Moving Geospatial Applications Towards a Mission Critical Scenario
Database Oriented GIS
Early GIS were based on CAD tools that were simple mapping tool. The graphical
data were stored in files direct into the file system. In the 90s the Geo-relational
model dominated the GIS. This model is based on attribute data stored in
database systems and the graphical data stored in files. Until now only a few
vendors have specified a GIS with all data inside a RDBMS and provided libraries
for 3rd generation languages to make spatial queries and/or have provided a 4th
generation language for the same purpose.
A couple of years ago the leaders of RDBMS started to support spatial data types
and spatial indexes in a object-relational model approach. They also supported the
SQL3 standard that allows more complex queries to the database, including spatial
predicates.
Projects based on GIS that use proprietary databases to store graphical
information will have many problems trying to integrate the geospatial information
to the enterprise architecture. They will not benefit from the standards, gateways,
reliability and security that are given by a commercial RDBMS.
Another important issue provided by GIS based on RDBMS is the capability to
handle short-transactions and long-transactions in the same environment. Short
duration transactions are the most popular model that all non-spatial applications
are based on. For engineering environment, this approach is not enough as a
work on the field can usually take a few hours and sometimes even weeks. The
concurrency control of short-transactions is well known and is implemented by all
RDBMS. The locks policies are very robust to support several transactions at a
time because they usually take milliseconds. This concurrency control is not
appropriate for long-lived transactions. (Dias 95)
This capability is necessary because of the nature of engineering and operational
environment. While the engineering area may be designing and constructing the
network, the operational area may be doing assignments, changing customer
addresses and selling new services. All these tasks must be done on the existing
network, i.e., existing data on the database. (Magalhães 97)
The capability of constructing very complex data models is another important issue
that database oriented GIS possesses. Data model is the logical and physical
representation of the real world objects. These objects can be more or less
complex depending on what their representation and the granularity necessary to
the application are. Usually telecommunication network is the most difficult network
to be modeled because of the enormous quantity of equipment, information
necessary about them, the amount of vendors for each and the lack of standards.
Abstraction is highly recommended in this environment with such a diversity.
The object-oriented data model helps construct complex objects because of its
characteristics like inheritance, encapsulation and polymorphism. Geographic
relationship between objects would also be mapped by an object-oriented data
model. The geographic operators (inside, pass-through, overlapping, adjacency)
should define the relationship of the objects. For example, aerial cables must be
laid through the poles, i.e., aerial cables must be laid adjacent to the poles.
