Not
exactly. One has to decide when mapping out of an information system
of a different
type
using XML which properties one must preserve losslessly, and then the choice of
elements vs attributes is not arbitrary. I gave an example of that
in a different email
when
comparing object oriented design where fields can have objects to XML design
where
attributes cannot have elements. At first, these appear to be
incompatible, then
pushing the containment relationship of fields and objects to the
elements and the
description up one metalevel,
<object><field><object></object></field></object>,
one
can make it work. Going in the other direction and keeping the mapping
of
fields to elements introduces microparsing and hiding types inside the
attribute values. It is semantically messy, but that is the choice
some make
to
keep the XML encoding looking as much like the object oriented encoding
as
possible.
Loose,
yes, but not arbitrary. One does have to understand, for example, the
structural constraints of XML.
le
For instance, the choice between when to use
elements versus when to use attributes is essentially arbitrary.
|