>> Denenberg, Ray:
>>> If [seasons in intervals] is to be added it would be with the restriction that if a season is supplied in one part of an interval then a season must be supplied in the other part.
>>> (Or more generally, they must both be in the same division class.)
Whether this restriction would be sensible depends on how specific seasons are to be specified in ISO 8601.
>>> Further, clarification will be added that a season is associated
>>> with the year in which it begins.
This would be inconsistent with how calendar weeks are defined by ISO. W01 is not the one with the first Monday of the year but the one with the first Thursday of the year! (In 2018 -W01-1 and -01-01 coincided, so the question of first week was trivial to decide.)
>>> The latter pertains specifically to winter: we are currently in winter 2017, not winter 2018.
By all conventions set forth by ISO 8601 until now, we are not (unless this "winter" is a period from, e.g., 1 November through 31 January).
You are obviously assuming an astronomical definition of seasons. This needs to be settled and stated in the standard. I strongly believe extensions to the MM field should be restricted to divisions that contain whole calendar months, which incidentally includes European meteorological definitions of the seasons. Introduce 'S1' through 'S4' or something else for astronomical seasons. Some academic trimesters and fiscal quarters (and other divisions) also use descriptive terms like "winter", but often agree with neither the meteorological nor the astronomical definition of seasons. EDTF may have been developed for bibliographic purposes where the astronomical, Northern-hemispheric definition predominates, but ISO 8601 has a much broader scope.
>>> This may cause discomfort because although this is the commonly accepted definition
ISO 8601 does not need to comply with external sources if that would break internal consistency. The whole point of a standard is to select a single definition for ambiguous terms.
> Hostage, John <[log in to unmask]> wrote:
>> I think there are two conflicting understandings of seasons at work here.
There are indeed.
>> One tries to nail down a precise definition that can be validated and manipulated mathematically.
This is the definition that would accommodate existing ISO 8601 practice.
>> The other realizes that seasons are bound up with language, culture, and the part of the world being talked about.
This one is what librarians and others involved with bibliographies would prefer apparently.
>> Maybe it’s futile to include seasons in a date and time standard because they are so subjective.
The beauty of a standard is that it can settle on a single definition that is most consistent with other definitions used in that standard.
>> It seems like a non-starter to say that if a publisher calls something “winter 2018” we have to represent it as 2017-24.
The current, cumbersome way would be, for instance, "2017-12/2018-02". You also have to translate natural language terms like "January" or "Monday" or "afternoon"; seasons are just a bit harder.
>> And if an issue says “winter 2017” we won’t know if that means 2017-24 under that definition or 2016-24, unless perhaps we have a lot of adjacent numbered issues to compare it to.
That is reason enough not to encode it in a ISO-8601 date notation, in my humble opinion..
Stephen Hearn <[log in to unmask]>:
> It might be simpler to dispense with specific seasonal terms
> and just authorize a relative sequence specified by year:
Indeed. Month names are also only informative, not normative in ISO 8601.
> The defined values could be extended to cover cases where a year is divided
> into more than four "seasons" or non-month units.
That is why it has been suggested to use the first digit semantically (as far as possible): Let it designate either the number of months in each division or the number of such divisions per year. Four seasons (=> quarters) of (about) three months each (=> trimesters) would thus preferably run from either 31 through 34 or from 41 through 44.
If the first division should (optionally) not start with 1 January, things get more complicated and weird. The earliest first day of a max. 91-day first season should be 18 October of the preceding year and the latest first day of that season should be 15 February, which would ensure that the majority of the days in the season is in the year their season nominally belongs to. This would be consistent with how ISO 8601 defines the first week of the year, i.e. the one with at least 4 out of 7 days within that year.
> The particular terms to be used for each serial's issue designations would
> have to be recorded some other way, as arguably they should be given the
> idiosyncratic nature of such term use.
I completely agree. This is how ISO date standards have worked for decades.