Dear Les,
I see that you "agreed" to correct two typos only.
I disagree with your reasons for not making the document more readable and
complete. Note that "more complete" doesn't necessarily mean "more
prescriptive". It means removing ambiguities that will continue hurting
interoperability of new and existing implementations.
Since your guiding principle is "don't rock the boat / don't describe what to
do - only what NOT to do", I see no point in arguing point-by-point.
I know that different Areas have different standards for their documents, so I
leave it to your community.
Thanks,
Orit.
-----Original Message-----
From: Les Ginsberg (ginsberg) [mailto:ginsberg(_at_)cisco(_dot_)com]
Sent: Thursday, April 13, 2017 8:06 PM
To: Orit Levin <oritl(_at_)microsoft(_dot_)com>; gen-art(_at_)ietf(_dot_)org
Cc: ietf(_at_)ietf(_dot_)org;
draft-ietf-isis-mi-bis(_dot_)all(_at_)tools(_dot_)ietf(_dot_)org
Subject: RE: Gen-ART Last Call review of draft-ietf-isis-mi-bis-02
Orit -
It is obvious you spent a lot of time on this review - and I do want to be
respectful of that.
However, there is a larger context here which I think has a significant bearing
on handling of many of your comments.
RFC 6822 was published over 4 years ago. Multiple interoperable implementations
exist. The bis version makes some modest - but significant changes. However, we
deliberately strived to keep the bis version as consistent as possible with RFC
6822 in order to minimize the possibility that aspects of the specification
which had NOT been changed would be reinterpreted simply because the wording
had changed. So, in cases where you suggest (below) that a different wording is
desirable I am very reluctant to make such changes because of the above concern.
If I do not indicate any response to a particular comment you can interpret as
meaning:
"Unnecessary changes relative to RFC 6822 are not desirable."
Inline.
-----Original Message-----
From: Orit Levin [mailto:oritl(_at_)microsoft(_dot_)com]
Sent: Thursday, April 13, 2017 4:59 PM
To: Les Ginsberg (ginsberg); gen-art(_at_)ietf(_dot_)org
Cc: ietf(_at_)ietf(_dot_)org;
draft-ietf-isis-mi-bis(_dot_)all(_at_)tools(_dot_)ietf(_dot_)org
Subject: RE: Gen-ART Last Call review of draft-ietf-isis-mi-bis-02
Hi Les,
Sorry for the delay in response.
Your feedback was very helpful. Below is a refresh of my comments. I
tried to make them more pointed and some are new.
Summary: This draft is "ready with issues" for publication.
General:
1) For implementers who are familiar with the history and the intent
of this extension, the information in the draft is probably sufficient
to serve as a check list for implementing a multi-instance IS-IS
router. For all other readers, the document doesn't contain an
overview of the new mode of operation, i.e. where the instances are
not a configuration and an internal implementation choice only, but
are exposed through the protocol to achieve the stated objective.
Lacking such an overview, the reader needs to reverse-engineer the logic
behind the documented guidance.
[Les:] I am not sure I understand your concern. You seem to be suggesting that
readers won't know when to use MI and when to simply create multiple non-MI
instances of the protocol.
The short answer to that is that it is not the purpose of this document to make
that decision. We have provided some new functionality - it is up to the user
to decide when it is appropriate to use the new functionality and when it is
not. We have provided some guidance in Sections 3 and 4 - but this is
non-normative - as it should be.
But your use of "reverse-engineer" confuses me, so likely I do not understand
your point.
As regards new vs old readers, the new version of the document provides as much
(or as little) guidance as RFC 6822 - so I do not see why new readers would
have any more issues than new readers of RFC 6822 had 4 years ago.
2) The draft talks about "extensions" in plural. Based on a single
extension on the wire and the overall goal of the new mechanism, I
would say that it is a single extension only. How many protocol
extensions does this document define? If they can be clearly
separated, then it needs to be clarified throughout the document.
Otherwise, the language throughout the document needs to be changed from
"extensions" to "the extention".
[Les:] There are multiple changes in protocol behavior described - hence the
term extensions is correct.
3) Editorial: Please, compare (Diff) the current draft with the
published RFC 6822. You will find that various RFC Editor corrections
got lost in this bis document. Some repeating examples of the lost
corrections are "instance- specific", " topology (or topologies)" and
"Type-Length-Value".
[Les:] Noted - thanx.
My comments below are a result of a reverse-engineering exercise.
Please, consider incorporating the suggested clarifications to improve
the document readability. I might have misunderstood some of the
parts; in such cases, please, provide an alternative text.
Abstract
1) Add clarification: "This document is not backwards compatible with
RFC 6822."
[Les:] This statement is made explicitly in the Appendix - and I believe that
is where it belongs for reasons I have previously stated.
2) Par. 2, replace the first two sentences with: "Configuration of
multiple protocol instances within a router allow the isolation of
resources associated with each instance. This document introduces a
new mode of operation where the protocol instances are not a matter of
configuration only, but are exposed through the new protocol extension
to achieve the objective stated above."
3) Par. 3 uses both present and future time. Does it mean before and
after the extension? Please, clarify by spelling out the intent and/or
changing the grammar.
[Les:] The sentence will in any case be changed when the document becomes an
RFC - not worth discussing.
1. Introduction
1) Move par. 3 to become the opening paragraph (i.e., par.1 ) in the
Introduction to improve its logical flow (also related to the next comment).
2) Par 4, sentence #2 says "This document defines an extension to
IS-IS to allow non-zero instances...". My assumption is that the
intent of this draft is to define a single extension, which will
improve the routing operations in a number of ways. If this is the
case, the quote needs to be replaced with "The MI-IS-IS extension,
defined in this document, also allows so-called "non-zero instances"..." .
3) Par 4, sentences #4 and #5 are not sufficient to describe the new
mode of operation introduced by this extension. Below is my attempt to
describe it.
Please, correct, if I got it wrong.
"IS-IS router instances that support this extension are preconfigured
with unique non-zero Instance Identifiers (IIDs) giving them the name
"non-zero instances". In addition, MI RTRs MAY (or SHOULD ?) implement
the legacy (or so-called "standard") instance of the IS-IS router for
backwards compatibility with legacy IS-IS routers. IID #0 is only used
by MI-RTRs to convey information associated with such standard
interface if implemented. See section 2.6 Interoperability Considerations for
more details."
4) Par 4, replace the two "may" with "can" to clarify the intent.
5) Par 5, change "defined" to "described" since the examples are not
normative.
5) Par 5, add references to the (sub)sections containing the
description of the two methods.
6) Par 7, move the last paragraph before listing the two examples and
adjust the text accordingly, to improve the logical flow.
7) In the end of the Introduction add a reference: "This RFC is not
backwards compatible with RFC 6822. Differences between this document
and RFC 6822 are described in Annex A." ... so that others don't skip
it by mistake as I did earlier...
[Les:] Reading of the Appendix is optional - the new version is usable without
any need to refer to RFC 6822 - and that is its primary purpose.
2. Elements of Procedure
1) It seems that this section (informational) and its subsections
(normative) use present and future times and, at times, the normative
language inconsistently both within the sections and among the
sub-sections. Please, explain the reasons and improve consistency accordingly.
2) Par 1, clarify the scope by adding "within a routing domain" at the
end of the first sentence.
3) Par 1, after the first sentence add a new sentence alone the
following lines "Routers form adjacencies and exchange routing updates
only if their IIDs correspond." This explains the basic premise of the whole
mechanism.
4) Par 2, change "may" to "can" or "might" for consistency.
Section 2.1
1) Par 2, change "may" to "MAY".
2) Par 3, remove "supported by legacy systems" from the first sentence
to avoid confusion. My understanding is that IID #0 is reserved for
use by MI- RTRs that also implement the standard instance and
advertise it in IIH using IID-TLV.
3) Par 3, change "except where noted" to "except as noted in section
2.6.2 (?)". This is an excellent place to explain the logic behind
this MUST NOT statement or, at least, state the general circumstances
where IID #0 is included in IID-TLV.
4) Par 9 (4th after the picture), change "as described later" to " as
described in section 2.6.2 (?)".
5) Par 13, change "recommended" to "RECOMMENDED".
6) Editorial: Par 13, change "particularf" to "particular".
[Les:] #6 - Agreed.
Section 2.2
1) Add that MI-RTR MAY (or SHOULD) implement the standard instance as
well and which packets are used to advertise it.
2) Rephrase "marks ... by including" to "MUST include" to use
requirements language.
Section 2.3
Editorial: Replace "normal" with "usual".
Section 2.4.1
1) Par 1. Replace "IID #0" with "standard instance".
2) Par 1. Replace "instances other than IID #0" with "non-zero instances".
3) Par 2 second sentence. What does it mean "normal expectations"? Is
this a network configuration requirement? Please, clarify in the text.
Section 2.4.2 Improve language consistency
1) Verbs are used inconsistently: some are used in present time,
others in future time.
Section 2.5, replace "exists" with "MUST be performed".
Section 2.5.1, replace "only operates" with "MUST only be performed".
Section 2.5.2, replace "This requires" with "It is REQUIRED".
Section 2.5.2 third sentence, after "inconsistent" insert "due to
their configuration". (Please, correct me if I am wrong.)
Section 2.6.1
1) Editorial: Par 1, first sentence, replace "not to cause" to "to avoid".
2) Par 2, remove "(IID #0)".
3) Par 2, replace "non-zero IID" with "non-zero instance".
4) Par 5 NOTE, replace "IID #0" with "standard instance".
Section 2.6.2
Replace all four appearances of "IID #0" with "standard instance".
Section 3.1
Replace the two "MAY" to "can".
7 Security Considerations
Discuss possible additional security considerations (or the lack of
them) related to the exposure of "instances" on the wire.
Reason: Beyond the normal IETF procedure, this is especially important
because "multiple instances allow isolation of resources..." Can this
isolation, if observed or interfered on the wire, be damaging beyond
the previous "standard instance" situation.
[Les:] There are no security concerns associated with the use of multiple
instances.
Security reviews of the document have been satisfied w the current contents.
Les
Thanks,
Orit.
-----Original Message-----
From: Les Ginsberg (ginsberg) [mailto:ginsberg(_at_)cisco(_dot_)com]
Sent: Saturday, April 8, 2017 4:23 PM
To: Orit Levin <oritl(_at_)microsoft(_dot_)com>; gen-art(_at_)ietf(_dot_)org
Cc: ietf(_at_)ietf(_dot_)org;
draft-ietf-isis-mi-bis(_dot_)all(_at_)tools(_dot_)ietf(_dot_)org
Subject: RE: Gen-ART Last Call review of draft-ietf-isis-mi-bis-02
Orit -
Thanx for the review.
Responses inline.
-----Original Message-----
From: Orit Levin [mailto:oritl(_at_)microsoft(_dot_)com]
Sent: Thursday, April 06, 2017 8:27 PM
To: gen-art(_at_)ietf(_dot_)org
Cc: ietf(_at_)ietf(_dot_)org;
draft-ietf-isis-mi-bis(_dot_)all(_at_)tools(_dot_)ietf(_dot_)org
Subject: Gen-ART Last Call review of draft-ietf-isis-mi-bis-02
I am the assigned Gen-ART reviewer for this draft. The General Area
Review Team (Gen-ART) reviews all IETF documents being processed by
the IESG for the IETF Chair. Please treat these comments just like
any other last call comments.
For more information, please see the FAQ at
<http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
Document: draft-ietf-isis-mi-bis-02
Reviewer: Orit Levin
Review Date: 2017-04-06
IETF LC End Date: 2017-04-07
IESG Telechat date: 2017-04-13
Summary: This draft is "ready with issues" for publication.
Major issues: None.
Minor issues:
1. Add text explaining the reason (or reasons) for replacing the
original RFC
6822 from 2012.
Reason: It is a "bis" draft and there is no mention about it in the text.
[Les:] Note that the latest revision of the draft correctly identifies
the draft as obsoleting RFC 6822. Previous versions had incorrectly
identified this as an update to RFC 6822.
This is then the new Standard for the IS-IS MI support.
There are two classes of future readers of this document:
a)Readers who are unfamiliar with RFC 6822. For them what changed
between RFC 6822 and this document is irrelevant.
b)Readers who are familiar with RFC 6822. For them it is useful to
know what changed - which is described in Appendix A.
In order not to distract readers of type "a" - as well as to provide
an "uninterrupted" description of the normative behavior I believe
placement of the change description in an Appendix improves the
readability of the document.
Does this make sense to you?
2. In Abstract, state clearly that this standard introduces the
support for instances vs. other already existing concepts also
listed in the Abstract (i.e., circuits, adjacencies, topologies, etc.).
[Les:] The Abstract currently says:
"This draft describes a mechanism that allows a single router to share
one or more circuits among multiple Intermediate System To
Intermediate System (IS-IS) routing protocol instances."
Previous to this extension, a router could have multiple instances of
the IS-IS protocol, but multiple instance could not be run over the same
interface.
So we are not introducing "instances", but we are introducing the
ability to enable multiple instances on the same interface.
Reason: The wording is not clear about what is the new feature vs.
what are the new benefits vs. what was the original baseline
3. Throughout the
document, use "standard instance" instead of "IID = 0" or "IID #0".
Reason: Expressions "standard instance", "IID = 0" and "IID #0" are
used interchangeably throughout the document. It seems that they all
refer to the same thing - the implementation of the original
protocol without the concept of instances. Please, correct me if I am wrong.
[Les:] I don't think this is possible without seriously compromising
the document. For example:
Section 2.1
" IID #0 is reserved for the standard instance supported by legacy
systems. "
Changing this to " Standard instance is reserved for the standard instance
..."
Is clearly nonsensical.
Later in Section 2.1
"When the IID = 0, the list of supported ITIDs MUST NOT be present."
What is being discussed here is what is the correct behavior when an
MI- capable router sends a PDU associated IID #0 and includes the new IID TLV.
Replacing this with "When the standard instance..." loses the
important point that the value of the IID in the IID TLV in this case is "0".
Hope this helps clarify things.
4. In section 2 par 3, change "support" and "operates" to "MUST
support" to use requirements language.
[Les:] I am on the fence as regards this change. Section 2 is an
introduction to the following sub-sections - which define the
normative behavior. But the introduction itself is not defining
normative behavior - it is providing a context in which the protocol
extensions defined in the following sub- sections can be understood.
I am more inclined to change the "MAY" used later in the same
paragraph you mention to "may" so it is consistent with the rest of this
section.
???
5. In section 2 par 2, change "may" to either "can" or "MAY" to
clarify the intent.
[Les:] Did you mean Section 2.1 para 2?
If so I agree to the change.
6. In section 2.1 par 3, clarify whether IID #0 is ever being used on the
wire.
[Les:] There are numerous places in the document where the legal use
of IID
#0 is discussed. I do not understand how a reader would conclude that
IID #0 is never sent on the wire.
Explain the concept of the "standard interface" (see previous
comment)
[Les:] There is no mention of "standard interface" - did you mean
"standard instance"?
If so, Section 1 paragraph 4 states:
"Legacy routers support the standard
or zero instance of the protocol."
Les