[
Lists Home |
Date Index |
Thread Index
]
Uche Ogbuji <uche.ogbuji@fourthought.com> wrote:
|> http://www.w3.org/RDF/Group/1998/02/NOTE-rdf-ns-reqs-0209
|
| So it looks as if this last link is the pertinent citation.
| [... 403 Unauthorized ...]
| I swear to the goddess that the W3C makes me want to spit up a flood,
| sometimes.
You are in luck. I found a copy quite accidentally.
====
NOTE-rdf-ns-reqs-0209
XML Namespace Requirements for RDF
W3C Note 09 Feb 1998
This Version:
http://www.w3.org/RDF/Group/1998/02/NOTE-rdf-ns-reqs-0209
Latest Version:
http://www.w3.org/RDF/Group/1998/02/NOTE-rdf-ns-reqs
Previous Version:
@@ or versions. hmm...
Editors:
Ralph Swick <swick@w3.org> W3C
Status of This Document
This document is generated by the Resource Description Framework Model and
Syntax Working Group as input to the XML namespace design activity.
Further
discussion on this document may be found on the Members-only working group
mailing list.
--------------------------------------------------------------------------------
Preface
RDF defines a data model for describing properties of Web resources.
The canonical representation in RDF of an instance of such a data
model is a directed labelled graph. In RDF the types of the nodes and
the arc labels within a single instance of a DLG may be defined by
independent parties without any coordination. RDF requires a
mechanism for labelling arcs that distinguishes labels created by
independent parties. RDF wishes to be able to use XML to encode and
exchange these DLGs. RDF therefore requires a mechanism in XML that:
1. Allows every element name to be assigned a globally unique identifier.
2. Allows every attribute name to be assigned a globally unique
identifier.
3. Accomplishes 1 and 2 without using a central registry of identifiers.
4. Allows for the definition of mappings from globally unique identifiers
to resources on the WWW.
5. Permits multiple namespace associations to be active simultaneously.
6. Allows the unique qualifiers to be written and read without having seen
a DTD.
7. Allows the attributes within a start tag to have namespace qualifiers
different from that of the element in that tag and different from other
attributes in that tag.
With these base requirements in mind, we submit the following comments on
Bray, Hollander, Layman "Name Spaces IN XML". The section names that
follow refer to the corresponding sections of that document.
Abstract
This section introduces the term "schema". Since the rest of the proposal
gives no particular semantics to the URI in the namespace PI (and we
support leaving this open at the level of the XML specification), we
wonder why it is necessary to introduce this term in this specification.
The generic phrase "defining document" has been used in discussion and its
use here will cause less confusion with independent activities to define
various forms of schemas.
1. Motivation and Summary
First paragraph:
We like the intent of this paragraph but we note that since "schema" is
undefined this paragraph has no meaning when separated from the context of
the mailing list discussions. We think the intent could be made clear
without having to appropriate this term, thereby leaving it available for
future specification.
Fourth paragraph:
We heartily endorse the use of URI as the mechanism for producing globally
unique identifiers. This satisfies requirements 3 and 4. In addition, it
is important that any legal URI be useable within a namespace declaration;
XML must not place limitations (such as character restrictions) on the
URI.
2.1 Declaring Namespaces
We fully endorse the principle that "some namespaces need no schemas; this
specification does not depend on their existance". This satisfies our
requirement 6.
Is it the intention that when 'href' is omitted the value of 'name' be
used to (attempt to) retrieve the defining document? In other words, if
the document would otherwise place the same value in both 'name' and
'href' would you encourage or discourage omitting 'href'? We wish to
encourage RDF applications to provide the defining document (in a form to
be specified by RDF). RDF "best practice" will be to provide both URIs,
which we expect to most often be identical.
2.2 Placing Declarations in Documents
First paragraph:
We think it would be a better design to have non-global scoping of
namespace declarations but in our discussions to date we have not
identified a specific RDF requirement that would absolutely preclude
global declaration scoping.
Without local declaration scoping it will be necessary on occasion to
rewrite portions of documents when combining them from existing pieces.
The namespace facility should be designed so that rewriting is not
required. Rewriting adds processing cost to applications, including the
requirement to buffer an entire document prior to transmission.
We anticipate that there will be applications of RDF that will appear with
documents as strictly localized markup in a small number of places.
It seems advantageous to us to keep the resulting textual change to the
document localized rather than splitting between the document prolog and
the location of interest.
We note that requiring all namespace declarations to appear in the
document prolog makes the embedded editor example of 3.1.1 more
difficult. The editor must report its namespace to the enclosing shell
editor separately from its content rather than using a locally scoped
declaration that is part of the content.
We further note that one recommendation coming from HTTP performance
studies is that document authors place important information in the first
few hundred bytes of the document. What constitutes "important" will be
application-dependent. XML should avoid as much as possible adding
further constraints on the author's ability to optimizing the content for
a specific purpose.
One possible local declaration scoping mechanism would be to introduce a
new tag; .... The namespace name introduced by this tag would be defined
only within the content of the tag and would override any definition using
the same name in outer content. See also Dave Raggett's suggestion of a
new global attribute, referred to in "Open Issues" below.
Well-Formedness Constraint - Unique Namespace Names
We see no purpose for this constraint if non-global scoping were
introduced.
2.3 Qualified Names
Well-Formedness Constraint - Namespace Name Declared
We endorse this constraint and note that it supports our requirement 6 by
removing the possibility of implicit namespace names other than 'xml'.
2.4 Using Qualified Names
Permitting both element types and attribute names to be qualified names
satisfies our requirements 1 and 2.
Furthermore, it is important that each attribute be able to be qualified
independently; see our requirements 5 and 7. This proposal as written is
sufficient for that purpose.
4. Issues Open for Discussion
Third Paragraph:
In RDF usage, attributes will not naturally reside in an element-specific
namespace. Attribute names will be global in an RDF schema just as
element names are global. More concretely, RDF defines a precise
equivalence between names used as elements and as attributes. This
is different from the usual SGML interpretation where attributes are
inherently qualified by the element name(s) with which they are used.
A simple minimization construct would be useful, though the consensus of
the RDF Model and Syntax WG is that namespace minimization is not a
requirement for the first version. Dave Raggett has suggested one
possible form (a global 'using' attribute) that combines both local
namespace declaration and minimization.
We don't know what other kinds of constraints on qualified attributes the
XML WG might be considering (as suggested by this paragraph); in general
we feel that XML should not constrain the use of qualified names.
4.1 Which Names Should be Subject to Qualification
The RDF Model and Syntax Working Group has not discussed a consensus on
use of qualified names for purposes other than elements and attributes.
We have discussed scenarios in which it would be useful to have qualified
values. It is likely we will want XML to make it clear that the data in
the namespace declaration must be available to the application for its own
processing purposes.
4.2 Validation
RDF Schemas will define constraints for correct RDF expressions
well beyond the constraints that XML DTDs are capable of encoding in a
practical way. We would like to clearly separate the concept called
"schema validation" in 4.2 from the concept of XML Validity. To make this
distinction clear we encourage the introduction of a third orthogonal
concept, for which we propose the term "correctness". Correctness refers
to semantic consistency with respect to one or more schemas. For example,
the statements "George is my biological father" and "George is my son"
might be always well-formed and always syntactically valid but not correct
when "George" binds to the same object.
Any document may be well-formed and correct (with respect to all the
included namespaces) without ever being tested (or testable) for
validity. In this sense, correctness is not a special case of
validity as suggested in the first paragraph.
4.3 Qualified Names vs. Reserved Attribute
Third Paragraph:
We endorse the decision that XML leave it up to the application to choose
whether and how to combine the URI,LocalPart pair and how to interpret
such a combination. RDF intends to define how to combine the namespace
URI and the LocalPart to access further machine-understandable data for
the corresponding element or attribute. Refer to our requirement 4; we
suspect that different applications will wish to define different
combination algorithms.
However, if it is decided that a single algorithm is needed for use
by all XML applications we believe that it should specify at most
the simple concatenation of the URI with the LocalPart.
--------------------------------------------------------------------------------
Ralph R. Swick Last updated: 1998-02-09T15:01Z
|
