EDTF date pattern {..1984} is sort of impossible to support?

[kspurgin]

Currently this displays as all of the dates: 1984 or some earlier date

Which is ambiguous/confusing because it seems to say "all of the dates, but choose between 1984 and an earlier date."

What I understand this pattern to actually mean is: 1984 and all earlier dates or up to and through 1984

Which...

First, what has ever happened on every single date from the beginning of time until 1984, that would ever need to be represented in Islandora? 🤯

And how do you handle this in Solr?

I think what you end up doing is choosing an arbitrary "earliest possible date" and setting the "beginning of time" to that under the hood, and that should be documented somewhere.

[kspurgin]

meta-issue: #1748

[seth-shaw-unlv]

The SOLR representation of this would be [* TO 1984]. See SOLR's "Working with Dates".

[seth-shaw-unlv]

Archivists will also make dates that use this concept, usually written as something like "sometime before 1984".

[kspurgin]

Oh sweet, Solr's easier for this than I feared. (Didn't so much tangle with Solr and dates).

Totally get the "before 1984" use case but then my literal brain drops into existential heck going, well if it is on paper, we know it wasn't before a certain date... like whenever paper was invented... so saying blanket "before 1984" is actually wrong. And yet it is practical and expedient and used!

I really need to stop thinking about date/time and code so much---it's unhealthy. 🤣

[kspurgin]

Except, to be pedantic, I think {..1984} is "sometime before 1983" since it includes 1984 in the set.

Gotta have that -1 in the display logic.

[kspurgin]

Sorry, no, I'm going back to this cannot reasonably be used, as it literally means ALL the dates represented in this set.

The archivist use case is [..1984] -- a date sometime in this range. Not all of them.

[kspurgin]

My understanding:
[..1984] = some date before 1983
{..1984} = all dates before 1983

[seth-shaw-unlv]

I see what you mean and I think your reading of the spec is correct.

Although, I think if we are being pedantic, nothing could use the {..1984} encoding because nothing that has happened since the beginning of time (which is what, the laws of physics? if even they are absolute, considering the big bang; when does 'time' begin, etc.) abruptly ended at the dawn of 1984.

In any case, SOLR isn't going to give us this granularity of date-range semantics. /1984, ../1984, [..1984], and {..1984} are all going to boil down to [* TO 1984].

Beyond that, is training the metadata creators. I know our librarians are simply using the interval signifier (e.g. 1950/1954 to express "some date in this range", rather than the set notation (e.g. [1950..1954]), or even further, strictly speaking, a photograph was taken on a day so it should be [1950-01-01..1954-12-31]. (Although I suppose you could take a year-long exposure if you got the chemistry right...) For better or worse, they are using the simplest form to express an idea where they trust users to understand what they mean from the context.

[kspurgin]

Totally agree with your 2nd paragraph. I think they were trying to be complete and cover all possible expressions. But some of them are just not that useful.

Was working through the Solr aspect thinking through how the heck it handles something like XXXX-12-XX -> "Some day in December in some year" and arrived at, "Yeah that's basically date = *:* in Solr." That loss of granularity in the index seems FINE given how people tend to expect to use this.

To your last paragraph, I think we have an opportunity here with the display-as-training and clear documentation of EDTF pattern-to-display mapping? Maybe? If what you enter results in a display that you recognize is inaccurate, then maybe you go back and enter to correct thing. Or you can consult the documentation looking for what you want the display to express, and then enter the right pattern.

[seth-shaw-unlv]

Ah, I see now... our librarians decided to stick with Level 1 semantics, which doesn't have set support, ergo... 1950/1954 is the closest they have to expressing the concept 'some day in that range'; although, again, perhaps 1950-XX-XX/1954-XX-XX would be more accurate for a photograph taken on a day in that range, even if the spec doesn't include any examples like that.

[seth-shaw-unlv]

Here are the examples they provide in the metadata guide for digital collections:

“Early” 1950s = 1950?/1953?
“Mid” 1950s = 1954?/1956?
“Late” 1950s = 1957?/1959?
“Early” 1900s = 1900?/1939?
“Mid” 1900s = 1940?/1969?
“Late” 1900s = 1970?/1999?
Approximately 1905 to circa 1920 = 1905~/1920~
Possibly 1953? = 1953?
Circa 1940s-50s? = 1940%/1959%
Circa 1930s = 1930~/1939~
Circa 1970s? = 1970%/1979%
Circa 1960s-1970s = 1960~/1979~
1953 - approximately 1958? = 1953/1958%
Maybe June 1, 1999 = 1999-06-01?
Valentine’s Day, year unknown (Feb. 14) = XXXX-02-14
June 25-27, 1908 (including 26th) = 1908-06-25/1908-06-27

[kspurgin]

Which is another use case for the existence of 1950-XX-XX that didn't come up when I was asking about it (why not just record 1950?) in Metadata Support Group.

Good lord it's all such a wormhole.

[seth-shaw-unlv]

And this, my friends, is why we have professional catalogers; because it takes a professional to have the patience and attention to detail to deal with this degree of specificity provided by the spec. Most of the rest of us don't have the patience or are too apathetic.