[
Lists Home |
Date Index |
Thread Index
]
"Thomas B. Passin" wrote:
> At first glance, this suggests that it is not __xml__ that needs a type
> system,
This is the fundamental truth of XML. However, it is my perception that over the
past two years, but especially most recently, the discussion on xml-dev is
motivated increasingly by the refusal to accept that premise. The types *of* XML
are those nominated in GIs. They are entirely lexical. In their nature is the
essence of XML for they are, entirely on their own terms, extensible, markup,
and the basis of any language which is given XML expression.
> but the infoset (maybe), schema, transformation, and query languages.
Undoubtedly these have that need, and for that reason are outside what *is* XML,
though they are entirely legitimate candidates for what is *done* with it.
> But there is a confounding factor, and probably that is what has lead to so
> much confusion and complexity.
>
> It is not enough to say "let this thing be an amy:duration". We also need to
> know "this thing in such and such a location, but not in that other location,
> must be an amy:duration". So the value types easily get mixed in with the
> element types, and bingo, the type system has become mixed up again with the
> xml, which we hoped to avoid.
Isn't preserving this distinction, while mapping from within the XML instance to
an appropriate datatype-based processing external to the instance, the function
of NOTATION?
> I think it will take some cleverness to see keep the type system from getting
> mixed up with the structure, but I think it will be worth it.
> Maybe a step towards keeping the type system out of the xml is to keep it out
> of the infoset too. Then the types get applied only after infoset creation,
> not during it.
Hear, hear. I believe that this is consonant with the original understanding of
the Infoset, as the lightest of abstractions from the lexical instance.
Respectfully,
Walter Perry
|
