Potential and Actual Collaboration Support for Distributed Pair-Programming

In Pair Programming – a software development technique that is part of Extreme Programming (XP) – two developers work side by side, on a single computer, to jointly produce an artifact. It has been reported that Pair Programming can be accounted for the development of higher quality software in half the time it required a single programmer. Pair programmers are meant to be collocated since they require continuous and fluid communication. However, the globalization of the software industry and the growth of open source software development are trends that motivate the need to support Distributed Pair Programming. For distributed pair programming to be effective, its support should facilitate not only actually collaborating in pairs, but also the pairing of distributed colleagues in an opportunistic and flexible manner. In this paper we present the COPPER collaborative editor, developed using the Collaborative Spaces Model, to support pair programming during actual collaboration, and for potential collaboration; awareness on the opportunities for collaboration when a distributed colleague is available or working on a file of common interest. We also present the results of an empirical evaluation of the actual collaboration features of the tool. The evaluation considered three different working conditions: pairs collocated sharing a single computer; distributed pairs using application sharing mode; and distributed pairs using collaboration aware facilities. In all three cases the subjects used the COPPER collaborative editor. The results support our hypothesis that distributed pairs could find the same amount of errors as their collocated counterparts. However, no evidence was found that the pairs that used actual collaboration awareness services had better code comprehension, as we had also hypothesized. Overall, according to comments from evaluation participants’, COPPER improves Distributed Pair Programming, in subtle but significant ways, by allowing concurrent work and better maintaining awareness on the concurrent actions of the pair, in contrast to collaboration-transparent applications which do not support these features.