Implementation Strategies for CSV Fragment Retrieval over HTTP

A large percentage of the data published on the web is tabular data, commonly in the comma-separated values (CSV) format. URI fragment identifiers for CSV provide possibilities for data publishers (web servers) to refer to specific parts (i.e., fragments) of a CSV file using URIs as well as for data consumers (web clients) to retrieve URI-identified fragments over HTTP for further processing. However, few studies have focused on CSV fragment retrieval, and to our knowledge, so far there are no implementations of CSV fragment identifiers compliant web clients/servers. This study addresses the issue of implementation strategies for CSV fragment retrieval over the HTTP protocol. Two strategies (i.e., client-side fragment evaluation and server-side fragment evaluation) including the fragment evaluation algorithms were proposed and then implemented in order to demonstrate the implementability of CSV fragment retrieval over HTTP and to experimentally observe and compare the run-time efficiency of the two strategies. The results indicate that server-side fragment evaluation is a more efficient strategy in terms of running time and bandwidth cost, but requires extensions for existing web servers, whereas client-side fragment evaluation is the opposite. Our methods and findings provide valuable reference for potential implementers.