Recommendations for RSVP-TE and Segment Routing (SR) Label Switched Path (LSP) Coexistence
RFC 8426

Note: This ballot was opened for revision 02 and is now closed.

Deborah Brungard Yes

Ignas Bagdonas No Objection

(Ben Campbell) No Objection

Comment (2018-05-09 for -03)
No email
send info
Most of my comments have already been made by others. I have only a couple of minor comments remaining:

§1: 
-It seems kind of weird to mix "requirements" and "assumptions" in the same list.
- Item 4: The "MAY" seems like a statement of fact, not permission.

Alissa Cooper No Objection

Benjamin Kaduk No Objection

Comment (2018-05-09 for -03)
No email
send info
TED/its expansion should be introduced in the Introduction, not the
Abstract, since the Abstract and the rest of the document must be
able to be read in a standalone manner.


In line with Alvaro's comment, I would suggest moving Sections 3.1 
through 3.4 into a new parent section "Discarded Options" or
similarly-named.


Section 3.5

   [...] It is RECOMMENDED that the IGP-TE update
   threshold SHOULD be lower in order to flood unreserved bandwidth
   updates often.

Lower than what?


Section 7

It feels odd to say that these "do not require any new security
considerations" and then go and list some new security
considerations.  I would suggest something like "The security
considerations of each protocol are unaffected by the presence of
the other, so only the interactions of the TED consistency solution
with the individual protocols needs to be considered."

I would probably also want to expand the text a bit, noting that:

A hijacked SR traffic stream could potentially cause disruption to
RSVP-TE LSPs in two ways: directly, but consuming sufficient
bandwidth to cause traffic loss, and indirectly, by consuming
sufficient traffic to result in the TED consistency solution
proposed in this document reducing the bandwidth available to
RSVP-TE paths.  The former is possible whenever RSVP-TE and SR
traffic share links, independently of whether this specification is
in use; the latter is new behavior with this specifciation but is
seen as preferrable to the alternative, since the impact to RSVP-TE
traffic can be controlled and paths re-routed.  However, some latent
risk of disruption remains because this specification is a reactive
protcol, relying on taking measurements and only adopting to new
traffic flows after the measurement period.  The finite duration of
the measurement window leaves open a period of potential disruption
before remediation can be applied.


Appendix A

This guidance seems like it could be (very) roughly characterized as
"these are the ranges of values that are not insane to use".  Is it
possible to give more precise/active guidance, such as (taking a
wild guess) "Values between 0.9 and 1.1 allow for taking accunt of
estimated future traffic growth during the current measurement
period while reducing the risk of leaving large amounts of bandwidth
underutilized.  The measurement period should be smaller than a
minute in order for this tight of a growth window to make sense."

Suresh Krishnan No Objection

Warren Kumari No Objection

Comment (2018-05-06 for -03)
No email
send info
If you haven’t already, please address Al Morton’s excellent OpsDir review.

Mirja Kühlewind No Objection

(Terry Manderson) No Objection

Alvaro Retana No Objection

Comment (2018-05-08 for -03)
No email
send info
(1) The Abstract says that this document provides "solution options" -- it would be very nice if a recommendation (or at least general guidance) was provided by the authors.  This point was brought up in the OpsDir review [1], to which the authors replied: 

   "The intent behind the recommendations is to not take a position on which solution is preferable. All solutions are valid but some do not satisfy all the requirements. However, if a solution is acceptable (based on their deployment of SR and RSVP and knowing which requirements are not satisfied) for an operator then such a solution can be chosen. "

That seems like a reasonable answer.  It would be good if, at least, text to the effect was included in the document.

(2) Given, from the text above, that not all requirements may be as important in a specific deployment, I find the use of rfc2119 language to describe them (in the Introduction) confusing and inappropriate.  Please consider not using Normative text in that section.

(3) While wondering whether there was a recommendation coming out of this document, I looked for other documents referencing it, and found 1: draft-ietf-mpls-rsvp-shared-labels ("Signaling RSVP-TE tunnels on a shared MPLS forwarding plane").  I realize that the topic is not exactly a match, but this sentence is included in it: "The RSVP-TE tunnels that use this shared forwarding plane can co-exist with MPLS-SR LSPs [I-D.ietf-spring-segment-routing-mpls] as described in [I-D.ietf-teas-sr-rsvp-coexistence-rec]."

I took a very quick look at draft-ietf-mpls-rsvp-shared-labels and the mechanism described there didn't look like any of the options described here.  Is the mechanism described in draft-ietf-mpls-rsvp-shared-labels an option to the problem addressed in this document?  If so, is it worth including a short description?  If not, then the sentence above sounds misleading.

[I realize I may be asking questions about a different document...but I'm taking the license given that the editor for this document is also an author in the other one. ;-) ]

(4) Nits:

(4.1) This text in the Abstract seems redundant: "In some instances, operators are also migrating existing services from RSVP-TE to SR LSPs...In other cases, services running on RSVP-TE might be migrated to run over SR."

(4.2) From §3.2:
   Note that it is not enough for the controller to just maintain the
   unified view of the available capacity, it must also perform the path
   computation for the RSVP-TE LSPs, as the reservations for the SR LSPs
   are not reflected in the TED.  This does not fit with assumption 2
   mentioned earlier.

Assumption 2 says that "Engines that compute RSVP-TE paths may have no knowledge of the existence of the SR paths in the same domain.", but in the scenario described here (where the controller "must also perform the path computation for the RSVP-TE LSPs"), then the assumption is not true as the controller would in fact have knowledge of the coexistence.  This is just a nit: I don't think the last sentence is accurate...

(4.3) This text is normatively redundant: "It is RECOMMENDED that the IGP-TE update threshold SHOULD be lower..."  RECOMMENDED and SHOULD mean the same thing.


[1] https://mailarchive.ietf.org/arch/msg/ietf/KwkK9VbeISPl9zgVmzdwLOB2FM0

Adam Roach No Objection

Martin Vigoureux Recuse

Comment (2018-05-07 for -03)
No email
send info
I have contributed to the document