- From: Xavier Franc <xfranc@online.fr>
- Date: Fri, 19 Dec 2003 23:29:35 +0100
- To: public-qt-comments@w3.org
Bernhard Bodenstorfer wrote: > The origin of most problems with durations is that > they depend on a reference dateTime so that they can > be unambiguously interpreted. > The origin of all problems with durations is that durations (I mean xs:duration) are conceptually lame! As long as a duration involves only seconds, minutes, hours, and days it is OK, it is just a syntactically equivalent way of representing a numeric duration in seconds (seconds are the legal unit for time). As soon as a duration involves months, it is no more OK, because a month is a fuzzy concept, something that lasts 28, 29, 30 or 31 days. And you are right, it is not possible to know the exact duration of a month without a reference date. [I feel I am just repeating obvious facts.] Then a sane approach of the question should be to avoid manipulating xs:duration as much as possible and to perform all computations with numeric durations in seconds. A complementary requirement is the capability to convert durations from and to numeric values, and to provide duration/date/time formatting functions. As I understand it, this is the point of view upheld by Michael Kay, and it seems sensible. On the contrary, the F&O WG has chosen to provide an arithmetic based exclusively on duration types. This seems to be based on considerations of compatibility with legacy languages like SQL (which has no less than 4 different types of durations!) Since xs:duration has proved intractable, the WG has gone further in the direction of complexity, and introduced two new subtypes supporting complex and rather odd arithmetic operations. [By the way this replacement of xs:duration by new types invalidates any argument of compatibility with legacy data...] Meanwhile, several issues remain virtually unaddressed in XPath/XQuery: - how to convert duration types to numeric values usable in 'real world' calculations (financial, scientific) - how to construct duration/time/dates from numeric values (this is very important for real calculations and also for converting durations/times/dates from localized formats) - how to format durations/times/dates in misc. formats I think it would be quite valuable to collect a number of real-life use-cases and see how they are solved by the current specifications or outsider proposals (that is, to write actual code snippets and see how they look). -- Xavier FRANC
Received on Friday, 19 December 2003 17:27:41 UTC