RE: midnight in ISO 8601 (was: Re: Notes on the say-as note)

On Thu, 2005-06-02 at 01:17, Pawson, David wrote:

>     I suspect Mr. Pawson was misled by the restriction to the 
>     range 00-23 in the profile at 
>     http://www.w3.org/TR/NOTE-datetime -- this is a change 
>     vis-a-vis ISO 8601, not a reflection of a restriction made 
>     by ISO 8601.

> If W3C notes are misleading, then perhaps that is the item that
> needs clarification.

The Note is quite explicit, in general, that it is defining a profile
of ISO 8601, not simply reproducing it.  It nowhere says or implies
that 8601 limits hour tokens to the range 0-23.  Like most technical
writing, the Note does assume an attentive reader.  But I don't think
that that fact makes it misleading.

> My goal was regularisation.

Is regularization a useful goal in the context of the SSML 1.0 say-as
element?  

That element is intended for tagging material which uses existing
notations, not for the production of new material which might
plausibly be restricted to particular formats.  Given that existing
notations, including that of ISO 8601, use the number 24 as well as
the number 00 for midnight, a system intended to allow existing
notations to be described needs to recognize that fact, just as it
recognizes the fact that existing notations use 'A.M.' and 'P.M.' in
various spellings or the fact that many existing documents will use
date formats like mm-dd-yy and dd.mm.yyyy.

>     In the context of the SSML 1.0 say-as element, it seems 
>     clear to me that Eira Monstad is right to suggest that a 
>     restriction to the range 0-23 is unhelpful in the task of 
>     describing time expressions in unconstrained 
>     natural-language text, and that the range 0-24 would be 
>     more useful.

> If a specific time & day is needed, then 24:00 hours on day n is
> exactly equivalent of 00:00 hours on day n +1.

This is quite correct, and is indeed part of Eire Monstad's motivation
for suggesting a revision of the say-as Note.

If 'hours' is restricted to the range 0-23 in the lexical analysis of
time expressions, then the standardized forms of the attributes of the
say-as element are unable to recognize, describe, rely on, or exploit
the fact that 24:00 hours on day n is exactly equivalent to 00:00
hours on day n+1.  It would seem to be a consequence of this that if a
voice synthesizer relies on the say-as element and encounters forms
like "24:00" in existing documents, it will be unable to recognize
these forms as time-of-day expressions.

It's not clear to me why you believe that state of affairs is
preferable to one in which the formats defined in the Note on the
say-as element allow the synthesizer to recognize "24:00" as a time
expression on the basis of annotation using the say-as element.

best regards,

-C. M. Sperberg-McQueen
 World Wide Web Consortium