[CDNi] draft-ietf-cdni-cache-control-metadata-00 submitted addressing initial draft feedback

[Glenn Goldstein <[e-mail]>]

draft-ietf-cdni-cache-control-metadata-00 has been submitted,
replacing draft-power-cdni-cache-control-metadata and addressing issues
identified by Kevin Ma as follows:

section 3: is "internal" mandatory-to-specify if "force-internal" is
specified? similarly, is "external" mandatory-to-specify if
"force-external" is specified? or, is the intent to support forced clearing
of the cache control policy?
** RESPONSE: clarification added. The default is “as-is””, regardless of
whether "force is unspecified, “True”, or, "False ".

section 3, "internal" description: "to be used" -> "that MUST be used" ?
** RESPONSE: fixed

section 3, "external" description: "to be used by" -> "that MUST be
returned to" ?
** RESPONSE: fixed. when with "MUST be expressed to"

section 3, example 3: the link to SVTA2032 gives me a 404 not found?
assuming that is resolved, and SVTA2032 adequately describes
MI.OriginResponseState, i'm not sure this example is necessary. it doesn't
enhance my understanding of how to parse the MI.CachePolicy.
** RESPONSE: 404 errors fixed. While example 3 may not enhance one's
understanding of how to parse the MI.CachePolicy, it does provide context
for how MI.CachePolicy would be used in a real-world scenario in
conjunction with other portions of the CDNI standard.

section 4: "useful when there a" -> "useful when a"
** RESPONSE: fixed

section 4, "error-codes" description: "or one of the special" -> "or any of
the special"? "one" implies to me that i can only pick "4xx" or "5xx" but
not both? also, ignoring the obvious redundancy inefficiencies, is there
any concern with specifying duplicate (e.g., ["404, "404"]) or overlapping
(e.g., ["404", "4xx"]) values?
** RESPONSE: fixed, with clarification on redundancies being allowed.

section 4, "error-codes" MtS: "MAY function" -> "functions" or "MAY be
used". though the use of this functionality is optional, the actual
override functionality is not optional.
** RESPONSE: fixed

section 4, example: perhaps add an "xx" to the example?
** RESPONSE: fixed

section 4, example: "origin.." -> "origin."
** RESPONSE: fixed

section 5: "to use by a dCDN" -> "the dCDN MUST use"?
** RESPONSE: fixed

section 5: "content be served" -> "content MUST be served" or "content MUST
only be served"?
** RESPONSE: fixed

section 5: what is the relationship between "stale-while-revalidating" and
"failed-refresh-ttl"? the latter is how long to wait until revalidating,
and the former is what to do before revalidating? I found it confusing,
specifically because the "failed-refresh-ttl" MtS sections states: "The
default is zero, in which no stale content is served.", but if true,
example 1 doesn't make sense? perhaps the "failed-refresh-ttl" MtS section
needs a further qualification? does "failed-refresh-ttl" only have meaning
if "stale-if-error" is specified? it is also not clear to me what the
difference is between "revalidating" and "refreshing" (they seem to be used
interchangeably in example 3, if so, should we make it consistent)?
** RESPONSE: "refreshing" changed to "revalidating" for clarity and
failed-refresh-ttl property renamed to failed-revalidation-delta-seconds

section 5, "stale-if-error" description: "or one of the special" -> "or any
of the special"? "one" implies to me that i can only pick "4xx" or "5xx"
but not both? also, ignoring the obvious redundancy inefficiencies, is
there any concern with specifying duplicate (e.g., ["404, "404"]) or
overlapping (e.g., ["404", "4xx"]) values?
** RESPONSE: fixed, with clarification on redundancies being allowed.

section 5, "stale-if-error" MtS: "may function" -> "functions" or "MAY be
used". though the use of this functionality is optional, the actual
override functionality is not optional.
** RESPONSE: fixed

section 5, example 3: the indentation is off (two extra spaces) for
"stale-if-error" and "failed-refresh-ttl".
** RESPONSE: fixed

section 6, are the "ProcessingStages" necessary for this draft? could we
just describe the "MI.CacheBypassPolicy" without the "ProcessingStages".
SVTA2032 is not a normative reference, but the encoding examples seem to
make it normative. i think it would be simpler to leave the processing
stages out of this draft?
** RESPONSE: fixed, with an example illustrating the use of these new MI
Cache Policy objects moved to a new "Informative Examples" section.

section 7, example 3: the link to SVTA2031 gives me a 404 not found?
assuming that is resolved, is SVTA2031 normative, i.e., do i need to
understand SVTA2031 to parse "expression"? "expression" is defined as a
String, so technically i can parse it and store it as a string, but the
type description says "An expression using the Metadata Expression Language
(MEL) [SVTA2031]". perhaps we could soften this language to allow for other
strings and make it more generic? should it support different expression
languages? would we want it to be extensible (e.g., have an additional type
field)?
** RESPONSE: SVTA2031 will be characterized as normative, as the intent is
to support only the Metadata Expression Language documented in that
specification. At some time in the future, if needed, other expression
languages may be supported.

section 8: i don't think a conclusions section is necessary
** RESPONSE: removed

section 10.1: this obviously needs to be filled out.
** RESPONSE: added