On Fri, 27 Sep 1996, Weibel,Stu wrote:
> I'd like to see a bit of discussion on this point, but I guess I have to
> say it seems reasonable. For the purposes of reaching some consensus,
> would someone summarize the proposed version identification approach and
> make a formal proposal? Jon?
OK, here's a summary/formal proposal for the proposed version identication
approach. I've just written it off the cuff so there are bound to be some
boo-boos in this:
Proposal for the addition of Version Numbers to Metadata Packages
1. Introduction
The Warwick Framework introduces the concept of packages of metadata.
These packages can be Dublin Core, PICS, USMARC or any number of other
metadata formats. One problem that has been encountered with packages
already is how to differentiate between different versions of a single
metadata format when the format does not contain an explicit protocol
version number itself. Often different versions of the same generic
metadata format can be mutually incompatible, which can lead to problems
for software that is attempting to automatically parse the metadata. Such
software should at least have the option of being about to determine the
version information concerning metadata held in packages if it wishes.
Demanding that all metadata formats incorporate a format version number is
not possible. It would it mean that existing metadata formats that are
already deployed without version numbers could no longer be carried in
Warwick Framework style packages. It would also mean that the version
numbers would have to be placed in the same position within _all_ formats
in order to allow software to automatically extract and parse the version
number information. Therefore version numbers of metadata formats are
something that has to be encoded in the packaging information surrounding
the metadata itself.
2. Embedding Metadata Packages in HTML 2.0
Many people have expressed a desire to embed metadata packages into HTML
2.0 documents to allow them to supply metadata about the document itself.
The metadata is held in the META elements inside the HEAD of the HTML
document. The META element in HTML 2.0 has two attributes that are of
interest in this application; ``name'' and ``content''. At the W3C
Distributed Indexing and Searching Workshop, May 28-29, 1996[1] a general
purpose convention was agreed upon for embedding the metadata itself into
the document. This was:
<META NAME="schema_identifier.element_name" CONTENT="string_data">
where ``schema_identifier'' is a short character string identifying a
particular metadata format, ``element_name'' is the name of an attribute
within that format and ``string_data'' is a string of metadata which can
be interpreted within the bounds of that metadata format.
This proposal attempts to extend that format by surrounding the metadata
from a single package by two further META elements which delimit the
beginning and end of the package and carry information on the version and
full name of the package's metadata format. These two elements will both
carry a ``name'' attribute value of ``package'' and will have following
format:
<META NAME="package" CONTENT="(TYPE=type)(VERSION=version) format_name">
where ``type'' is either ``begin'' at the start of a package or ``end'' at
the end of a package, ``version'' is the version number of the metadata
format that this package of metadata conforms to and ``format_name'' is
the name of the metadata format. The ``TYPE'' part of the content
attribute is mandatory and must be present for all META elements with a
name attribute having the value "package". The ``VERSION'' part and
format_name are optional but recommended. If the ``VERSION'' part is
missing, processing software that handles the particular metadata format
held within the package should assume that the metadata conforms to some
well known version (presumably the one either in widest use or as backward
compatible as possible) or omit processing that metadata package
completely. The ``TYPE'', ``VERSION'', their values and the
``format_name'' should all be case insensitive.
An example of a package encoded in this format is:
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<HTML>
<HEAD>
<TITLE>Sex, drugz and rock'n'roll</TITLE>
<META NAME="package" CONTENT="(TYPE=begin)(VERSION=1.2.4) Foo Metadata">
<META NAME="foom.title" CONTENT="Sex, drugz and rock'n'roll">
<META NAME="foom.author" CONTENT="Jon Knight">
<META NAME="package" CONTENT="(TYPE=end)(VERSION=1.2.4) Foo Metadata">
</HEAD>
<BODY>
<H1>Sex, drugz and rock'n'roll</H1>
<P>Sex, drugz and rock'n'roll are what makes the world go round</P>
</BODY></HTML>
3. Dublin Core Versioning
It is further proposed for the Dublin Core package, a version number of
0.1 should immediately be ascribed to Dublin Core metadata that uses the
elements and sub-elements described in Paul Miller's Ariadne article
entitled ``Metadata for the masses''[2] and the list of known
sub-element identifiers and values[3]. This makes use of the original
thirteen elements that were agreed upon at the first Dublin metadata
workshop and the subsequent workshop in Warwick. This should be
considered as the basic version of the Dublin Core metadata format.
Future versions may expand or contract the number of elements or change
the way in which elements and sub-elements are formated. If no version
number is supplied for a package of Dublin Core metadata, it should be
assumed that the package contains version 0.1 Dublin Core metadata.
4. Security Concerns
This draft does not address any security concerns.
5. Acknowledgements
Thanks to <<your name here>> for their comments on earlier drafts
of this document.
This work was supported by grants from the UK Electronic Libraries
Programme (eLib), and the European Commission Telematics for
Research Programme.
6. Author
Jon Knight,
ROADS Project,
Department Computer Studies,
Loughborough University,
Ashby Road,
Loughborough,
Leics. ENGLAND
LE11 3TU
Email: [log in to unmask]
Phone: +44 1509 228237
7. References
[1] Stuart Weibel, "A Proposed Convention for Embedding Metadata in HTML",
report from W3C Distributed Indexing and Searching Workshop, May 28-29,
1996, <URL:http://www.oclc.org:5046/~weibel/html-meta.html>
[2] Paul Miller, "Metadata for the masses", Ariadne Web Edition, Issue 5,
Sept. 1996, ISSN 1361-3200,
<URL:http://www.ukoln.ac.uk/ariadne/issue5/metadata-masses/>
[3] Jon Knight and Martin Hamilton, "Dublin Core Sub-Elements", revision
1.10, 19th September 1996,
<URL:http://www.roads.lut.ac.uk/Metadata/DC-SubElements.html>
Tatty bye,
Jim'll
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Jon "Jim'll" Knight, Researcher, Sysop and General Dogsbody, Dept. Computer
Studies, Loughborough University of Technology, Leics., ENGLAND. LE11 3TU.
* I've found I now dream in Perl. More worryingly, I enjoy those dreams. *
|