Hi, Greg,

Adding the IPPM mailing list, since you reference it below, and point to RFC 
7799.

Please see inline.

> On Apr 17, 2019, at 3:00 PM, Greg Mirsky <[email protected]> wrote:
> 
> Hi Carlos,
> thank you for sharing your view on the design of the VXLAN-GPE protocol. 
> Please find my comments in-line tagged GIM>>.
> 
> Regards,
> Greg
> 
> On Fri, Apr 12, 2019 at 5:23 AM Carlos Pignataro (cpignata) 
> <[email protected]> wrote:
> Greg,
> 
> You seem to be confusing and mixing things up.
> 
> Please see inline.
> 
> > On Apr 12, 2019, at 2:50 AM, Greg Mirsky <[email protected]> wrote:
> > 
> > Dear Editors, et al.,
> > I've read the latest -07 version and would like to share my comments and 
> > questions with you:
> >       • in the earlier version of the draft, the O-bit was introduced and 
> > defined as:
> > OAM Flag Bit (O bit):  The O bit is set to indicate that the packet is an 
> > OAM packet.
> > At the same time, in the latest version, the new Next Protocol value (0x81) 
> > to identify iOAM Data was introduced. Hence are my questions:
> >       • What must be the value of the O-bit when the value of the Next 
> > Protocol field is iOAM?
> 
> The O bit “is set to indicate that the packet is an OAM packet”.
> IAOM is not an OAM packet. Hence, O bit clear if VXLAN-GPE encapsulates a 
> non-OAM packet and adds an IOAM shim.
> GIM>> It is a very unexpected statement from you,

I would tell you ‘expect the unexpected’, but that is not the case here... :-)

If you actually read my statement, you will see "IAOM is not an OAM *packet*.” 
New emphasis added — it is _not_ *an OAM packet*.

I stand by it. It’s an augmented data packet of interest.

> one of the co-authors of draft-ietf-ippm-ioam-data  where the first paragraph 
> of the Introduction section defines iOAM as Hybrid Type 1 OAM, according to 
> RFC 7799 classification:

This is still quite consistent with that definition, which I recommend you 
re-read:

RFC 7799 is about metrics and methods, not packet definitions.

You will see that Active methods include a “dedicated measurement packet 
stream”. “Active Methods generate packet streams.”

However, Hybrid Type I leverages “Augmentation or modification of the stream of 
interest"

Following your logic, if you change a packet field such as TTL/HL to provide 
packet coloring, do you need to set the O-bit?

>    IOAM is to complement
>    mechanisms such as Ping or Traceroute, or more recent active probing
>    mechanisms as described in [I-D.lapukhov-dataplane-probe].  In terms
>    of "active" or "passive" OAM, "in-situ" OAM can be considered a
>    hybrid OAM type.  While no extra packets are sent, IOAM adds
>    information to the packets therefore cannot be considered passive.
>    In terms of the classification given in [RFC7799] IOAM could be
>    portrayed as Hybrid Type 1.
> 
> >       • Do you plan to define the Next Protocol values for active OAM 
> > protocols, e.g., Echo Request/Reply, BFD, Performance Monitoring?
> 
> I cannot speak to “plans” (replying as “et al.”, not as “editor”), but I 
> expect OAM documents ought to have those requested, and you do not need 
> necessarily all of those — the nex protocol IPv4 / IPv6 can encapsulate ICMP, 
> UDP/BFD, etc., etc., etc., etc.
> GIM>> Of course, use of the well-known port number, usually it is UDP 
> destination port, is one of the methods to demultiplex protocols. This method 
> is broadly used, for example, in MPLS OAM. But this method has some 
> disadvantages that were pointed out in the discussion of BFD over GENEVE in 
> Prague by David Black (the last presentation in the minutes):
> David: I want the BFD header to be as close to the thing whose liveness I'm 
> testing. The more headers you add in the middle, the more ways there are to 
> break BFD without breaking the forwarding engine. The further away I move BFD 
> from the chunk of hardware's liveness I care about, the more opportunities 
> are for BFD and the hardware to not to tell me the same thing.
> 

Theory and practice match in theory, less so in practice — unless you are 
planning to update RFC 5881.

This can be argued both ways (e.g., better to have common handlers, etc.), but 
I take no position. I do say that in presence of deployed OAM encapsulations, 
there’s no basis for your your comment about defining a protocol type for 
“Performance Monitoring” as a protocol...


> 
> >       • How to interpret the situation when the O-bit value is 1 but the 
> > value of the Next Protocol field is, for example, NSH, i.e.., not any of 
> > OAM protocols?
> 
> I expect it means there’s OAM within that NSH (sometimes there’s no such 
> things as “OAM Protocols”), or an unhandled OAM protocol.
> GIM>> Should that, i.e., OAM in NSH or immediately following NSH be signaled 
> using SFC NSH methods that are already defined in 
> draft-ietf-sfc-multi-layer-oam?


Non sequitur — I do not follow what signaling has to do with the conversation, 
nor how that draft might be relevant.

> Using the server layer, as you've suggested, seems as layer violation. 
> 

This is a funny statement — NVO3 is a layer violation :-) So are Pseudowires. 
Please do not take this statement too seriously, it is not inviting discussion.

> >       • I believe that the use of the dedicated OAM flag and the Next 
> > Protocol field for a fixed-size header that cannot include meta-data is 
> > unwarranted and adds unnecessary complexity.
> 
> Disagree. 
> 
> See example where protocol is IP carrying an OAM packet.
> GIM>> When IP/UDP encapsulation is used for OAM, there's no, in my opinion, 
> any need to use the O-bit. The destination IP address must be as described in 
> RFC 8029 or RFC 5884 and the demultiplexing is done based on the destination 
> UDP port number. Clearly, O-bit is unnecessary if IP/UDP encapsulation for 
> OAM being used.
> 

Fortunately, bit usage is a matter of definition and not opinion :-)

Kind regards,

Carlos.


> > I suggest not to use O-bit in the VXLAN-GPE header and release it into the 
> > Reserved field.  I don't see the apparent benefit of using the flag, as the 
> > VXLA-GPE uses the fixed size header and the header cannot carry OAM data in 
> > it. The only role the VXLAN-GPE header must perform, in my opinion, is to 
> > unambiguously identify the payload type that immediately follows the header 
> > as OAM (demultiplexing OAM protocols may be done in OAM Header shim).
> > Much appreciate your consideration.
> > 
> 
> Best,
> 
> —
> Carlos Pignataro, [email protected]
> 
> “Sometimes I use big words that I do not fully understand, to make myself 
> sound more photosynthesis."
> 
> > Regards,
> > Greg
> > _______________________________________________
> > nvo3 mailing list
> > [email protected]
> > https://www.ietf.org/mailman/listinfo/nvo3
> 

_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to