Re: [GROW] Genart telechat review of draft-ietf-grow-bgp-reject-07

[Job Snijders <[e-mail]>]

Dear Dale,

Thank you for your time reviewing this document.

On Fri, May 19, 2017 at 11:19:34AM -0700, Dale Worley wrote:
> Summary:  This draft is basically ready for publication, but has nits
> that should be fixed before publication.
> 
>    1. Introduction
> 			
>    BGP routing security issues need to be addressed in order to make
>    the Internet more stable.  Route leaks [RFC7908] are part of the
>    problem, but software defects or operator misconfiguration can
>    contribute too.  This document updates [RFC4271] in order to
>    improve the default level of Internet routing security.
> 
> This paragraph is a good introduction, but it isn't very cohesive.  I
> suggest revising the third sentence to something like:
> 
>    This document updates [RFC4271] so that routes are neither imported
>    nor exported unless specifically enabled by configuration, thus
>    reducing route leaks, and so improving the default level of
>    Internet routing security.
> 
> Then again, considering section 5 paragraph 1, perhaps this update
> reduces all three causes (route leaks, software defects, and operator
> misconfigurations), in which case you could say
> 
>    This document updates [RFC4271] so that routes are neither imported
>    nor exported unless specifically enabled by configuration, thus
>    reducing the consequences of these problems, and so improving the
>    default level of Internet routing security.

I've accepted the latter suggestion.

> --
> 
>    BGP speakers following this specification do not use or send routes
>    on EBGP sessions, unless configured to do otherwise.
> 
> This sentence seems to be correct as written, but somehow it reads
> awkwardly to me.  I think the problem is that "do otherwise" is used,
> when the "otherwise" is "do not use ...".  I think it would read more
> smoothly to say:
> 
>    BGP speakers following this specification do not use or send routes
>    on EBGP sessions, unless specifically configured to do so.
> 
> Perhaps the Editor should be consulted about this.

I've accepted this suggestion.

> --
> 
>    2.  Terminology
> 
>    [RFC4271] describes a Policy Information Base (PIB) which contains
>    local policies that can be applied to the information in the
>    Routing Information Base (RIB).  This document distinguishes the
>    type of policy based on its application.
> 
> Here, you want to say "the type of a policy" or "the type of a
> particular policy", because "policy" refers to a specific policy
> within a set of policies, rather than being a mass noun.

accepted.

> 
>    3. Changes to RFC4271	
> 				
>    This section describes the Updates to [RFC4271] that define the
>    default behavior of a BGP speaker when there are no Import or
>    Export Policies associated with a particular EBGP session.
> 
> Of course, there is no need to capitalize "Updates".
> 
> The wording "describes the updates" is awkward.  Really, it lists or
> specifies the updates, rather than describing them.  Also, the use of
> "updates ... that ..." suggests that there is a larger set of updates
> from which "the updates that define the default behavior" are
> selected, and that smaller set will be described in this section.
> Instead, there are updates, and those updates define the default
> behavior.
> 
> So I think a better wording is:
> 
>    This section updates [RFC4271] to change the default behavior ...".
> 
> It's probably worth consulting the Edtior to see if a better wording
> is possible.

I suggest: "This section updates <xref target="RFC4271" /> to specify
the default behavior of a BGP speaker when there are no Import or Export
Policies associated with a particular EBGP session."

> --
> 
>    The following paragraph is added to Section 9.1 (Decision Process)
>    after the fifth paragraph ending in "route aggregation and route
>    information reduction":
> 
> Strictly, this says that there are five paragraphs which end in "route
> aggregation and route information reduction", and the fifth of them is
> being discussed.  The correct wording is 'the fifth paragraph, which
> ends in "route aggregation and route information reduction"'.

accepted.

>    The following paragraph is added to Section 9.1.3 (Phase 3: Route
>    Dissemination) after the third paragraph ending in "by means of an
>    UPDATE message (see 9.2).":
> 
> Similarly, this should be 'the third paragraph, which ends in "by
> means of an UPDATE message (see 9.2)."'

accepted.

>    5. Security Considerations	
> 
>    Permissive default routing policies can result in inadvertent
>    effects such as route leaks [RFC7908], in general resulting in
>    rerouting of traffic through an unexpected path.
> 
> The word "rerouting" emphasizes that the traffic's route has been
> changed, whereas I think the problem you are concerned with is simply
> that the traffic is routed through an unexpected path.  That suggests
> changing "rerouting" to "routing".
> 
> Then again, perhaps routing people always conceptualize an incorrect
> route as a change from an expected or correct route, in which case
> "rerouting" is the best word to use.

"routing" is perfectly fine too in this context since we are explicit
about "unexpected paths". Accepted.

>    Appendix A.  Transition Considerations
> 
>    It is anticipated that transitioning to a compliant BGP
>    implementation will require a process thay may take several years.
> 
> You probably want to s/a compliant BGP implementation/compliant BGP
> implementations/, unless you are describing the process for an
> individual operator, not for all operators collectively.

The process refers to the vendors of BGP implementations, not operators.
Given that the appendix is targetted mostly to the vendor audience,
would you have a suggestion within that context?

>    A.1. N+1 N+2 Release Strategy	
> 
>    An implementer could leverage an approach described as "the N+1 and
>    N+2 release strategy".
> 
> I prefer reducing the words within the quotation marks.  (But probably
> it's best to ask the Editor.)  That would give:
> 
>    An implementer could leverage an approach described as the "N+1 and
>    N+2" release strategy.
> 
> The section title is difficult to understand without context.  I
> suggest revising it in a way that parallels the way you use the term
> in the text of the section, such as:
> 
>    A.1. Using an "N+1 and N+2" Release Strategy	
> 
> This conveys the maximum possible amount of information to a reader
> (like me) who doesn't know what the "N+1 and N+2" strategy is, namely
> that there is a known release strategy with the name "N+1 and N+2",
> and that the section describes how to use it in this context (and
> hopefully, defines it as well). -- (All of which expectations are
> met.)

I moved the quotation marks around according to your suggestions.

>    "ebgp insecure-mode"
> 
> I think that in this phrase, "ebgp insecure" modifies "mode", and so
> it would be written "ebgp-insecure mode".  (As opposed to "ebgp"
> modifying "insecure mode", in which case it would indeed be written
> "ebgp insecure-mode".)

The current phrasing aligns with industry standards (for better of for
worse), one should consider the suggestion arbitrary, it is merely a
configuration keyword suggestion and not normative. 

I've attached a -07 -> -08 htmldiff which incorporates your suggestions.
After we covered the 's/a compliant BGP implementation/compliant BGP
implementations/' subtopic, and my co-authors reviewed the proposed
changes, we can proceed to cut & publish a -08.

Kind regards,

Job