It is easy to be critical of Wang and Armstrong (2009) as someone with no background knowledge of cyberinfrastructures (CIs). Before getting into a discussion of the actual content of this paper, I must comment (as others on this blog have done before me) on the authors’ gratuitous use of jargon and unclear writing. This paper is incredibly inaccessible to anyone outside of the field.
Despite the authors’ aims to develop a broad theoretical approach for the use of CI in geospatial analysis, the specificity of the techniques discussed made it difficult for me to imagine how I might use such techniques (or cyberinfrastructures more broadly) in my GISciences research project. If the goal of parallel processing methods is to reduce the computational load on a given system, what is the threshold at which one should decide to use a parallel processing method? Does a researcher ever run the risk of making a computational task more complex by attempting to use a CI unnecessarily?
While I am critical of the specificity with which the authors talked about CI techniques, I wish that they had gone into greater detail about the significance of geospatial data type on these parallel processing methods. Wang and Armstrong briefly make the distinction between object-based and field-based representations, but do not go into much detail about how fundamental characteristics of specific data types might impact parallel processing methods. I would also be very curious to learn how CIs can be applied to handle analytical tasks with streaming data.
Wang and Armstrong’s discussion of “computational intensity and computing resources requirements of geographical data and analysis methods” (1) points towards a future for geography that is closely tied to the field of computer science. While many of the details of this paper went over my head, I believe that there is a lot of potential in a closer relationship between these two fields.