Re: [Gen-art] Gen-ART review of draft-ietf-dhc-dhcpv6-failover-protocol-04

[Tomek Mrugalski <[e-mail]>]

W dniu 01.02.2017 o 21:57, Christer Holmberg pisze:
>
> 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.
>
Hi Christer!
Thanks a lot for your review. See my responses below.
>
>  
>
> For more information, please see the FAQ at
>
>  
>
> <https://trac.ietf.org/trac/gen/wiki/GenArtfaq>.
>
>  
>
> Document:                                     
> draft-ietf-dhc-dhcpv6-failover-protocol-04.txt
>
> Reviewer:                                        Christer Holmberg
>
> Review Date:                                  01.02.2017
>
> IETF LC End Date:                          19.01.2017
>
> IESG Telechat date: (if known)    02.02.2017
>
>  
>
> Summary:                                       The document is almost
> ready for publication, but there are some editorial nits that I’d like
> the authors to address.
>
>  
>
> Major issues:                                 None
>
>  
>
> Minor issues:                                 None
>
>  
>
> Nits/editorial comments:
>
>  
>
> INTRODUCTION:
>
>  
>
> Q1:        In the first sentence of the Introduction, I suggest to say:
>
>  
>
> “The failover protocol defined in this document provides…”
>
Sure, will update as suggested.
>
>  
>
> Otherwise it’s a little unclear what failover protocol you are talking
> about.
>
>  
>
> Q2:        In the Introduction, before the first sentence, shouldn’t
> there be some background text, including some information about the
> problem that the document solves. I know there is something in the
> Abstract, but I think there should also be something in the
> Introduction, before jumping into the solution.
>
There's a whole RFC about it: RFC7031. It was intended as a gentle
introduction to the DHCP failover problem: what is in and outside of
scope, when failover is useful and when it may not be the best tool. The
DHC WG decided to split it into separate document, because there are
lots of things to explain and the core document is already obscenely
large. This RFC is clearly linked at the end of Introduction section.
Perhaps it would be more useful to move this reference to the front and
add a sentence or two of explanation? How about this text added as a
first paragraph?

This document defines a DHCPv6 failover protocol, which is a mechanism
for running two DHCPv6 servers with the capability for either server to
take over clients' leases in case of server failure or network
partition. For a general overview of the DHCPv6 failover problems, use
cases, benefits and shortcomings, see RFC7031."
>
>  
>
> Q3:        In the Introduction, I suggest adding a reference to the
> first occurrences of “DHCP service” and “DHCP server”.
>
DHCP service is not a formal term. It was intended as a general
description of the DHCP server availability. How about we replace "or
provide a DHCP service" with "to provide service to DHCP clients"?
>
>  
>
> Q4:        In the Introduction, you switch between “This protocol” and
> “The failover protocol”. Please use consistent terminology. This
> applies to the document in general.
>
There are 8 instances of "this protocol" and 20 of "the failover
protocol". We will go with the latter, ok?
>
>  
>
> SECTION 4:
>
>  
>
> Q5:        In the Abstract and Introduction it is said that DHCPv6
> does not provide server redundancy. Then section 4 talks about
> failover concepts and mechanism.
>
>  
>
> Are those concepts something used for DHCPv6 today, but for some
> reason do not fulfil the failover protocol requirements?
>
>  
>
> OR, are these general concepts that will be supported by implementing
> the failover protocol?
>
>  
>
> I think it would be good to have an introduction statement clarifying
> that.
>
These are not being used by any non-failover DHCPv6 today. There is one
implementation of this draft that uses them. Most of those mechanisms
are required by any sane failover design (e.g. any kind of failover-like
solution has to use lazy updates or it would be undeployable due to
performance), but some are specific to this particular failover design
(e.g. the allocation algorithms). Would the following text added at the
beginning of Section 4 be sufficient?

"The following section describes core concepts that are a foundation of
this failover protocol. Allocation strategies described in Section 4.2
are specific to this particular failover design, while the concept of
Lazy updates (Section 4.3) and MCLT (Section 4.4) seem to be necessary
for any failover design."

Also, if you think it's useful, we can repeat a 7031 reference there,
possibly to Section 5.2 that discusses lazy updates
(update-after-response) and its alternative (update-before-response).

Thanks again for your review,
Tomek