paper discusses some trends in spatial database
technology and implementation and the way in which
these interact with the trends in update technology. To a
large extent this interaction promises to be a beneficial
symbiosis, but some problems still to be solved are
identified and probed.
Some key developments in spatial database technology
are described, including object-orientation and
versioning. The impact of these on the utility of large
spatial databases is considered along with further
developments in client/server architecture, distributed
processing and Internet access. The critical role of
database integrity emerges, and this is examined
particularly in the context of revision and update. The
parallel tasks of information refinement and enhancement
are also examined.
1.2 Data Transfer Standards
The relevance of emerging trends in data transfer
standards is briefly addressed, including the transfer of
update information. Short and longer term approaches to
multiproduct databases, supporting both a range of
cartographic products and a range of digital data
products, are discussed with particular reference to
update processes. Finally the view is advanced that by
the Millennium the convergence effect that is already
apparent in update technology, both from field sources
and from imagery, will have resulted in such a degree of
standardisation that there will be very little in the way of
discriminating factors between rival offerings. The
discriminating factor will be the database, and the ease
and efficiency with which any particular update system
can integrate with it.
2. TRENDS IN SPATIAL DATABASE TECHNOLOGY
2.1 Value in Information
Investors in spatial databases, whether they be
governmental agencies or commercial companies, are
being driven to increase the value of their information. In
addition to refining its accuracy and currency they seek
to add to its utility and range of application. Ideally they
seek to do this without having an ever increasing
proliferation of independent databases, all demanding
costly maintenance. In practice, this aim is not
immediately achievable. A viable approach is to reduce
the number of independent databases, and to automate,
or at least orchestrate, the posting of updates across
them. Software tools which identify, validate and codify
changes together with flowline management tools and
data models which provide means of administering
changes provide the essential framework. A conceptual
model is shown in Figure 1.
956
2.2 Object-Orientation
Object technology provides a very productive approach
to increasing the value of spatial information. Modelling
of the subject is in terms of objects that closely fit the
real world, rather than just in terms of geometries and
relational tables. Object-orientation (O-O) goes a step
further, and allows the behaviours of objects to be
modelled. O-O databases and O-O programming
languages are well-established in mainstream Information
Technology (Informatics) and are being effectively
deployed to handle large volumes of spatial information
in a very flexible manner. In particular O-O spatial
databases provide very efficient support for topology and
spatial generalisation, and have very positive scaleability
characteristics. They can provide large area support from
a common base for both a range of map scales and a
diversity of data products (e.g. link-node transportation
networks and polygonised boundary data). An overview
of Object-Orientation in the context of Geographic
Information is to be found in (Woodsford, 1995) and a
more popular treatment from a general management
perspective in (Taylor, 1990).
2.3 Versioning
Database Versioning is a second crucial advance. It
provides support for very large continuous spatial
databases in an economical and manageable fashion.
Versions are maintained in a tree structure, with only
change information recorded, rather than complete
copies. Multiuser update is supported, with each update
process having its own logical copy of the complete
dataset, without the overheads of providing a physical
copy to each process. The concepts are illustrated in
Figure 2, which also shows how versioning can be used
in conjunction with long transaction support. An Update
process (or user) has exclusive write access to a defined
segment, which is simply a logical set of objects within
the database. The set may be defined by a spatial extent,
a set of object classes or any other logical rule.
Segments have to be logically exclusive, although any
process can have read access to the whole data.
Updates can be validated for internal correctness as they
are generated, and validated for consistency with the
database as a whole prior to merging the updated
version into the master data. The mechanism for
validation is the use of methods, which are the general
mechanism in an O-O system for invoking behaviours.
General and specific validation methods can be built into
the object database schema (i.e. not at the application
programme level) and used to ensure integrity of the data
across update processes.
A powerful concept is that of the Object Lifecycle,
illustrated in Figure 3. This provides a framework for
managing all phases of the life of an object, including
updates over time, by defining methods to be invoked at
each relevant stage in transactions involving objects, and
ensuring they are invoked. The mechanism is analogous
to that of triggers in relational databases.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
ste
wit
ad
en:
prc
enm
ime
3.2
Thi
an
jus
hig
inc
tha
evc
(ret
refi
it w
enc
to t
The
har
usu
The
car
giar