Re: [sfc] Progression of OAM work in the SFC WG

[Greg Mirsky <[e-mail]>]

Dear Carlos,
top-posting your questions addressed to my proposal to deprecate the O-bit.
Please find my answers and notes in-lined under the GIM>> tag below.
CMP: RFC 8924 already includes ICMP and BFD as example SFC Action OAM
protocols. Those can be encapsulated in IP, in which the Next-Protocol
indicates IP — and (as per RFC 8300) the O-bit is set to indicate an OAM
packet.
GIM>> I cannot find in RFC 8924 the explanation of the encapsulation of the
BFD Control message. Should we assume that IPv4/IPv6 is assumed? If that is
the case, which IP address is to be used as the destination IP address? And
one more, What special processing is ensured by setting the O bit?
CMP: The proposal of deprecating the O-bit breaks those SFC OAM protocols
already included in an RFC.
GIM>> Could you please clarify in which part the removal of the O bit
breaks, for example, ICMP or BFD?
CMP: That proposal of deprecating the O-bit also breaks Section 4.1 of IOAM.
GIM>> Could you please clarify to which IOAM document you are referring?
CMP: This expired I-D yet implemented in open source uses the O-bit:
https://www.ietf.org/archive/id/draft-penno-sfc-trace-03.txt
GIM>> I recall that the WG discussed and agreed to significantly modify the
NSH format. That possibly had affected some existing implementations.

Regards,
Greg



On Sun, Feb 13, 2022 at 8:06 PM Carlos Pignataro (cpignata) <
[e-mail]> wrote:

> Replying to this specific email in the thread, only because it is the
> latest one at this point in time, but bundling comments from various places
> on the thread. Please find some follow-up comments and questions marked
> “CMP”, indicating the “From”, and organized chronologically:
>
> *James N Guichard <[e-mail]
> <[e-mail]>>:*
>
> 1) The chairs have reviewed the O bit definition in RFC 8300.  That
> definition is at best open to interpretation and therefore incomplete.  For
> example, the clear intention is only to mark packets which are intended for
> SFC OAM at the SFC service layer.  But that is not what the current text
> says.  There is also, unfortunately, ambiguity as to what constitutes an
> OAM packet.  So it is reasonable for documents to update 8300 to clarify
> the exact applicability and action for the O-bit.
>
>
> CMP: It is unclear to me what exact issues the chairs are calling out.
> There are two issues mentioned:
> CMP: 1. "But that is not what the current text says.” —> what does the
> text say precisely?  Can you point to the specific RFC 8300 text that is
> believed to be incomplete, and a detailed use case that showcases the
> potential issue? An extrapolated counter-example just to make a point: the
> text does not  say what days of the week OAM packets can be marked, yet
> there’s an expectation that they will work Mondays and Saturdays all the
> same — even holidays.
> CMP: 2. What specifically is the ambiguity with what an OAM packet is?
> Please grep the RFC series for “oam packet”, it is not a new term.
> CMP: Net-net, O-bit=1, OAM packet.
>
> 2) However, related to point 1), we can't have multiple documents updating
> the definition differently.  As such, the authors of the SFC iOAM draft and
> the SFC multi-layer-oam draft need to come together and figure out what the
> clarification is for the definition of that bit. We do not believe as
> chairs that either of these documents can move forward from the WG until
> such clarity has been reached.
>
>
> CMP: If there is still believed to be an ambiguity or a clear opportunity
> for improving the definition, I agree with consolidating that update. I had
> suggested before that if such new text or real clarification is agreed, a
> separate document that only makes that O-bit clarification is the most
> clear and clean way to go. That doc should be a one-pager or less. I’d be
> happy to contribute to that minimal doc, if it is shown to be needed.
>
> CMP: For completeness also, the "SFC multi-layer-oam” document has many
> outstanding issues besides the O-bit.
>
> *Greg Mirsky <[e-mail] <[e-mail]>>*
>
> I've reviewed our SFC OAM documents and draft-ietf-sfc-nsh-tlv. As I
> understand these documents, the Active SFC OAM and IOAM are identified by
> the respective values for the NSH Next Protocol field (to be assigned by
> IANA).
>
>
> CMP: There is a major issue with the sentence above. Active SFC OAM (as a
> category), as per RFC 8300, needs to be identified by setting the O-bit. No
> way around that based on RFC 8300. However, the text "the Active SFC OAM”
> creates confusion since it refer to a specific OAM protocol which has the
> same name as a the category as well. That name collision is quite
> unfortunate in my opinion.
>
> At the same time, so far no OAM-specific meta data TLV has been defined.
> Thus, it appears that one way forward could be to not involve the O bit in
> the active SFC OAM or IOAM altogether. In other words, to deprecate the NSH
> O bit.
>
>
> CMP: RFC 8924 already includes ICMP and BFD as example SFC Action OAM
> protocols. Those can be encapsulated in IP, in which the Next-Protocol
> indicates IP — and (as per RFC 8300) the O-bit is set to indicate an OAM
> packet.
> CMP: The proposal of deprecating the O-bit breaks those SFC OAM protocols
> already included in an RFC.
> CMP: That proposal of deprecating the O-bit also breaks Section 4.1 of
> IOAM.
> CMP: This expired I-D yet implemented in open source uses the O-bit:
> https://www.ietf.org/archive/id/draft-penno-sfc-trace-03.txt
>
>
> *Gyan Mishra <[e-mail] <[e-mail]>>*
>
> I agree that what makes sense to me as a path forward is to deprecate the
> O bit and not use in SFC Multilayer OAM and SFC IOAM, as both SFC
> Multilayer OAM and SFC IOAM are identified by the respective values for the
> NSH Next Protocol field (to be assigned by IANA), as well as so far no
> OAM-specific meta data TLV has been yet defined.
>
>
> CMP: What is the “SFC Multilayer OAM”?
> CMP: The O-bit (reading RFC 8300) does not indicate if there is OAM
> Metadata. It indicates an OAM packet.
>
>
> *"Joel M. Halpern" <[e-mail] <[e-mail]>>*
>
> As far as I can tell, if we deprecate the O-bit and rely on the next
> protocol field, we are saying that in practice (not by rule) NSH metadata
> can not be used for OAM.  That's fine with me as long as we agree on that.
>
>
> CMP: I do not fully follow this. Can you explain the relationship of the
> O-bit with Metadata?
> CMP: Here’s an example: RFC 8924 shows ICMP as Active OAM. The SFC next
> protocol is IP, the IP next protocol is ICMP, the O-bit needs to be set.
>
> CMP: Consider the following document, independent submission and in the
> RFC Publication queue already:
> CMP:
> https://datatracker.ietf.org/doc/draft-mymb-sfc-nsh-allocation-timestamp/
> CMP: Is a Timestamp in Metadata an OAM element?
>
> CMP: Consider this proposal
> https://datatracker.ietf.org/doc/draft-mirsky-sfc-pmamm/
> CMP: Where the base header borrows a bit for an OAM function.
>
> CMP: Neither of those have any impact on the O-bit.
>
>
> *Greg Mirsky <[e-mail] <[e-mail]>>*
>
> I agree that, as we find SFC NSH now, no OAM metadata has been defined. It
> appears to me that deprecating the O bit does not affect any of the already
> defined mechanisms in the SFC NSH.
>
>
> CMP: It actually does, see my example above.
>
> I think that if a new, for example, MD Type = 2 TLV that is used for any
> OAM functionality to be proposed in the future, deprecated the O bit
> would not prevent using such NSH TLV.
>
>
> CMP: I agree with this — because there is no correlation between MD Type
> field and O-bit as defined.
>
> Best,
>
> Carlos.
>
>
>
> 2/13/22 午後12:32、Gyan Mishra <[e-mail]>のメール:
>
> Hi Joel
>
> I am in agreement as well with what you have stated that once the O bit is
> deprecated that NSH metadata cannot be used for OAM and that we would have
> to now from that point forward rely on the next protocol field as we will
> be doing for SFC Multilayer and SFC IOAM drafts as well as will have
> backwards compatible with any already existing defined mechanisms in SFC
> NSH.
>
> I don’t see any issues or impact with moving forward.
>
> Kind Regards
>
> Gyan
>
> On Sun, Feb 13, 2022 at 12:04 PM Greg Mirsky <[e-mail]>
> wrote:
>
>> Hi Joel,
>> I agree that, as we find SFC NSH now, no OAM metadata has been defined.
>> It appears to me that deprecating the O bit does not affect any of
>> the already defined mechanisms in the SFC NSH. I think that if a new, for
>> example, MD Type = 2 TLV that is used for any OAM functionality to be
>> proposed in the future, deprecated the O bit would not prevent using such
>> NSH TLV.
>> What do you think?
>>
>> Regards,
>> Greg
>>
>> On Sun, Feb 13, 2022 at 6:55 AM Joel M. Halpern <[e-mail]>
>> wrote:
>>
>>> There is an implication of deprecating the O-bit that I would like to
>>> hear from more WG participants about.
>>>
>>> As far as I can tell, if we deprecate the O-bit and rely on the next
>>> protocol field, we are saying that in practice (not by rule) NSH
>>> metadata can not be used for OAM.  That's fine with me as long as we
>>> agree on that.
>>>
>>> Yours,
>>> Joel
>>>
>>> On 2/13/2022 2:52 AM, Gyan Mishra wrote:
>>> >
>>> > Hi Jim, Joel & SFC WG,
>>> >
>>> > I agree that the RFC 8300 definition of O bit is incomplete and not
>>> > clear as to its intended use.
>>> >
>>> > That is a problem that I agree needs to be rectified.
>>> >
>>> > I understand that we need to get this resolved before we can progress
>>> > Multilayer SFC OAM draft-ietf-sfc-multi-layer-oam-18 and SFC IOAM.
>>> >
>>> > I agree that what makes sense to me as a path forward is to deprecate
>>> > the O bit and not use in SFC Multilayer OAM and SFC IOAM, as both SFC
>>> > Multilayer OAM and SFC IOAM are identified by the respective values
>>> for
>>> > the NSH Next Protocol field (to be assigned by IANA), as well as so
>>> far
>>> > no OAM-specific meta data TLV has been yet defined.
>>> >
>>> > So we have I believe solid solution and path forward and I support
>>> > deprecating the O bit.
>>> >
>>> > Kind Regards
>>> >
>>> > Gyan
>>> >
>>> > On Wed, Feb 2, 2022 at 5:42 PM Greg Mirsky <[e-mail]
>>> > <mailto:[e-mail]>> wrote:
>>> >
>>> >     Thank you, Jim and Joel, for guiding the SFC OAM work and pointing
>>> >     out the issue that must be addressed.
>>> >
>>> >     I've reviewed our SFC OAM documents and draft-ietf-sfc-nsh-tlv. As
>>> I
>>> >     understand these documents, the Active SFC OAM and IOAM are
>>> >     identified by the respective values for the NSH Next Protocol field
>>> >     (to be assigned by IANA). At the same time, so far no OAM-specific
>>> >     meta data TLV has been defined. Thus, it appears that one way
>>> >     forward could be to not involve the O bit in the active SFC OAM or
>>> >     IOAM altogether. In other words, to deprecate the NSH O bit.
>>> >
>>> >     I greatly appreciate your comments on the proposal to deprecate the
>>> >     NSH O bit.
>>> >
>>> >     Regards,
>>> >     Greg
>>> >
>>> >     On Wed, Feb 2, 2022 at 10:36 AM James Guichard
>>> >     <[e-mail]
>>> >     <mailto:[e-mail]>> wrote:
>>> >
>>> >         Hi WG:____
>>> >
>>> >         __ __
>>> >
>>> >         Having reviewed all of the OAM related documents in our WG, the
>>> >         chairs would like to provide a few comments to hopefully
>>> >         generate discussion and forward progress of this work:____
>>> >
>>> >         __ __
>>> >
>>> >         1) The chairs have reviewed the O bit definition in RFC 8300.
>>> >         That definition is at best open to interpretation and therefore
>>> >         incomplete.  For example, the clear intention is only to mark
>>> >         packets which are intended for SFC OAM at the SFC service
>>> >         layer.  But that is not what the current text says.  There is
>>> >         also, unfortunately, ambiguity as to what constitutes an OAM
>>> >         packet.  So it is reasonable for documents to update 8300 to
>>> >         clarify the exact applicability and action for the O-bit.____
>>> >
>>> >         __ __
>>> >
>>> >         2) However, related to point 1), we can't have multiple
>>> >         documents updating the definition differently.  As such, the
>>> >         authors of the SFC iOAM draft and the SFC multi-layer-oam draft
>>> >         need to come together and figure out what the clarification is
>>> >         for the definition of that bit. We do not believe as chairs
>>> that
>>> >         either of these documents can move forward from the WG until
>>> >         such clarity has been reached. ____
>>> >
>>> >         __ __
>>> >
>>> >         3) Related to the SFC iOAM, we need a clear definition of
>>> iOAM.
>>> >         There seem to be differences between the definitions in
>>> >         published RFCs, the usage (which is not a definition) in the
>>> SFC
>>> >         draft, and the various ippm drafts.  Any such definition will
>>> >         need to be vetted with the ippm working group.____
>>> >
>>> >         __ __
>>> >
>>> >         Again, it would be good if members of the working group beyond
>>> >         the two author teams spoke up about their readings of the
>>> >         documents, and their understandings of what we need.____
>>> >
>>> >         __ __
>>> >
>>> >         Yours,____
>>> >
>>> >         Jim and Joel____
>>> >
>>> >         __ __
>>> >
>>> >         __ __
>>> >
>>> >         __ __
>>> >
>>> >         _______________________________________________
>>> >         sfc mailing list
>>> >         [e-mail] <mailto:[e-mail]>
>>> >         https://www.ietf.org/mailman/listinfo/sfc
>>> >         <https://www.ietf.org/mailman/listinfo/sfc>
>>> >
>>> >     _______________________________________________
>>> >     sfc mailing list
>>> >     [e-mail] <mailto:[e-mail]>
>>> >     https://www.ietf.org/mailman/listinfo/sfc
>>> >     <https://www.ietf.org/mailman/listinfo/sfc>
>>> >
>>> > --
>>> >
>>> > <http://www.verizon.com/>
>>> >
>>> > *Gyan Mishra*
>>> >
>>> > /Network Solutions A//rchitect /
>>> >
>>> > /Email [e-mail] <mailto:[e-mail]>//
>>> > /
>>> >
>>> > /M 301 502-1347
>>> >
>>> > /
>>> >
>>> >
>>>
>> --
>
> <http://www.verizon.com/>
> *Gyan Mishra*
> *Network Solutions A**rchitect *
> *Email [e-mail] <[e-mail]>*
>
>
> *M 301 502-1347 *
>
> _______________________________________________
> sfc mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/sfc
>
>
>