Web Services Composition and Geographic Information - Geographical Information Systems: Trends and Technologies - page 135

encoding and decoding, the verbosity of XML and its redundant textual
characteristics. First of all, the latency and the network traffi c produced by
SOAP are considerably higher compared to technologies like CORBA or
Java RMI. These issues can have an even greater performance impact when,
for example, the client is a mobile device where the available bandwidth
can be low and the latency very high. Moreover, the process of converting
a memory object into an XML object and vice versa is a quite expensive
computational task, able to consume over 90 percent of the entire end
to end SOAP processing time. The performance problems are amplifi ed
also by diffi culties encountered by traditional hardware architectures to
simultaneously evaluate multiple conditions, which represents a central
issue in XML string and character processing. Finally, the addition of security
policies to SOAP messages adds another source of overhead. In fact, the
use of WS-Security protocol has a big impact on both the processing and
response time and on the SOAP message size. In this case, the problem is
due to the need of providing for a message level security instead of the
traditional channel-level security, where SSL/TLS on the HTTP protocol
is used.
An empirical proof of the performance problems occurring when using
W3C service technologies can be found in Zhang et al. (2007). In this paper,
a prototype system built to evaluate the performance of composition and
invocation of geospatial functionalities by using Web services is discussed.
The experimental results agree with the above discussion and show that
the trade-off between convenience and overheads are acceptable for small
volumes of geospatial data, although the large response time for ample data
volumes represents a problem that cannot be underestimated.
The Representational State Transfer paradigm
A totally different architectural style for the development of distributed
applications is represented by the Representational State Transfer (REST)
paradigm (Webber et al. 2010). Many researchers and practitioners judge this
paradigm as a feasible way to realize distributed applications eliminating
the intrinsic complexity of the Web services standards (Pautasso et al.
2008).
Two fundamental notions in this context are the concept of resource (i.e.,
any meaningful information that can be addressed) and the representation
of it (i.e., a document that captures its current state).
The REST architectural style is based on four principles (Pautasso et
al. 2008):