RE: XML message format draft

[Graham Klyne <GK@xxxxxxxxxxxxxx>]

At 11:45 AM 2/7/03 +0000, Jon Hanna wrote:

> > As for time format, I'd like to move to a generic ISO8601-based form per
> > RFC 3339 [1] or XML schema [2].
> >
> > #g
> > --
> >
> > [1] http://www.ietf.org/rfc/rfc3339.txt
> >
> > [2] http://www.w3.org/TR/2001/REC-xmlschema-2-20010502/#dateTime
>
> Agreed.
>
> I would suggest that we define a datetime which:
>
> 1. Is a subset of those allowed in the two works cited above, and also in
> NOTE-datetime <http://www.w3.org/TR/NOTE-datetime> (commonly referred to as
> W3CDTF), and as such also compatible with Dublin Core's W3CDTF (which is
> just a reuse of the former, with a DC-supplied URI).

I think RFC 3339 satisfies that goal, when used with one additional 
restriction:  the letters T and Z MUST be in uppercase.  (Note:  RFC3339 
and the W3CTDF were both derived from the same original.)

> 2. Had the same range and precision as that used in RFC2822.

You mean not including sub-second resolution?   Hmmm... maybe.  Treating a 
Time element as corresponding exactly to RFC2822 then that makes 
sense.  Then we'd have to use a different element if we really wanted to 
express times with sub-second resolution.

I'm undecided.


> 3. Was always expressed in terms of UTC/GMT.

Er, in a sense, that requirement conflicts with your 2.   RFC2822 
explicitly allows (and, I think, encourages) use of non-UTC 
timezones.  RFC3339 allows just Z or a numeric offset, so normalizing to 
UTC should always be easy to do.


> As such we would have a datetime format expressed as the following ABNF:

[...]

I'm reluctant to create yet another definition of date-time format.  The 
work Chirs Newman and I did on RFC3339 was intended to avoid just that 
outcome.  If some changes are really needed, I think it would be better to 
specify

#g



-------------------
Graham Klyne
<GK@xxxxxxxxxxxxxx>