Operational Management of Loop-Free Alternates
draft-ietf-rtgwg-lfa-manageability-11

1. The abstract mentions a lot of things that this document covers: “provides operational feedback on LFA, highlights some limitations, and proposes a set of refinements to address those limitations.  It also proposes required management specifications.”  The Introduction presents a quick guide to what some sections cover (good idea!), but not all sections are mentioned there.

It would be nice to cover all the sections in the “document map” in the introduction.


2. In section 3.1 it may not be clear to all readers why P4 is not an LFA for P8.  In all the other cases there is an explicit statement (a sentence or two) that explains and clarifies.


3. In Section 6.2.4.2 the document talks about signaling color information, it includes a set of requirements..and it reads “How signaling is done is out of scope of the document”, but then you go on and point to a specific solution.  Even if there might be a high certainty that the solution you point at is moving on in the process, is good, should be used, etc..  I think this document would be better served by just defining the requirements (specially if you’re pointing at the solution as out of scope).   You do the same in 6.2.4.4.


4. The IS-IS overload bit is mentioned in several places as important to consider.  Are there similar considerations related to the use of the OSPF MaxLinkMetric or the R-bit?  If so, please include them..if not, please explain why in the document.  [BTW, there is no reference pointing to the OL bit.]

I only have a few editorial things to add:

-- Section 4 --

   Implementers SHOULD document their LFA selection algorithms (default
   and tuning options) in order to leave possibility for 3rd party
   modules to model these policy-LFA expressions.

I'm not going to whine about this *too* much, but I don't see this as an appropriate use of 2119 key words.  I'd be happier if it were changed to not be a 2119 word.  That said, there's no need for discussion.  Do as you think best.

-- Section 6.2 --

   3.  If the defined policy does not permit to determine a unique best
       LFA,

The wording here is really awkward English.  The verb "permit" needs a direct object here, as in "does not permit <someone> to determine <something>."  You can't just omit the <someone>.  The easiest fix is to change "to determine" to "the determination of".  (I would also use "allow" rather than "permit", but that's just a personal wording preference.)

-- Section 6.2.4.1 --

   When SRLG protection is computed, and implementation SHOULD permit to

Same problem here as in 6.2, above, but because of the list the fix isn't as easy.  First, "and" should be "an".  Second, the fix should probably be like this:

NEW
   When SRLG protection is computed, an implementation SHOULD permit the
   following:

   o  Exclusion of alternates violating SRLG.

   o  Maintenance of a preference system between alternates based on
      SRLG violations. [...etc...]
END

In the first sub-bullet, "Preference based on number of violation," both instances of "violation" should be "violations".

-- Section 7.3 --

   o  MUST be able to display, for every prefixes, the primary next hop
      as well as the alternate next hop information.

"prefixes" should be "prefix".

-- Section 7.4 --
When you say "provide an alert system" here, I think you mean "provide an alert", or, perhaps better, "trigger an alert" or "generate an alert".  Presumably, in order to do that, an alert system has to be available.  But the point here is that you saying when specific alerts should be triggered/generated.

I've got nothing that rises to the level of a discuss, but I do have a few comments:

*** Substantive Comments: ***

-- section 4, last paragraph:

This is an odd thing to make normative. It seems more a question of business and ecosystem decisions.

-- 6.2.4.2:

Can you explain the "color" metaphor, or reference an explanation?

-- 8:

The entirety of the security considerations are of the form of "No new considerations compared to [RFC5286]". Please offer supporting arguments for that. For example, a high-level description of the nature of the changes, new behaviors, or clarifications introduced in this draft, and how those do or do not impact the security considerations.


*** Editorial Comments: ***

-- General:

There are pervasive grammatical errors, especially incorrect use of singular and plural forms, missing articles, and incorrect use of commas. The RFC editor will fix these, but you could save them quite a bit of work by making another proofreading pass.

Also, throughout the draft, there are lists of examples in the form of ":A,B,C...". I suggest using the form "A,B,C etc.", or even better "For example , A, B, and C" or "e.g., A, B, and C"

-- section 2, first bullet:

Last sentence is a fragment.

-- 3.1, last paragraph:

This seems to say that, in addition to the technical issues mentioned here, service providers simply might not like it. Is that really what you mean to say?

-- 4, last bullet:

Missing "to" . Otherwise, this renders to "... may be able monitor constantly..."

-- 5, first paragraph: "As all FRR mechanism, ..."

I think there's one or more missing words. Do you mean "As [in/for] all other FRR mechanisms, ..."?

"Depending of the hardware..."

s/of/on

'... compared to the amount of destinations in RIB."

I suggest " ... compared to the number of destinations in the RIB."

-- 2nd paragraph " ... permit to compute ... "

"Permit" needs a direct object. That is "... permits [something] to compute...". What is the something?

(Note: This construction appears several times)

-- 6.1, first paragraph: "The granularity of LFA activation should be controlled ..."

-- 6.1, last bullet in first list: "... SHOULD have a better priority ... "
s/better/higher

-- 6.2, item 3 in the numbered list: "... an implementation SHOULD pick only one based on its own
       decision, as a default behavior."

A "default behavior" implies that people might make non-default choices. I suggest striking the phrase", as a default behavior".



I'm not sure what that means.

-- 2nd paragraph: " ... following criteria:"

I suggest s/criteria/granularities

(Note: I see quite a bit of the use of the word "criteria" when I think you mean something else. )

-- 6.2.3:

What does "enhanced criteria" mean? Also the word "Downstreamness" seems a bit of a stretch, but if the RFC editor lets that get by then more power to you :-)

-- 6.2.4.1:
Please expand SRLG on first mention.

-- 6.2.5.4: 5th paragraph "... it is needed to limit..."

perhaps "... implementations need to limit..."

-- 9 and 10:

Section 9 is out of place (should probably go right before references), and 10 is empty.

I agree with Avaro's Comments #2 and #3.

Similar to others, I find the readability/English poor. I think it will be too much for the RFC Editor and the RFC Editor will not catch some of these errors e.g.:
- 6.2.4.1 SRLG "preference system"
- 6.2.4.2 "Link coloring is a powerful system"
I would say (similar to Alvaro's comment on "alert system"), it's not a "system". A better wording would be "technique" or "method". Agree with others, a proof reading pass is really needed.

Alexey Melnikov's Gen-ART review comments were worth considering.

Ron Bonica's Opsdir review.

Folks,

I have reviewed this document as part of the Operational directorate's ongoing effort to review all IETF documents being processed by the IESG.  These comments were written with the intent of improving the operational aspects of the IETF drafts. Comments that are not addressed in last call may be included in AD reviews during the IESG review.  Document editors and WG chairs should treat these comments just like any other last call comments.

This document is on the Standards Track. It provides operational feedback on LFA, highlights some limitations, and proposes a set of refinements to address those limitations.  It also proposes required management specifications.

The document is well-written and nearly ready for publication.

Major Issues
----------------
None

Minor Issues
---------------
- Please run this document through the NIT checker and address the NITS

- I am not sure how the sitting IESG feels about the use of lowercase "must", "should" and "may". You may want to check this before the IESG review.


Ron Bonica

---

example that I would cite as good to all caps

6.1
...


   o  Per prefixes: prefix protection SHOULD have a better priority
      compared to interface protection.  This means that if a specific
      prefix must be protected due to a configuration request, LFA must
      be computed and installed for this prefix even if the primary
      outgoing interface is not configured for protection.

LFA MUST

since it's a requirement

in most other cases I see a lower cast must what is being described is the logic that draws you to a conclusion, and those are ok.