Re: [RTG-DIR] Rtgdir early review of draft-ietf-idr-segment-routing-te-policy-18

[Ketan Talaulikar <[e-mail]>]

Hi Mohamed,

Thanks for your very detailed review and helpful suggestions. Please check
inline below for responses.

We will post the update once the submission tool reopens.


On Fri, Jul 8, 2022 at 6:35 PM Mohamed Boucadair via Datatracker <
[e-mail]> wrote:

> Reviewer: Mohamed Boucadair
> Review result: Has Issues
>
> Document: draft-ietf-idr-segment-routing-te-policy-18
> Reviewer: Mohamed Boucadair
> Review Date: 08/07/2022
> IETF LC End Date: N/A
> Intended Status: Standards Track
>
> I appreciate the effort that was spent to progress this draft since more
> than 6
> years!
>
> Before reviewing the document, I started first by re-reading
> RFC8024/RFC9012
> and reading draft-ietf-spring-segment-routing-policy for establishing the
> context. Overall, the approach documented in
> draft-ietf-idr-segment-routing-te-policy is sound and straightforward.
>
> I didn’t find major concerns from a routing standpoint other than the need
> to
> motivate some few claims (see the detailed review file about RRs, for
> example)
> and the lack of considerations related to the handling of the various
> sub-TLVs
> by intermediate routers (if any).
>
> However, there are a number of generic issues that I would recommend to
> consider (see the detailed review for the full list). All these are
> easy-to-fix
> issues.
>
> # General Comments (in no specific order)
>
> ## Consistency
>
> ### Single or multiple paths
>
> There is an apparent inconsistency in the document about the handling of
> multiple paths. For example, Section 1 says :"Selection of the best
> candidate
> path for an SR Policy" while the same section says also “this will result
> in
> one or more candidate paths being installed into ..”.
>

KT> The first is about the selection of the best candidate path for an SR
Policy by the SRPM - this is what gets installed in the forwarding. The
second is about the installation of the received candidate paths into the
BGP table. There is no inconsistency.


>
> If multipath is supported, then please add an explicit statement and make
> sure
> the overall text is consistent.
>

KT> Only a single CP is selected for a given SR Policy. This is per the
draft-ietf-spring-segment-routing-policy and this document does not change
that.


>
> ### Value 0 is marked as reserved for some registries, while that value is
> associated with a meaning for other registries.
>
> Is there any reason why a consistent approach isn’t followed here? what is
> the
> issues if value 0 is open for assignment?
>

KT> It is normal routing protocol practice to not assign the TLV 0 values.
Can you indicate where the TLV code point 0 is being assigned?


>
> ## Modifications to the format of the Color Extended Community
>
> The text says that you are modifying the format the Color Extended
> Community,
> while this is not true. What this draft does is just associating a meaning
> with
> some bits. I would update the text accordingly.
>

KT> We are changing the format of only the Flags field and not of the
entire EC. Flags are normally independent bits and here we are combining
two bits to convey 4 values. Clarified this in the Introduction section.


>
> ## Normative language
>
> The use of the normative language should be double-checked. The most
> apparent
> concern is related the statement related to the handling of the reserved
> bits
> (SHOULD) while this RFC9012 uses MUST (which is correct, IMO).
>

KT> Ack. I will fix it and change it to MUST.


>
> I tagged many others in the full review, fwiw.
>
> ## Lack of description
>
> Many fields are provided without acceptable description (e.g., “Local IPv4
> Address: a 4-octet IPv4 address.” or “Preference: a 4-octet value” !!).
>

KT> These fields are in the context of a sub-TLV. There is text in the
description of that sub-TLV that provides a reference (e.g., to the
draft-ietf-spring-segment-routing-policy section or a segment type, etc.)
There is no need to repeat a detailed description for each field IMO.


>
> Also, some fields are provided with a structure but the text says also that
> these are reserved (e.g., 2.4.2 says “TC, S and TTL (Total of 12 bits) are
> RESERVED”).
>

KT> This is the MPLS label field. I am not sure that I follow your concern
here.


>
> I wonder whether you can add a statement to say that multiple flags can be
> set
> simultaneously unless this is precluded by future flag assignments.
>

KT> Not sure that is necessary. In most cases, the bits/flags are
independent. Where they are not, there is generally text explaining their
relationship or dependency.


>
> Last, the document does not include the expected behavior of intermediate
> routers (e.g., whether it is allowed or not to alter some fields). I guess,
> they must not touch the content of the attributes but it is better if this
> is
> explicitly mentioned in the text.
>

KT> Yes, the contents must not be altered. Will clarify in sec 4.2.4.


>
> ## Reserved vs. Unassigned
>
> Almost all the “reserved” bits in the spec can be assigned in the future. I
> would use “Unassigned” as per RFC8126.
>

KT> Ack. Will change in a few places where this has been missed.


>
> FWIW, 8126 says the following:
>
>       Unassigned:  Not currently assigned, and available for assignment
>             via documented procedures.
>
>       Reserved:  Not assigned and not available for assignment.
>             Reserved values are held for special uses, such as to extend
>             the namespace when it becomes exhausted.
>
> ## Deprecated values
>
> The document includes notes about some “deprecated” codepoints. I’m not
> sure
> there is a value in having such notes in the final RFC.
>

KT> Yes, there is a need. One is to avoid them being used for any other
sub-TLV in the future. Two is that there are early implementations out
there that have some degree of support - even if they are just doing some
parsing/showing.


>
> ## IANA considerations
>
> ### The document uses a mix of TBD statements (e.g., Section 2.4.3) and
> hard-coded values (early assignments). Not sure what’s was the rationale
> especially that code 20 was assigned but not listed as such.
>

KT> Fixed.


>
> ### The IANA actions should be more explicit and ask IANA to update
> existing
> entries. For example, the current registry for code 73 points to
> [draft-previdi-idr-segment-routing-te-policy]. Need to update that entry
> and
> similar ones.
>

KT> Have fixed the text. IANA will update "This document" to the RFC number
before publication. There is no need to keep changing the draft name
through its lifecycle.


>
> ### The document lists (under IANA section) some values that are
> deprecated.
> The document should be clear whether these codes are available for future
> assignment or not.
>

KT> Deprecated means they are not available for assignment by IANA unless
the IETF changes that via an RFC.


>
> ### Many sub-TLVs have flag bits but not all of them have a registry to
> track
> future flag bit assignments.
>

KT> The registries would be added by future documents that start using
those flags.


>
> ## Manageability considerations
>
> No such considerations are included in the document.
>

KT> Will add.


>
> # Detailed review
>
> FWIW, you can find my full review at:
>
> * pdf:
>
> https://github.com/boucadair/IETF-Drafts-Reviews/raw/master/draft-ietf-idr-segment-routing-te-policy-18-rev%20Med.pdf
> * doc:
>
> https://github.com/boucadair/IETF-Drafts-Reviews/raw/master/draft-ietf-idr-segment-routing-te-policy-18-rev%20Med.doc
>
>
>
KT> This was helpful and have incorporated most of those suggestions.

Thanks,
Ketan