[dsdl-discuss] Re: FW: DSDL: Physical validation

With reference to Part 9:

> * The document must use namespaces sanely (as defined in
> http://lists.xml.org/archives/xml-dev/200204/msg00170.html)

As far as Part 9 is concerned, the intention is that all namespaces will be
globally declared and so are in scope for the whole DTD. We should perhaps
consider making it mandatory that no two prefixes be assigned to the same
namespace URI.

> * The document must associate the namespace
> "http://www.w3.org/1999/XSL/Transform" with the prefix "xsl"

I presume this is part of the 'reserialisation' issue connected with the
first point. If we insist upon a one-to-one relationship between prefix and
URI, this requirement becomes easier to enforce.

> * Namespace declarations must be placed on particular elements (this
> requirement is particularly relevant when mixing namespaces and DTDs;
> see eBooks as an example)

I hadn't envisaged there being a mechanism in Part 9 for insisting that a
namespace declaration be placed on a particular element, but perhaps we
should consider adding this feature.

> * Content of the <script> element must be escaped with a CDATA section

Not a DTD issue.

> * The <foo> element must be inserted using the &foo; entity

Not a DTD issue.

> * Specific characters must be represented using particular entities
> (e.g. non-breaking spaces as &nbsp;)

Another reserialisation issue? In the general entity case this is surely not
achievable, but in SGML applications SDATA entities are generally
reserialised as entity references...? Of course, in the general SGML
application one can restrict the document character set (in the SGML
declaration) to exclude characters that you want to force the user to encode
as SDATA entities.

> * Specific characters must be included as native characters rather than
> as entities or character references

In an SGML DTD one can specify the content of an element to be CDATA, i.e.
not contain any markup characters, in which case neither entity references
nor character references are recognised.

> * Character references must use the hexadecimal version rather than the
> decimal version of the character's codepoint

Not a DTD issue.

> * Particular elements, when empty, must be represented with a start and
> end tag rather than an empty tag (and vice versa).

Not a DTD issue?

> Of course, there's scope for specifying other physical
> constraints, such as
> which kind of quotes should be used around attribute values; what
> whitespace
> is allowed within tags or in element-only content; and so on.
> A physical schema language would have to decide which physical features to
> constrain.
>
> Is there any room in DSDL for such a schema language?

Many of the above cannot, as I see it, be specified by any of the schema
languages that we currently have in DSDL, but whether any or all of these
are within scope I'm uncertain.

Francis Cave

--
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)