If uncertainty one must start off with a model of how to describe uncertainty. I provided one such model some time ago. There are others. That is the starting point. I see these as modifiers to our unknown predicate. The term really should be "reliability" rather than unknown or uncertain. A date, after all, that is unknown is unknown. Reliability and trust in the correctness of a given date, by contrast, can be qualitatively described: a continuum from probably wrong to highly certain.

On Mon, 24 Jan 2011 15:31:25 -0500, Ray Denenberg, Library of Congress wrote
> Simon - I am trying to respond here to your message "essentials" discussing "uncertainty" (and have changed the subject line accordingly).


>

 


> My first problem is that I don't really understand how you are defining "uncertainty", so could you please give some concrete examples.


> There is one mention of "uncertain" in the current draft spec, #316,


> Uncertain, but known to be one of a set." for example [1667,1668, 1670-1672]  One of the years 1667, 1668, 1670, 1671, 1672


> But I suspect that's not what you have in mind.  There are also several examples of "unspecified", for example,


>         "199u: some unspecified year in the 1990s "


> and there is "approximate year" e.g. 1992~


> and "questionable year" e.g. 1992?


> For purposes of this discussion let's call all of these "uncertainty"  - whether approximate, questionable, or unspecified . 


> I am assuming that your requirement is to attach some *level*  of uncertaintly, in contrast to these examples, where the uncertainly is unqualified.  e.g.. 1992? just says "questionable", unqualified, that it, it doesn't specify to what degree it is questionable.


> There are use cases for all of these examples, and in all of these cases there is no requirement to qualify the degree of uncertainty.  However that is not to say that qualification is not a requirement, but it is to say that there is a need to be able to express unqualified uncertainty, in as simple a manner as possible using a syntax that is unburdened by the existence of additional syntax to express qualification.


> Thus for example we want to say "1992?", and if the syntax allows the uncertainty to be qualified, we don't want to have to include an explicit indicator to say it is unqualified.


> Does this make sense, and are we talking about the same thing?


>

 


> --Ray


>

 


>

 


>
> From: Discussion of the Developing Date/Time Standards [mailto:[log in to unmask]] On Behalf Of Simon Grant
> Sent: Thursday, January 20, 2011 9:47 AM
> To: [log in to unmask]
> Subject: [DATETIME] essentials


>

 


> Trying to balance closure with reopening cans of worms...


>
>

 


>
> I'm sorry, I seem to have gone a bit too far and too quickly in the below.


>
>

 


>
> I think what I am really saying is that it would be a good idea to design a spec for representing uncertainty so that the essentials are independent of date and time system (calendar, units, etc.) and then to fit this to the ISO 8601 norm -- meaning Gregorian years, months, days, etc.


>
>

 


>
> To get there, I would loosen up on particular units, and focus first on the essentials of what we are trying to represent.


>
>

 


>
> However, I am fully aware that the priority of uncertainty is mine, and that other people may have other priorities.


>
>

 


>
> I would be perfectly happy to work on a profile (both restricting and extending) of ISO 8601 to express uncertainty, using many of the ideas that have been discussed here.


>
>

 


>
> Is that, perhaps, what we should be doing? Decoupling, not lumping together, specifications for different purposes. We can still coordinate them, of course, but define them separately, so we can reach consensus in a finite time...


>
>

 


>
> Thanks


>
>

 


>
> Simon


>
>

 


>
> On 23 December 2010 19:21, Bruce D'Arcus <[log in to unmask]> wrote:


> Not been following this much, and am on vacation, but in my experience these disagreements often result from unclear use case and requirements understandings.  Is that possible here?


>
> On Dec 23, 2010 2:07 PM, "Ray Denenberg" <[log in to unmask]> wrote:
> > Simon - I am preplexed, and indeed somewhat confused by your response.
> >
> >
> >
> > Why prioritize a decade?
> >
> > One, because we can. We can do it by masking a single character. Two, since most of the world uses a decimal system, the decade, as a reference point, just as the century (though less well-defined), have become popularly used reference points. If nine had been the chosen radix for our number system rather than ten (which probably would have been a beter choice) then we would be distinguishing nine year and 90 year intervals instead.
> >
> > If you have no requirment to represent a decade that's fine, nor do I, but if Ed does - and he is a stakeholder in this process - we should find a way to support this requirement.
> >
> >
> >
> > I am trying to understand your position on the meaningfullness (or lack) of a date, and take it that you don't think that the representation of days, years, etc. are meaningful (and thus not worth standardizing) unless they are "aligned to common chronological divisions". I confess that I may not fully understand what you mean but I take it to mean that a year, say, 1984, is not sufficiently defined without a precise start and end point including a time zone and a calendar designator. Please forgive me if I am misrepresenting your position.
> >
> >
> >
> > What we are trying to do here is build a specification that meets needs which range from simple to complex, with solutions whose complexities are proportional to the complexities of the needs. Well that goes without saying, that's what standards are supposed to do, but my point is, this work started off as an effort on behalf of data entry people who want to do things like enter a year of publication into a form, a year like 1986, and "1986" has quite enough specificity for their needs. Those users represent a large part of the constituency of this effort. We also want to satisfy more complex scientific requirments if possible but not by sacrificing the simple solutions for simple requirments.
> >
> >
> >
> > --Ray
> >
> >
> >
> >
> >
> > From: Discussion of the Developing Date/Time Standards [mailto:[log in to unmask]] On Behalf Of Simon Grant
> > Sent: Tuesday, December 21, 2010 4:30 PM
> > To: [log in to unmask]
> > Subject: Re: [DATETIME] precision and decade
> >
> >
> >
> > This discussion seems rather shaky to me, because it seems to build too much on weak foundations.
> >
> >
> >
> > I maintain that, at the level at which it is of interest to us, there is no coherent concept of a particular day/week/year/decade/century as an entity in its own right, and that therefore there is no point (and that's on the generous side) to elaborating this.
> >
> >
> >
> > What does make common sense (and is therefore worth elaborating if possible) is
> >
> > (a) that an event happened within a given time interval, and for ease of reference (only) the boundaries of that time interval may be aligned to common chronological divisions
> >
> > (b) that there is a set of events (constrained in some other way, as it is impossible to catalogue all events in any time interval) that happened within a certain time interval -- as (a), for convenience common boundaries are often used. This may be what people have in mind when they refer to a day, week, month, year, decade, century or whatever. I maintain it is the only sense in which those time concepts can be meaningfully reified. Note, however, that those with different calendars will be grouping together different sets of things when they say, for instance, "that was a hard month".
> >
> >
> >
> > There is nothing universal about any particular time interval. Why prioritise a decade in a decimal calendar over some other interval of years in some other calendar? The closer I consider things, the more arbitrary a decade or a century appears. Hence my preference (and willingness to argue) for a primary, general system of points and intervals, and only secondary and less important (and only if trouble-free) some notations for selecting ranges of years etc. etc.
> >
> >
> >
> > The precisions mentioned here are fully dependent on the time system and calendar, and therefore only "scientific" inasmuch as our current science uses these units and this numbering system. I do *not* believe they are worthy of standardization.
> >
> >
> >
> > Simon
> >
> > On 21 December 2010 19:37, Denenberg, Ray <[log in to unmask]> wrote:
> >
> > Rereading the recent (and not so recent) discussion I'm trying find a way to move this along, particularly the issue of precision. Ed seems to be the one most interested and he said:
> >
> >> At the heart of things I also don't want us to confuse dates with
> >> intervals.
> >> If I say something occurred in the 1960s I don't want to have to use
> >> intervals
> >> just as I don't have to use intervals to talk about 12 Sept 1933 (which
> >> is
> >> again saying something different from 1933-09-12T00:00Z/1933-09-
> >> 12T23:59Z).
> >>
> >> I suggest we in generally have the following precisions:
> >> - second
> >> - minute
> >> - hour
> >> - day
> >> - week
> >> - month
> >> - year
> >> - decade
> >> - century
> >
> > We have the precision Ed seeks for: second, minute, hour, day, week, month, and year.
> >
> > Which leaves decade and century. Century is a separate discussion unto itself. I will treat that in a separate thread.
> >
> > So let's just talk about decade for the moment.
> >
> > Ed supports the 'x' approach: where we let 196x mean the decade, 1960s.
> >
> > There are reasonable arguments against this, but I'm willing to go along with it if it will move us forward. We're only talking about this for decade (and possibly century).
> >
> > Comments, please.
> >
> > --Ray
> >
> >
> >
> >
> > --
> > Simon Grant
> > +44 7710031657
> > http://www.simongrant.org/home.html
> >


>
>

> --
> Simon Grant
> +44 7710031657
> http://www.simongrant.org/home.html




--

Edward C. Zimmermann, NONMONOTONIC LAB
http://www.nonmonotonic.net
Umsatz-St-ID: DE130492967