Ann raises a couple of interesting issues:
1. The context for a given XML rendering of a particular semantic standard
2. Validation of semantics as expressed in a schema
In the first issue, I think perhaps there is some confusion about what OAI
is. Use of OAI protocols is a model for what I will refer to as
pre-coordinated agreements about semantics and syntax. It is distinctly in
the tradition of all parties agreeing about an idiom of description, and
sharing technological models for how to deploy those agreements.
In the OAI world you have the additional twist of supporting
discipline-specific metadata as well. So, the context for this schema
declaration is OAI, and others who might adopt a similar approach. This
seems both reasonable and useful, though probably not universal.
The second issue is thornier. Specification of semantics within a schema
does not confer the ability to validate adherence to that definition in
instance data. It DOES confer the ability to validate structure and syntax
of the metadata, but semantics has to do with assertions about concepts and
relationships, and validating the such assertions is beyond current
capabilities of machine processing.
This raises the issue of trust. Do we trust
[libraries|governments|museums|publishers|...] to be absolutely accurate in
all instances? No. Do we generally expect them to be truthful? In general.
In the Web, users have to make critical judgements about whose metadata to
trust in a given context. I think this is actually good news, in that the
metadata ecology will tend to evolve in favor of organizations that can
provide reliable resource descriptions.
stu
-----Original Message-----
From: Ann M Wrightson [mailto:[log in to unmask]]
Sent: Wednesday, April 24, 2002 9:55 AM
To: [log in to unmask]
Subject: Re: OAI and DCMI announce DC-Simple Schema Declaration
I would put this comment round a different way:
The schema provides XML element definitions for dc elements, without
providing an explicit or implicit context for their use. What usage in XML
documents is this intended to support? Is this schema intended for reuse
within local metadata standards? - if so, then our experience suggests that
providing XML Schema Type rather than Element definitions is more helpful,
since local standards are likely to restrict values permitted for elements.
We have drafted a metadata standard for XML schemas within the UK Government
GovTalk collection - this is based on the UK e-Government Metadata Standard,
which is itself based on Dublin Core. Having two levels of inheritance of
metadata element definitions involved (with specialization at each level)
brought out some interesting design issues in the XML representation - not
least, that the XML representation had to be a fairly uncomfortable
compromise between the semantics (intended meaning) of the metadata, and
what could be said easily in an XML schema to validate the metadata.
The document is expected to go out for public consultation soon on the UK
GovTalk website (http://www.govtalk.gov.uk) - consultation responses from
the dc community would be v. welcome.
Ann W.
Ann M Wrightson MA MBCS
Prif Ymgynghorydd / Principal Consultant
alphaXML Cyf/Ltd
http://www.alphaxml.com
Gwasaneuthau XML: e-Lywodraeth, e-Fasnach, e-Gyhoeddi
XML services to Government and Industry
-----Original Message-----
From: The broadest of mailing lists related to the international Dublin
Core effo [mailto:[log in to unmask]]On Behalf Of Renato Iannella
Sent: 24 April 2002 13:27
To: [log in to unmask]
Subject: Re: OAI and DCMI announce DC-Simple Schema Declaration
--On 24/04/2002 6:51 -0400 Weibel,Stu wrote:
> DCMI and OAI are pleased to announce an XML schema for unqualified DC
> metadata that facilitates the declaration of modular metadata components.
There is a small but significant error in this XML Schema.
No "root" element has been defined, and as the 15 DC elements
can only accept string values, you can only have one element
in your DC metadata instance.
(This has been verified using XML Spy V4.3.)
A new root/parent element, such as <dc-record> should be added.
For example:
<xs:element name="dc-record">
<xs:complexType>
<xs:choice minOccurs="0" maxOccurs="unbounded">
<xs:element ref="title"/>
<xs:element ref="creator"/>
...
<xs:element ref="coverage"/>
<xs:element ref="rights"/>
</xs:choice>
</xs:complexType>
</xs:element>
Also, there is a Group defined in the XML Schema (elementsGroup)
but is not used anywhere.
I noticed the use of the xml:lang attribute for all elements.
Did you consider adding an "encoding" attribute to support
the specification of encoding schemes (eg URI, W3CDTF, RFC1766)?
DCMES V1.1 mentions a few of these as "recommended best practice".
If so, you can simply add:
<xs:attribute name="encoding" type="xs:string" "use="optional"/>
to the "elementType" complex type.
Cheers...Renato <http://purl.net/renato>
Chief Scientist, IPR Systems Pty Ltd <http://iprsystems.com>
Open Digital Rights Language (ODRL) <http://odrl.net>
|