RE: XSLT applied to RDF?

The following is the abstract of a presentation that I was supposed to give
(entitled "Using RDF for Data") at XTech 2000 last week.  I was told at the
last minute by our legal department that I couldn't give the presentation
because we are in the "quite period" before going IPO.  However, I am
attaching below the abstract of the talk, since it has already been made
public.

Paragraph 6 briefly describes the canonical form of RDF/XML that we use
(fully conformant to RDFMS 1.0) and how we are using XSLT to tranform that
RDF/XML into DHTML.


Samuel C. Yang
Senior Software Architect
Opus360 Corporation - West Coast Office (formerly PeopleMover, Inc.)
1500 Rosecrans Avenue, Suite 310
Manhattan Beach, CA 90266

Putting People and Projects Together (sm)

Ph: 310-819-1900 x127
Fx: 310-819-1901
Email: syang@peoplemover.com <mailto:syang@peoplemover.com>
Web: www.peoplemover.com <http://www.peoplemover.com>

========================================================================


This presentation will show how PeopleMover has successfully used RDF
throughout our n-tier thin client framework to represent, transport, and
manipulate all data, not just metadata.  I will also present information
about our RDF Java library, and describe the features that make it ideal for
those purposes.

The RDF data model is essentially the same as the relational data model.  We
are able to automatically generate a RDF schema from a DB table's metadata,
and then automatically map any row in that table to an equivalent RDF
resource description that conforms to the generated RDF schema.  We are also
able to go in the opposite direction to create, update, or delete rows in
the DB using RDF data that conforms to the generated RDF schema.

Each RDF resource (r) that corresponds to a row in the same DB table (T) has
the same generated URI namespace.  However, the local part of each r's URI
is different, and is based on the primary key value of the corresponding row
in T.  Also, we map each column name in T to a unique RDF property name
shared only by the r's that represent rows in T.

Whenever the DB model changes, we use a DB-to-RDF metadata tool to generate
a new equivalent RDF schema, and store it in an RDF/XML document.  The
generated RDF schema contains metadata to describe every table, every
column, and every relationship.

At system startup time, a DB metadata service reads that RDF/XML document
and caches its information in memory.  Another DB service can then accept
requests with RDF/XML arguments, and perform the appropriate CRUD (create,
read, update, or delete) operations on the DB with the help of the cached DB
schema.

We also use XSLT to transform RDF/XML into DHTML.  However, to ensure that
we can do so, our RDF library generates RDF/XML that has a (perfectly legal)
RDF syntax that satisfies a simple constraint.  That constraint is: a RDF
resource must be physically represented "one-level-deep".  That is, any
property value that refers to another resource must do so using the
rdf:resource attribute.  Resources cannot physically embed other resources.
This constraint is required for at least two reasons: (1) this ensures that
any RDF graph can be represented as an XML tree.  All cycles in the RDF
graph are physically broken, but are logically maintained through URI
references.  Without this constraint, a cycle can cause a tree of infinite
depth.  (2) XSLT stylesheets do not have to "understand" the full RDF/XML
syntax, and can make very simple assumptions about the structure of the
RDF/XML document: (a) Every RDF resource is an immediate child of the
RDF/XML document's document element; and (b) A property's value is given by
either literal (i.e., purely textual) content, or by a r:resource attribute.

Our RDF library, PmRdf, is written in Java, and uses the collections library
(either the JDK 1.1 or 1.2 versions) to efficiently store and handle the RDF
data.  PmRdf parses RDF/XML documents using the W3C SiRPAC parser, and can
therefore "import" RDF/XML documents that are not "one-level-deep" into a
PmRdf 
object.  However, the library always "exports" the data from a PmRdf object
into RDF/XML that is "one-level-deep".  An instance of any resource or
property can be obtained in constant time by providing its URI.  PmRdf has a
very extensive API, and also supports RDF containers.  Although designed to
be a generic RDF API, it has proven to be very easy and efficient to use for
our specific data manipulation needs.

Sam


-----Original Message-----
From: alison@cee.hw.ac.uk [mailto:alison@cee.hw.ac.uk]
Sent: Thursday, March 09, 2000 1:38 AM
To: caro@Adobe.COM
Cc: www-rdf-interest@w3.org
Subject: Re: XSLT applied to RDF?


> 
> Has anyone toyed with applying XSLT transformations to RDF to generate
> alternative representations?  For example, the simplified "Just The

I toyed with this last summer, to eliminate alternative abbreviated
syntaxes to make subsequent use of XSLT to present (simple) RDF simpler.
(Written up as paper for WWW9). I started writing a "normalising"
XSLT stylesheet for RDF, but got bored/ran out of time. For my purpose
I needed a canonical non-abbreviated syntax to make
presentation/transformation
of non-parsed RDF simple.

		Alison

Received on Thursday, 9 March 2000 15:01:01 UTC