[dsdl-discuss] Re: Numeric precision

Francis

Thanks for responding to my proposal re:

> > <!ELEMENT Number (#PCDATA)>
> > <!ATTLIST Number
> > Type (Complex|Real|Rational|Integer) Integer
> > Precision (16bit|32bit|64bit|unlimited) 32bit
> > Signed (yes|no) yes
> > Accurate (yes|no) no >
>
> How are you going to represent a complex number in a single data element?
> Don't you need sub-elements for the real and ocmplex parts? The
alternative
> is that you adopt one of the various conventional notations for writing
> complex numbers, in which case you'll either alienate someone by not
> selecting their favourite notation or you'll have to introduce a Notation
> attribute. I'm not convinced that Complex should be there at all.

I'm not convinced is should be there either. I was presuming that we would
define a single notation within the standard. (Others could be defined for
other namespaces of datatypes.)

> What about Orthography (this might be covered by xml:lang, but I have my
> doubts) and Base as additional attributes? Or are you only talking about
> base-10 numbers using 'arabic' digits?

So far all the datatype proposals I have seen have a built in presumption of
base-10 for numeric data, and arabic digits within the ASCII codes set part
of ISO 10646. I certainly am not claiming that any numeric datatype
introduced into DSDL should be able to cope with all bases and
orthographies. As far as I am concerned our aim is to define a minimum set
of datatypes to enable the development of schemas for which DSDL can
guarantee interoperability, not to define a set that covers all needs of all
communities.

Martin Brayan

--
DSDL members discussion list
To unsubscribe, please send a message with the
command  "unsubscribe" to dsdl-discuss-request@dsdl.org
(mailto:dsdl-discuss-request@dsdl.org?Subject=unsubscribe)