Re: [mpls] Poll 2: New SPL vs re-purposed ELI for carrying Network Action Indicators

[Greg Mirsky <[e-mail]>]

Hi Robert,
thank you for getting back to me. I have several additional notes tagged by
GIM2>> below.

Regards,
Greg

On Fri, Jun 17, 2022 at 1:02 PM Robert Raszuk <[e-mail]> wrote:

> Hi Greg,
>
> > as the declaration of non-support of the application of IOAM in, for
>> example, MPLS networks?
>>
>
> Nope. It can clearly be supported by MPLS networks. But it does not need
> to be part of the label stack. Subtle but extremely significant difference.
>
GIM2>> Can you point out to me the document that proposes encoding the IOAM
Trace header and Trace Data into the label stack, supporting IOAM in MPLS
using ISD?

>
> GIM>> I am not sure that PSD can be used in the scope
>>> of draft-decraene-mpls-slid-encoded-entropy-label-id. I think I need
>>> Bruno to help me here. I'll ask him in the other discussion thread.
>>>
>>
> Indeed me too. Bruno's draft defines placeholder for slice_id. To me this
> is except the name the same as my RAF proposal. So no ISD nor PSD needed as
> per MNA framework.
>
GIM2>> I've read the opinion that Bruno's draft can support up to eight
MNAs. I assume that some would require some data associated with the
particular MNAI. It seems to me that Bruno suggested that PSD can be used
as part of draft-decraene-mpls-slid-encoded-entropy-label-id. Let's wait
for his clarification.

>
> Slice_ID would be programmed among other things to execute IOAM on a
>>> packet.
>>>
>> GIM>> Would kindly elaborate on how you envision such a mechanism?
>>
>
> What I mean that you carry the IOAM data outside of the label stack. label
> stack is to be kept condense and lean and not take a balloon shape to slow
> each packet down. Imagine aerodynamics analogy to electrons or
> wavelengths :)
>
GIM2>> I believe that there are different options when it comes to
collecting and transporting IOAM Trace data, not only those described in
RFC 9197. IOAM Direct Export and Hybrid Two-step are good examples that
keep the footprint of IOAM in data packets under control, limited to what
is functionally necessary.

>
> Best,
> R.
>