Database Server Models for WMS and WFS Geographic Web Services - Geographical Information Systems: Trends and Technologies - page 160

this link. A value depends on the physical implementation of the database.
For example, a specialization can be translated into a set of relational joins
or a set of projections and unions.
We emphasize that the goal here is not to propose a standard solution
like the OGC Table Joining Service (TJS) (OGC 2010a). TJS offers a method
of data research, access and use from multiple sources and, dynamically, to
supply the databases, to perform analysis. Indeed, we use the same schema
of database to determine the relevant attributes.
Synthesis
The concept of relevant link (RL) presented in this section widely extends
the functional dependence. They take into account more semantics such
as the ones that may exist between attributes and between attributes and
methods. The main advantage of this proposal lies on semantic defi nitions.
Information will be processed with links of relevance and not only from
their graphical representations.
Figure 7 presents the relationships between the various concepts of our
meta-model formalized here in UML.
This diagram indicates the following rules:
• A class can have the role of source class and/or of target class within
a RL.
• A class is composed of attributes, methods, and possibly RL.
• An attribute can be spatial or alphanumeric.
• An attribute can have the role of a source attribute or of a target attribute
within a RL.
• An attribute has a single type.
• A method has a list of parameters and a return value defi ned by a
type.
• A parameter can be an attribute or the result of a method.
• A constraint (pre-conditions and/or post-conditions) may be associated
with an RL.
• A RL can be typed by RLintA-A, RLintA-M or RLextA-A.
• A RLintA-A establishes a link between two sets of attributes of a
class.
• A RLintA-M is built with a set of attributes and methods of a class, the
parameter list of this method and its result. A typed attribute denotes
the result.
• A RLextA-A can bind two sets of attributes of two different classes.
• A RLextA-A does or does not involve a spatial query.