W3C home > Mailing lists > Public > public-rif-wg@w3.org > December 2005

[UseCase] HP-1 Message transformation

From: Dave Reynolds <der@hplb.hpl.hp.com>
Date: Fri, 02 Dec 2005 11:27:48 +0000
Message-ID: <43902FB4.5040405@hplb.hpl.hp.com>
To: public-rif-wg@w3.org

[This is the first of a series of messages in response to the request for 
example use cases.  It seemed best to put the use case examples in 
different messages but tag them. There is no significance to the ordering 
or naming convention.

These are quick sketch examples, not detailed, considered, formal use
case proposals.

Feedback on style and grain size as well as content is welcome. I've
kept them relatively abstract rather than use a more narrative style,
just a matter of taste.


** HP-1   Message transformation

* Outline:

There is an IT management eco-system in which a number of management
components interact over a service oriented architecture. One vendor's
management component (A) exports a model of the resources it manages
as an RDF instance model (IA) according to ontology (OA). A second vendor's
management control desk (B) requests the resource model in it's ontology (OB).

A mediation component (may be a separate service within the SOA or a
sub-component of either A or B, irrelevant for current purposes) performs 
the translation of instance model IA to an instance model IB according to a
publishable set of mapping rules.

In this use case the models being translated have a natural boundary,
they may have been derived from multiple web based sources but the
translation occurs on a defined bounded document or message.

* Implications

In practice many of the required mappings are simple
(renaming of classes and properties, transformation of literal values,
composition of properties, aggregation of literal values).

Three less trivial requirements arise:

(1) IA and IB are often not isomorphic and the mapping can require
object introduction. For example IA may represent a computing node as
a single resource with properties such as network interface address,
whereas IB may represent the compute node and its network interface as
separate resources. In RDF terms this may be handled through the use
of bNodes but might also require algorithmic construction of new
resource URIs.

(2) IA may not be sufficient to completely instantiate IB and the
mapping sometimes requires the introduction of defaults. This
requires negation over the extensional data in the closed model IA
(not necessarily negation over intensional predicates).

(3) Some mapping rules require quantification over predicates. For
example OB may include a notion of "dependsOn", predicates in OA may
be marked with annotations to indicate if they induce dependency
relations. Or a resource in IA may need to have all of its attribute
values "copied" across to a corresponding but distinct resource in IB,
where the set of attributes is open ended.
Received on Friday, 2 December 2005 11:28:19 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:47:36 UTC