Re: [alto] AD review of draft-ietf-alto-cost-calendar-07

[Mirja Kühlewind <[e-mail]>]

Hi Sabine,

please see inline.

On 14.08.2018 19:03, Randriamasy, Sabine (Nokia - FR/Paris-Saclay) wrote:
> Hello Mirja,
>
> Thank you so much for your thorough review and suggestions. The authors will discuss open points such as those relating with metrics and your they will be integrated in a next version.
>
> Best regards,
> Sabine and co-authors
>
> -----Original Message-----
> From: Mirja Kuehlewind (IETF) [mailto:[e-mail]]
> Sent: Monday, August 06, 2018 4:12 PM
> To: [e-mail]
> Cc: [e-mail]
> Subject: AD review of draft-ietf-alto-cost-calendar-07
>
> Hi all,
>
> I reviewed this draft and there are a few minor fixes that we need before we can start IETF last call:
>
> 1) Please remove the following sentence, or refer to the respective sections instead:
> „IANA considerations and security considerations will be completed in
>     further versions."
> [[SR]] [[SR]] OK, will do and will remove previous sentence on section 5 that is no longer applicable.
>
> 2) Please fix everywhere in the doc:
> "Content-Length: TODO“
> [[SR]] the value "TODO" was left on purpose, in case we would be asked to change the names of the metrics. The plan was to fill them up when all the example metrics will be approved by the reviewers.
Which reviewers? I think the should be filled before we go to IETF last 
call, so please do it now.


>
> 3) Also please make all occurrences of RFC7285 an actually reference ("[RFC7285]" instead of only „RFC7285“).
> [[SR]] OK, will do
>
> 4) Then I also have a question about this example in Sec 4.1.2:
> "For example: if the "calendar-start-time" member has value "Mon, 30
>     Jun 2014 at 00:00:00 GMT" and if the value of member "repeated" is
>     equal to 4, it means that the calendar values are the same values on
>     Monday, Tuesday, Wednesday and Thursday.  The ALTO Client thus may
>     use the same calendar for the next 4 duration periods following
>     "calendar-start-time“.“
> I think this example only makes sense if also the duration period based on "time-interval-size“ and "number-of-intervals“ with „1 hour“ and „24“ respectively is given, right? Can you please add this here.
> [[SR]]  Would the following re-phrasing be fine?
> "For example: suppose the "calendar-start-time" member has value "Mon, 30 Jun 2014 at 00:00:00 GMT", the "time-interval-size" member has value "1 hour", the "number-of-intervals" member has value "24" and the value of member "repeated" is equal to 4. This means that the calendar values are the same on Monday, Tuesday, Wednesday and Thursday on a period of 24 hours starting at 00:00:00 GMT.
> The ALTO Client thus may use the same calendar for the next 4 days starting at "calendar-start-time" and will only need to request a new one for Friday July 4th at 00:00:00 GMT."
Yes, please.

>
> And one minor (technical) comment/question that I would like to discuss before we go into IETF last call:
> Why is "time-interval-size“ combining the value and unit in one element, instead of e.g. using "time-interval-unit“ and "time-interval-value“? Would that not make the implementation of the parsing much simpler?
> [[SR]] Parsing 2 separate fields would avoid ambiguities indeed, the idea with this proposed format was to spare one member to convey in the responses. We took inspiration from the encoding format of constraints by an ALTO Client in 11.3.2.3 of RFC 7285 that follows a similar pattern, e.g. "le 15". Would it help if we rephrase the format specification of the "time-interval-size“  value in section 3.1 to avoid parsing errors ?
I think the approach is fine. I was just wondering why it was chosen.

>
> And finally, you use the cost type names "num-routingcost", "num-latency", "num-pathbandwidth" and "string-quality-status“ as well as metrics "routingcost", "latency" and "bandwidthscore" in the examples and say that "The cost types in this example are either specified in the base ALTO
>     protocol or may be specified in other drafts see
>     [draft-ietf-alto-performance-metrics] or defined in this draft for
>     illustrative purposes.“
> However, non of these metrics are actually defined in [draft-ietf-alto-performance-metrics], nor are they „defined“ in this doc. Would it maybe make sense to actually use cost types and metrics from [draft-ietf-alto-performance-metrics] in these examples (or remove the reference and provide some kind of definition)?
> [[SR]]  We propose to revise the related parts in section 3.3 as follows:
> - only reference [draft-ietf-alto-performance-metrics] in section 2.2.1 that explains how Calendars support all kinds of metrics and modes and remove it from section 3.3,
> - keep the "routingcost" metric,
> - define example illustrative costs and metrics in section 3.3, that have names such as "num-owdelay", "num-throughput",  "string-service-status". If we are asked to replace them by cost types with nonsensical metric names such as "shoesize" or "cattle-head-count", we will do so.
Yes, please define the metric accordingly. I'm okay to have metric that 
actually make some sense. Just make sure that these metric are 
understood as example metrics only and should not be used. Or would it 
make more sense to "just" use some metrics from 
[draft-ietf-alto-performance-metrics]?

Mirja

>
> Given I reviewed the whole doc, also a couple of editorial comments/proposals below (however, I fully leave it to the authors' judgement to apply these or not).
>
> Thanks!
> Mirja
> [[SR]] Thanks a lot for your guidance and suggestions
> Sabine
>
>
> ————————————————
> Other editorial comments:
>
> 1) The abstract is quite long. I think if it could be formulated more crisp, it would be easier to read, e.g. see the text in the shepherd write-up
>
> "This document is an extension to the base ALTO protocol (RFC 7785).  It extends the ALTO cost information service such that applications decide not only 'where' to connect, but also 'when'.  This is useful for applications that need to perform bulk data transfer and would like to schedule these transfers during an off-peak hour, for example.“
> [[SR]] OK will do
>
> 2) I know that IRD is on the known abbreviation list, maybe still spell it out at first occurrence for the ease of the reader…?
> [[SR]] OK will do
>
> 3) As the alto base spec is already published for a while, maybe:
> OLD
> "IETF is currently standardizing the ALTO protocol which aims at
>     providing guidance to overlay applications…“ NEW "The ALTO protocol provides guidance to overlay applications…“
> [[SR]] OK will do
>
> 4) Maybe:
> OLD
> “...for example by deferring backup to night
>     during traffic trough.“
> NEW
> "for example by deferring backups or other background traffic to off-peak hours.“
> [[SR]] OK will do
>
> 5) To be inline with previously used wording, maybe OLD „...we expect to further
>     gain on storage and on the wire data exchange…“ NEW "we expect to further save network and storage resources…“
> [[SR]] OK will do
>
> 6) sec 2.2 "The protocol extension placeholders for an ALTO Calendar are: the
>     IRD, the ALTO requests and responses for Cost calendars.“ Not sure I fully understand the word „placeholder“ here, maybe:
> „To realize an ALTO Calendar, this document extends the
>     IRD, the ALTO requests and responses for Cost calendars.“ ?
> [[SR]] OK will do
>
> Also further
> "Extensions are designed to be light and ensure backwards
>     compatibility with base protocol ALTO Clients and with other
>     extensions.  It uses section 8.3.7...“ What is „it“ here?
> Maybe:
> „This extension is designed to be light and ensure backwards
>     compatibility with base protocol ALTO Clients and with other
>     extensions. As recommended, it relies on section 8.3.7…“ ?
> [[SR]] OK will do
>
> 7) Sec 4.1.1:
> "A Calendar-aware ALTO client supporting single cost type values, as
>     specified in RFC7285, MUST provide an array of 1 element:
>
>                            "calendared" : [true];“ This could be stated more clearly e.g.
> "A Calendar-aware ALTO client only supporting single cost type values, as
>     specified in RFC7285, that aims to request a Calendar MUST provide
>     an array of 1 element:
>
>                            "calendared" : [true];“
> [[SR]] OK will do
>
> 8) 4.2.2
> "If the value of member "calendared" is equal to 'false' for a given
>     requested Cost Type, the ALTO Server must return, for these Cost
>     Types, a single cost value as specified in RFC 7285.“ Probably use normative MUST here instead.
> [[SR]] OK will do
>
> 9) Probably RFC5246 does not need to be a normative reference for this doc (as it is already normative for RFC7285).
> [[SR]] OK, will move it to Informative
>
>
>