Re: [alto] Benjamin Kaduk's Discuss on draft-ietf-alto-cost-calendar-19: (with DISCUSS and COMMENT)

["Randriamasy, Sabine (Nokia - FR/Paris-Saclay)" <[e-mail]>]

Hello Benjamin,

Thanks a lot for your review and feedback. Regarding the discuss, indeed, this extension does not support constraints on calendars and definitely it is better to explain that it is not present and why. We will get back to you with proposals for text to be added in relevant sections of the document and also respond to your comments.

Best regards,
Sabine
  

-----Original Message-----
From: Benjamin Kaduk via Datatracker <[e-mail]> 
Sent: Tuesday, March 3, 2020 11:15 PM
To: The IESG <[e-mail]>
Cc: [e-mail]; [e-mail]; [e-mail]; Vijay Gurbani <[e-mail]>; [e-mail]
Subject: Benjamin Kaduk's Discuss on draft-ietf-alto-cost-calendar-19: (with DISCUSS and COMMENT)

Benjamin Kaduk has entered the following ballot position for
draft-ietf-alto-cost-calendar-19: Discuss

When responding, please keep the subject line intact and reply to all email addresses included in the To and CC lines. (Feel free to cut this introductory paragraph, however.)


Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
for more information about IESG DISCUSS and COMMENT positions.


The document, along with other ballot positions, can be found here:
https://datatracker.ietf.org/doc/draft-ietf-alto-cost-calendar/



----------------------------------------------------------------------
DISCUSS:
----------------------------------------------------------------------

What's the justification for removing the 'constraints' field of ReqEndpointCostMap, compared to RFC 7285?


----------------------------------------------------------------------
COMMENT:
----------------------------------------------------------------------

Section 1

   In this document, an "ALTO Cost Calendar" is specified in terms of
   information resources capabilities that are applicable to time-

nit: "information resources capabilities" has two plurals.  Either make "resources" possessive ("resources'") or just use "resource".

Section 1.1

I'm not sure how much *archival* value there is in the current discussion of SENSE and Unicorn.

Section 3.1

   sent to the ALTO Clients needing it.  The "ALTO Incremental Updates
   Using Server-Sent Events (SSE)" Service
   [I-D.ietf-alto-incr-update-sse] can be used to update the calendar
   faster if supported by both the Server and the Client.

nit(?): faster than what?

Section 3.3

   The present document extends the definition of a legacy cost map
   given in [RFC7285] to allow a cost entry to be an array of values,
   with one value one per time interval, instead of being just one
   number.  Therefore the implementor of this extension MUST consider
   that a cost entry is an array of values.  Specifically, an

nit: this is not quite true, strictly speaking -- the cost entry is only an array when the cost calendar functionality is active, which is not a subset of when the extension is implemented.
Also, if the cost entry is definitively an array, then this would be replacing the definition, not extending it.

   implementation of this extension MUST parse the "number-of-intervals"
   attribute of the "calendar-attributes" in an IRD entry announcing a
   service providing Cost Calendar.  The implementation then will know
   that a cost entry of the service will be an array of values, and the
   expected size of the array is that specified by the "number-of-
   intervals" attribute.

Is it a protocol error if the cost entry is a scalar or an array of different size than expected?  Where do we specify error handling for those cases?

   To realize an ALTO Calendar, this document extends: the IRD and the
   ALTO requests and responses for Cost Calendars.

nit: no colon needed.

   o  it allows an ALTO Server to offer Calendar capabilities on a cost
      type, with attributes values adapted to each information resource.

I'm not entirely sure what this is intending to say.  Is the idea that this is a general mechanism that could be applied to capabilities of all cost types (as opposed to, e.g., making a new cost mode for "calendar of numerical" that would need many new types to support different types of calendar)?

Section 3.3.2

   The ALTO protocol extensions for Cost Calendars have been defined so
   as to ensure that Calendar capable ALTO Servers can provide legacy

nit: hyphenate "Calendar-capable" (and similarly throughout).  I see that "calendar-aware" is already hyphenated, which is good.

   ALTO Clients with legacy information resources as well.  That is, a
   legacy ALTO Client can request resources and receive responses as
   specified in [RFC7285].

Should we say anything about Calendar-aware ALTO Clients being able to get useful responses from legacy Servers as well?

Section 4

   The Calendar attributes in the IRD information resources capabilities
   carry constant dateless values.  A Calendar is associated with an

"constant" in what sense?

Section 4.1

   types.  A cost type name MUST NOT appear more than once in the
   "calendar-attributes" member of a resource entry; multiple
   appearances of a cost type name in the CalendarAttributes object of
   the "calendar-attributes" member MUST cause the ALTO Client to ignore
   any occurrences of this name beyond the first encountered occurrence.

It seems that this is most important in that the given resource entry has an array of calendar-attributes, and we need to know which element of that array to use when processing calendars for that cost type.
Indicating the extra layer of structure in the description around this requirement would help the reader.

   An ALTO Server SHOULD specify the "time-interval-size" in the IRD as
   the smallest it is able to provide.  A Client that needs a longer
   interval can aggregate multiple cost values to obtain it.

nit: we haven't defined "time-interval-size" yet, so either moving this later or giving a bit more explanation might be useful.

   o  "cost-type-names":

      *  An array of one or more elements indicating the cost-type-names
         in the IRD entry to which the capabilities apply.

Which capabilities?

      *  is the duration of an ALTO Calendar time interval in a unit of
         seconds.  A "time-interval-size" value contains a non-negative
         JSONNumber.  Example values are: 300 and 7200, meaning that
         each Calendar value applies on a time interval that lasts 5
         minutes and 2 hours, respectively.  Since an interval size
         (e.g., 100 ms) can be smaller than the unit, the value
         specified may be a floating point (e.g., 0.1).  Both ALTO
         Clients and Servers should be aware of potential precision
         issues caused by using floating point numbers; for example, the
         floating number 0.1 cannot be represented precisely using a
         finite number of binary bits.  To improve interoperability and
         be consistent with [RFC7285] on the use of float point numbers,
         the Server and the Client SHOULD use IEEE 754 double-precision
         floating point [IEEE.754.2008] to store this value.

nit: please be consistent about using "floating-point number" (vs., e.g., "floating point" or "float point").

   - Providing attribute "cost-type-names" together with "time-interval-
   size" and "number-of-intervals" improves the readability of the
   Calendar attributes specified for an IRD resource and avoids
   confusion with Calendar attributes of other cost types.

I'm not sure I understand either of these points (how readability is helped and how confusion is avoided).

Section 4.2

   It may be useful to distinguish IRD resources supported by the base
   ALTO protocol from resources supported by its extensions.  To achieve
   this, one option, is that a "root" ALTO Server implementing base
   protocol resources and running at a given domain, delegates
   "specialized" information resources such as the ones providing Cost
   Calendars, to another ALTO Server running in a subdomain.  The "root"

How would a Client know that this mechanism is in use?

   This document provides an example, where a "root" ALTO Server runs in
   a domain called "alto.example.com".  It delegates the announcement of
   Calendars capabilities to an ALTO Server running in a subdomain
   called "custom.alto.example.com".  The location of the "delegate
   Calendar IRD" is assumed to be indicated in the "root" IRD by the
   resource entry: "custom-calendared-resources".

This is "assumed" only for the purpose of the example, and not as a general protocol mechanism, right?

   Another benefit of delegation is that some cost types for some
   resources may be more advantageous as Cost Calendars and it makes few
   sense to get them as a single value.  For example, if a cost type has
   predictable and frequently changing values, calendared in short time
   intervals such as a minute, it saves time and network resources to
   track the cost values via a focused delegate Server rather than the
   more general "root" Server.

Is the idea just that you compartmentalize the fast-changing stuff from the slow-changing stuff, so that your listing of what is changing quickly only includes the things actually changing on that timescale, so you don't end up also listing the calendar for the slowly-changing things in the same response?
Also, nit: s/few/little/


Is there a reason for "map" and "calendar" to be in different orders in "filtered-cost-map-calendar" and "endpoint-cost-calendar-map"?

   o  the Calendar for "owdelay": is an array of 12 values each provided
      on a time interval lasting 300 seconds (5 minutes).

nit: s/owdelay/num-owdelay/

Sectgion 5

I'd consider using a more recent Date in the example.

Section 5.1.1

   The input parameters of a "legacy" request for a filtered cost map,
   defined by object ReqFilteredCostMap in section 11.3.2 of [RFC7285],
   are augmented with one additional member.

There's probably some pedantic consideration here about "other extensions", such as the "multi-cost-types" member from RFC 8189.

   This field is an array of 1 to N boolean values, where N is the
   number of requested metrics.  Each entry corresponds to the requested

Maybe reference RFC 8189 so the reader doesn't get confused by RFC 7285 specifying only a single metric type?

Section 5.1.2

   The non Calendar specific "meta" fields of a calendared Filtered Cost
   Map response MUST include at least:
     [...]

side note: this structure where we effectively repeat the requirements of all previous specifications is not going to scale well with future extensions.

   o  each "CalendarResponseAttributes" object in the array is specified
      for one or more cost types for which the value of member
      "calendared" is equal to 'true' and for which a Calendar is
      provided for the requested information resource.

Member 'calendared' of what data structure?

   o  "cost-type-names": is an array of one or more cost-type-names to
      which the capabilities apply and for which a Calendar has been
      requested.  The value of this member is a subset of the "cost-
      type-names" array specified in the corresponding IRD Calendar
      attributes.

Just to check my understanding: in the IRD Calendar attributes, there's a "calendar-attributes" member whose value is an array of objects, and each of those objects has a "cost-type-names" member whose value is an array of cost-type names.  So what we're saying here is that the "cost-type-names" in a CalendarResponseAttributes entry are a subset of the union of the "cost-type-names" members in all of the entries in the "calendar-attributes" array of objects in the IRD calendar attribute, which is not exactly what the quoted text seems to be saying.

   o  "time-interval-size": as specified in Section 4.1 and with the
      same value.

   o  "number-of-intervals": as specified in Section 4.1 and with the
      same value.

nit: "with the same value" could perhaps imply that there is a requirement to have actually published an IRD calendar for this resource and literally take the value from that existing calendar; "with the same semantics" should relax that (perceived) requirement.

Section 5.1.3

   An example of non-real time information that can be provisioned in a
   Calendar is the expected path throughput.  While the transmission

nit: "non-real time" and "non-real-time" have different meanings, and I think the latter is the intended one.

   usage patterns.  In this example, we assume that an ALTO Client
   requests a Calendar of network provider defined throughput ratings,

nit: hyphenate "network-provider-defined".

     Content-Length: 1013
     Content-Type: application/alto-costmap+json

Please double-check the content length (for all the examples) -- it's a bit annoying to do from the text rendering of the I-D that applies a left indent, but assuming that the initial '{' is supposed to be the first byte of the response body and the rest of the whitespace is part of the response, I get something more like 1043 octets of content.

Section 5.2.1

We should probably say that the interpretation of the various fields is the same as the RFC 7285/8189 ReqEndpointCostMap, and "calendared" the same as for ReqFilteredCostMap.

Section 5.2.3

   o  C1 for Monday, Tuesday, Wednesday, Thursday, (weekdays)

   o  C2 for Saturday, Sunday, (weekend)

   - C3 for Friday (maintenance outage on July 4, 2014 from 02:00:00 GMT
   to 04:00:00 GMT, or big holiday such as New Year evening).

I'd consider using a more recent date than 2014 throughout this example.
Also, nit: please use a consistent character to represent the bullet point.

  Host: alto.example.com

Would it be more consistent with previous examples to use custom.alto.example.com here (and in other examples)?

Section 7


   [RFC8446] specifies TLS 1.3 and writes in its section 1: "While TLS
   1.3 is not directly compatible with previous versions, all versions
   of TLS incorporate a versioning mechanism which allows Clients and
   Servers to interoperably negotiate a common version if one is
   supported by both peers".  ALTO Clients and Servers SHOULD support
   both TLS 1.3 [RFC8446] and TLS 1.2 [RFC5246], and MAY support and use
   newer versions of TLS as long as the negotiation process succeeds.

I know this document has been in the works for a long time and so there may be reluctance to make changes on this front, but I'll note that RFC
8446 has been out for a year and half, so under normal conditions we'd just say "use TLS 1.3 or newer" without mentioning 1.2.

   participate in a DDoS attack.  The Calendar information would be
   valuable information for when to persecute a DDoS attack.  A

nit: "persecute" is an unusual word here; "execute" or "perform" seem like better alternatives.

   Hence, a more robust ALTO Client should adapt and extend protection
   strategies specified in Section 15.2 of the base protocol.  For

It's probably better to use the RFC number than just "the base protocol".

Section 10.2

I think [IEEE.754.2008] needs to be normative, as do RFCs 5246 and 8446, and draft-ietf-alto-incr-update-sse.