated Channel (G-ACh) Label (GAL) [RFC5586] at the
   bottom of the label stack followed by Associated Channel Header
   (ACH).  Channel Type field in ACH MUST be set to BFD CV [RFC6428].


First, there’s no definition for non-IP BFD in RFC 5586 — only in RFC 5885.
GIM>> RFC 5586 defined the use of GAL. I think that this reference is appropriate. I agree that the second reference should be to RFC 5885, not RFC 6428. Will make the change.

CMP: Thank you. However, RFC 5885 is in the context of PW VCCV — is there a missing definition in the specs for BFD over G-ACh generically?
GIM2>> I think that the following quote from RFC 5586 set the relationship between Channel Type field in PW ACH and G-ACh:
    Channel Types for the Associated Channel Header are allocated from
    the IANA "PW Associated Channel Type" registry [RFC4446].
I understand that that there's one and only one registry and channel values are equally applicable to PW ACH and G-ACh. And full name of the registry now is MPLS Generalized Associated Channel (G-ACh) Types (including Pseudowire Associated Channel Types).

CMP: That is correct. I was curious as to whether additional control plane is needed for this support.


Second, the specification in RFC 6428 applies to MPLS Transport Profile only. NOT for MPLS, and explicitly NOT for P2MP!

https://tools.ietf.org/html/rfc6428#section-1

   This document specifies the BFD extension and behavior to satisfy the
   CC, proactive CV monitoring, and the RDI functional requirements for
   both co-routed and associated bidirectional LSPs.  Supported
   encapsulations include Generic Associated Channel Label (GAL) /
   Generic Associated Channel (G-ACh), Virtual Circuit Connectivity
   Verification (VCCV), and UDP/IP.  Procedures for unidirectional
   point-to-point (P2P) and point-to-multipoint (P2MP) LSPs are for
   further study.


So, no, this does not work.

RFC 6428 does not have scope for P2MP.
And RFC 5586 does not specify anything for BFD. Instead, what needs to be cited (appropriately and expanded) is RFC 5885
GIM>> RFC 5586 specifies the use of GAL and G-ACh and the reference is used in this context.

CMP: This is the same comment as above.


https://tools.ietf.org/html/rfc6428#section-4
      RFC 5884 - BFD CC in UDP/IP/LSP
      RFC 5885 - BFD CC in G-ACh
GIM>> I'd point that it is for p2p BFD CC, and p2mp BFD uses different from p2p BFD method to demultiplex BFD control packets.


CMP: Apologies I did not understand this response.
GIM2>> Apologies for sending partial explanation. P2MP BFD cannot use Your Discriminator field to demultiplex the recieved BFD control packet. BFD for Multipoint Networks defines the special procedure that requires the use of Source ID. When the encapsulation of BFD control packet does not include IP/UDP header, the Source ID can be provided as Source MEP-ID TLV in MPLS-TP BFD CV. This draft proposes the new IP Address TLV for that. Thus I have to correct myself and re-state the earlier text in the draft that the value in the Channel Type filed of G-ACh must be MPLS-TP CV (0x0023).

CMP: I understood you said above that the reference to RFC6428 was incorrect.

CMP: Now, just to understand the approach:

CMP: Are you suggesting that the IP header is not used with BFD and instead a new TLV (of which information structure?) carries the IP address that you removed before? Seems like a musical-chairs arrangement of the data. I may very likely be missing something. Apologies in advance if that is the case.

CMP: Also, is the applicability MPLS-TP? What is the normative reference for MPLS-TP P2MP?
GIM3>> I should have explained why I think that MPLS-TP CV message (0x0023) type is more suitable than BFD Control, PW-ACH encapsulation (without IP/UDP Headers) (0x0007). The latter includes only the BFD control packet while the format of the former includes Source MEP-ID TLV that immediately follows the BFD control packet. And with the new proposed IP Address TLV the 0x0023 G-ACh channel works better for p2mp BFD over p2mp MPLS LSP. Alternative, of course, would be to define the new G-ACh type for p2mp BFD over p2mp MPLS LSP.


Thanks — I appreciate that a lot of packet formats could be drawn. My question is really regarding motivation for the work (the vague “ In some environments”)

Thanks,

— Carlos.

Thanks,

Carlos.


CMP: Thanks again for considering the comment to improve the document.

Thanks,

Carlos.


      RFC 5085 - UDP/IP in G-ACh
       MPLS-TP - CC/CV in GAL/G-ACh or G-ACh



Thanks,

— Carlos Pignataro

On Oct 13, 2018, at 4:24 PM, Greg Mirsky <[e-mail]<mailto:[e-mail]>> wrote:

Dear WG Chairs, et al.,
as the author of the BFD for Multipoint Networks over Point-to-Multi-Point MPLS LSP (draft-mirsky-mpls-p2mp-bfd) I would like to ask you to consider WG adoption call of the draft. The document addresses non-IP encapsulation of p2mp BFD over MPLS LSP that may be useful if the overhead of IP, particularly IPv6, encapsulation is the concern. The base specification of BFD for Multipoint Networks is at this time in IESG LC.

Regards,
Greg
_______________________________________________
mpls mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/mpls