Lists Home |
Date Index |
At 09:51 AM 9/30/2002 -0400, W. E. Perry wrote:
>Jonathan Robie wrote:
> > I think there *is* broad agreement for integers, floats, doubles, strings,
> > etc. across programming languages and data representations.
>Yes, there is, just as there is 'broad agreement for' many other abstractly
>understood datatypes. What there is not agreement on in an XML-centric view is
>any particular instantiation of any of these types.
So far so good - and incidentally not any different from the traditional
relational world, in which various programming languages or tools could
instantiate an integer in different ways, with different capacities, or
cast the integer from the database into anything else the developer could
>That is, in fact, the point
>of text-based markup: accepting as a first premise that binary instantiations
>(let alone operations predicated upon those instantiations) are utterly local
>and shared only sporadically. Markup must therefore function as its creator's
>preferred identification of particular content. Where there is a contract in
>the form of a DTD or other schema, it is binding only upon the data
>chooses to identify and then to abide by it. From the data recipient's
>view, that contract is no more than the data creator's undertaking to present
>particular classes of data in a particular consistent form.
I'm not sure that I understand. If a data creator and a data recipient
agree to use a given schema, then the data recipient may reject any data
that does not validate according to the schema, on the grounds that it does
not conform to the agreement.
>What any data recipient does internally with the data it receives is invisible
>in theory to the data creator, and in my experience most often invisible in
>practice as well. What the data recipient can do with the data it receives is
>utterly dependent upon how that data is instantiated for use. In short, to
>up to some external datatype specification control of the specific local
>instantiation of data is to give up the possibility of uniquely local
>processing, and thereby unique local value added from that processing. Is that
>a reasonable expectation of a competitive business model which relies upon
>local, often proprietary, expertise?
Obviously, once a program gets ahold of data, it can do whatever it wants
with that data. Sometimes that may involve casting. I am not sure what
issues you are concerned about here.