[rssac-caucus] RSSAC002v2 published; future work items

[Wessels, Duane]

Dear Caucus members,

As you may know, the RSSAC002v2 document was published a couple of weeks ago.  That document can be found at https://www.icann.org/en/system/files/files/rssac-002-measurements-root-07jan16-en.pdf

I've been trying to keep a list of all the other proposed changes that people have mentioned while we were working on and waiting for v2 to be published.  My notes are below.  Please let me know if I've missed anything.

I'd like to suggest that RSSAC establish a work party to tackle some of these issues and come up with v3 of the document.  Unless anyone is opposed to that idea I will propose it at the next call.

DW

========================================================================


- more realistic YAML examples (Akkerhuis)

  Steve Sheng has already updated the master copy of the document
  with real examples from A-root.

- zone-size metric.

  Is this metric worth the trouble?  It is somewhat difficult to define.
  Every operator that reports this is measuring it "out of band" because
  measuring it "in band" is all but impossible.  All letters report the
  same value for this metric which leads to wasted effort.

- load-time (Crabill, Kash)

  This could be clarified with respect to large number of instances

- Are there additional metrics to consider? (Crabill)

- publication-time (Kash, Akkerhuis)

  is this still true?: "There are current DNS software logging limitations
  that inhibit the perfect collection and resolution of latency in publishing
  available data values due to the lack of zone serial numbers in AXFR/IXFR
  logging statements.  "

  For NSD this is fixed starting with version 4.12, since 4.1.4 it will
  now also be logged at a less verbose log level (on request of Anand
  Buddhdev, Ripe-NCC).

- RCODE distribution (Roy Arends)

  Section 2.5 reads "The RCODE distribution is a raw count of the RCODE
  values observed in responses during the reporting period.

  This single sentence is ambiguous, since it is interpreted in two ways.
  Either it includes responses received, or it doesnt.  That needs to be
  clarified, otherwise the numbers are meaningless as they cant be compared.

  Section 4.3 has a similar ambiguity.  Does responses-sent mean sent by
  the root server, or sent by something else, and observed by the root
  server.

- RCODE distrubtion (Ondrej Sury)

  I would move the text from 4.5 (or add short note) about extended RCODEs
  to 2.5.

- traffic-volume  (Ray Bellis)

  descriptive text in section 2.3 doesn't talk about responses.  it is
  titled "The number of queries"

  The stated rationale (measuring increasing loads over time) doesn't need
  response volumes, and in the presence of RRL it's not a useful proxy
  measurement for the reliability of a server either.