Print

Print


Hi Betsy, thanks for your comments and welcome to our forum. 

> 1.) I�m not clear about item #2 under Standard Features.� The label
> says �Date/time - without hyphen/colon.�� However, the example uses
> hyphens:
> 2001-02-03T093001

That was an error, thanks for catching it, I've removed the hyphens.

> 2.) Hyphens:� The option express to a date without a hyphen is
> appealing. Without the hyphen, the date looks elegantly
> simple.� However, this option creates too much complexity.� Systems
> will possibly have an extra format to parse.� It also leads to possible
> data errors.� How to remember it is okay to use 19990415 to represent
> year, month, and day, but not okay to use
> 199904 to represent a year and month?

I don't think there is any way around this problem though I would welcome
further discussion.  I do not think we could get agreement to profile out
the separated forms (hyphens in dates and colons in times) because that's
the form that W3CDTF uses and I don't think this spec is going to go
anywhere if it does not accommodate W3CDTF.  And on the other hand, part of
the motivation for beginning the work on this spec was to add the
non-separated forms. I think we're stuck with this complexity.


> 3.) Further complicating the hyphen issue, it looks like section 3 only
> permits questionable dates to be represented without the hyphen:
> 20040611? Is 2004-06-11? permissible?� I�m not seeing an example
> formatted that way.

Yes, there's no intention to restrict "questionable" (etc.) to unseparated
values.  The description of the '?' etc. characters' syntax is that "they
apply either to the entire portion of the string to their left, unless
immediately preceded by parenthesis, in which case they apply to the portion
of the string delimited by parentheses."  The key being "the string"
whatever that string may be.  The examples are not intended to be
comprehensive. The spec at this point is a tool to achieve a stable
consensus on the proposed standard, and then a more formal spec perhaps
including a BNF definition will need to be written. 


> 4.) Unspecified years, days, and months can be represented with the
> letter u.� From the examples: 199u, 1999-uu, 199901uu.� However, when
> only the century is known, the letter u is not used to fill in the
> missing digits: 19 is used for sometime in the 20th century.� It seems
> like if we�re going down the path of using the letter u, we might as
> well use it in the century coding (19uu) as well in order to be
> consistent.� 

Actually, no, '19' would not mean "SOMETIME in the 20th century" it would
mean "THE 20th century".

There has been a great deal of discussion of this and surrounding issues but
let me start out by saying that the only reason why in our spec '19'
means"the 20th century" is that that's what it means in ISO 8601, and one of
the principles of this work is that whatever we want to express, if it is
expressible in ISO 8601 then this spec should at least specify that syntax
even if it perhaps specifies alternative syntaxes not specified in ISO 8601.


The case of century is particularly troubling because we cannot even agree
what it means. Some of us believe that the 20th century began in 1901 and
others believe it began in 1900. ISO 8601 doesn't help resolve this because,
even though it tells you how to represent a century, it doesn't tell you
what a century is.  And it is fairly clear to me that we are not going to
agree to a definition and are going to have to leave it undefined. 

But there is a way to represent the interval consisting either of the years
(1) 1900 through 1999, or (2) 1901 through 2000:

(1) 1900/1999  or 1900--1999
(2) 1901/2000 or  1901--2000

So I think we should keep '19' in the spec to mean "19th century"
(undefined) for reasons I stated above, with the general guidance that if
you have a specific interval you want to represent you'd be better off using
one of these two forms.  If on the other hand you want to designate during
what century World War II occurred, '19' is probably good enough. 

But I digress from the point you raised. The spec would indeed support
'19uu' to mean "some unspecified year between 1900 and 1999".   If what you
want to say is the event occurred "sometime during the period between 1900
and 1999" -- it is a matter somewhat outside the scope of the spec to say
whether 1900/1999 means
(1) an instant in time during that period, or 
(2) the entire interval

That would be a semantic matter for the applcation which references the spec
to specify.

I hope this helps. Thanks again. 

--Ray