Well, it looks like it's broken, then—at least the ruby parser (https://github.com/inukshuk/edtf-ruby) and the javascript parser (https://github.com/inukshuk/edtf.js) agree—neither considers 1990?-uu-uu a valid EDTF date, which agrees with my reading of the BNF.

- David Newbury
-----------------------------------
p. (773) 547-2272
e. [log in to unmask]

On Thu, Mar 9, 2017 at 11:50 AM, E. Zimmermann <[log in to unmask]> wrote:

The ? applies to the 1990. If the BNF grammar disallows this then it is broken!

 

 

From: Discussion of the Developing Date/Time Standards [mailto:[log in to unmask]GOV] On Behalf Of David Newbury
Sent: Thur
sday, March 09, 2017 4:34 PM


To: [log in to unmask]
Subject: Re: Uncertain, unspecified dates

 

1990?-uu-uu appears to be what state my intentions, but from my reading of the BNF it does not appear to be a valid EDTF date.  

 

(internalUncertainOrApproximate and internalUnspecified appear to be mutually exclusive)

 

 


- David Newbury
-----------------------------------
p. (773) 547-2272
e. [log in to unmask]

 

On Thu, Mar 9, 2017 at 9:58 AM, E. Zimmermann <[log in to unmask]> wrote:

If I am uncertain of the year being 1990 and don’t yet know the month and day but expect to be able to fill them in at some time…..

1990?-uu-uu

 

Uncertainty is quite different from “approximately”..   That means “in the neighborhood of”…

 

So if I think it was somewhere around 1990 etc. etc.

1990~-uu-uu

 

The character ‘u’ is more than just a placeholder BUT I think it is probably best to think of them as just that..

 

The problem with (uu)? is that it is saying “I expect to be able to put in some unreliable numbers”.. Normally we expect or hope to get good data.. Once we collect something we can speak about trust.. like.. well I got a day but I don’t 100% trust it.. In fact at some point we did have a notion of belief in the reliability of a value..

See EDTF’s list of possible future features..

 

Right now ‘?’ means just “unreliable”..  (its level being unstated  whence implementation defined and might include another man’s date without qualification.. See the current discussion on “alternative facts”)

 

 

From: Discussion of the Developing Date/Time Standards [mailto:[log in to unmask]GOV] On Behalf Of David Newbury
Sent: Thursday, March 09, 2017 3:34 PM
To: [log in to unmask]
Subject: Re: Uncertain, unspecified dates

 

What I am trying to express is:

 

I know an event happened on a specific day.  My research has indicated that it took place on some specific day in 1990, but has not indicated yet a month or day, but I do not believe the knowledge is unattainable.  

 

That is appropriately recorded as 1990-uu-uu

 

However, I am uncertain that my research is correct.  It may, in fact, have been 1995, or 1850, or some other year entirely.  I would like to indicate that the date, while known, is untrustworthy.  In my reading of the specification, that appears to be the meaning of a date such as 1990-10-10?:  That it is a recorded date, but that is is uncertain or untrustworthy.

 

I would like to be able to state both of these things, and they seem to not be in conflict.  My certainty/uncertainty of a date is not related to the precision to which I know that date.

 


- David Newbury
-----------------------------------
p. (773) 547-2272
e. [log in to unmask]

 

On Thu, Mar 9, 2017 at 7:34 AM, E. Zimmermann <[log in to unmask]> wrote:

What does 1990-uu-uu say? It says that I KNOW that an event took place in 1990. I currently don’t know the day or month but I expect to get this data at some time in the future.

 

What could uncertainty mean here? That I don’t expect to get the missing data or that I’m not quite sure that I don’t know the data?

The ‘u’ character in your use case is nothing other than a placeholder.. Perhaps you get some information that it was on the 15th of some month BUT you are uncertain.. that is where the ‘?’ as an expression of uncertainty enters..

 

To round up:

-       You know that an event took place an some instant in a day

-       You know that it occurred in 1990

-       You don’t know the month or day (yet)

-       You may NEVER be able to know the month or day—but you may one day and by best practice you indeed expect to be able to “fill in the blanks”..

1990-uu-uu

 

If, however, you can’t measure the day or even month (and with current knowledge don’t really expect to).. But know it was 1990 then

 

1990

 

Just as a sample I put on a scale may weigh 20g on my crude balance.. It weighs 20g.. If later I get a better balance.. I can change the information. Say to 19.789g

Same with dates..  Dates are not carved in stone!

 

The new addition we have provided to ISO 8601 is that we know have finally a concept in date analogous to weighs and lengths.

 

 

From: Discussion of the Developing Date/Time Standards [mailto:[log in to unmask]GOV] On Behalf Of David Newbury
Sent: Wednesday, March 08, 2017 8:55 PM
To: [log in to unmask]
Subject: Uncertain, unspecified dates

 

I'm trying to implement a system using EDTF, and I'm trying to record an event which:

 

1. Took place with day precision. (It is a discrete event, and my system does not record with more than day precision)

2. Is uncertain (I have a date, but I am unsure if it is accurate)

3. Is unspecified.  (I know the year it took place in, but I cannot yet determine the month or day).

 

It seems logical that this would be recorded as  1990-uu-uu?, but that does not appear to be a valid EDTF date.

 

Is there a reason that "uncertain" and "unspecified" cannot be used together?  They seem to be orthogonal concerns.

 

- David Newbury
-----------------------------------
p. (773) 547-2272
e. [log in to unmask]