Lists Home |
Date Index |
Chiusano Joseph wrote:
> I am not saying that given sufficient _other_ information, that a
> program cannot so interpret the snippet ... for example, application
> level semantics.
> Yes - I was actually thinking of "human-level" semantics; the ability of
> a human to interpret the meaning of data based on its surrounding XML
That's called a _specification_. There are loads of these. Some are good,
some aren't. Nonetheless.
> On the other hand, the XML Schema specification does not provide
> such semantics.
> I beg to differ - and I'm sure it's because I am looking at semantics
> from a different standpoint here. I respectfully acknowledge our
> different viewpoints regarding semantics, and do not in any way think
> that yours (or anyone else's here) is incorrect. Just that I view the
> meaning of "semantics" very differently. I'll respectfully step aside
> from this debate, so as not to muddy the waters.
As I said. The XML Schema specification does not specify a mechanism by
which an XML Schema processor _alone_ makes use of <xsd:documentation>
" Annotations do not participate in ·validation· as such. Provided an
annotation itself satisfies all relevant ·Schema Component Constraints· it
cannot affect the ·validation· of element information items. "
Now certainly a human can read such <xsd:documentation> items and draw
conclusions as to the intended semantics of certain XML elements and
attributes. What I am saying is that such conclusions are not defined
_within_ the XML Schema specification, rather outside the specification.
Your use of the term "semantics" for human readers seems quite correct. My
point is intended more specifically for machine processable semantics, and
particularly which specification licenses which semantics -- I've started by
saying that XML 1.0 itself does not specify much in the way of semantics.
The intention is that specifications which are layered on XML 1.0 provide an
increasing degree of semantics. XML Schema does provide for simple datatype
semantics (e.g. numbers, dates -- to some extent) but most applications need
more. RDF and OWL provide another layer, though I expect that many
applications, for example those that use RSS, might need even more, and in
most cases we still need human level semantics i.e. readers of plain text.