Por Qué Los Proyectos De Software Se Están Moviendo A Los Sistemas De Control De Versiones Distribuidos?

Why Are Software Projects Moving From Centralized to Decentralized Version Control Systems?
Brian de Alwis Dept of Computer Science University of Saskatchewan Saskatoon, SK, Canada
brian.de.alwis@usask.ca

Jonathan Sillito Dept of Computer Science University of Calgary Calgary, AB, Canada
sillito@ucalgary.ca

Abstract
Version control systems are essential for co-ordinating work on asoftware project. A number of open- and closedsource projects are proposing to move, or have already moved, their source code repositories from a centralized version control system (CVCS) to a decentralized version control system (DVCS). In this paper we summarize the differences between a CVCS and a DVCS, and describe some of the rationales and perceived benefits offered by projects to justify thetransition.

1. Introduction
For many projects, their version control system (VCS), along with other tools such as the issue tracking system, is central to how development work is organized. The central challenge in managing software development is scaling the change process up to large numbers of possibly geographically-distributed software developers without sacrificing quality or introducingundue overhead. To a large extent these tools determine how easily people can contribute to a project, how new feature development is co-ordinated, how often separate development lines are merged, how code is reviewed and how the support of already released code is organized. Despite their importance, the impact these tools can have on the development work and the trade-offs involved is not yet wellstudied. A new generation of VCS, called decentralized VCSs (DVCS), have emerged that address some of the limitations of current centralized VCSs (CVCS), such as CVS [1, 5] and Subversion [4], to better support decentralized workflows. Some of these new DVCSs, such as GIT,1 M ERCURIAL,2
retrieved 2009/01/20 2 www.selenic.com/mercurial/; retrieved 2009/01/20
1 git-scm.com;

and B IT K EEPER,4have become sufficiently mature that many open- and closed-source projects are proposing to move, or have already moved, their source code repositories to a DVCS. As part of a larger research project to explore practices and tool support around version management, we have begun a qualitative study to answer two research questions. First, what do these projects see as the benefits of a DVCS?Transitioning a source code repository to a new VCS requires significant effort,5 and so we suppose that there must be compelling reasons for switching. Second, what changes have these projects made to their development processes in making the switch? To answer these questions, we examined publicly-available documents and mailing list discussions for four open-source projects (Perl, OpenOffice, NetBSD, Python)that have moved or are contemplating moving to a DVCS to identify rationales and perceived benefits offered by projects to justify such a transition. In this paper we present some initial observations from our ongoing analysis. This paper is structured as follows. In Section 2, we provide some background, summarizing the uses of VCSs and a brief explanation of what makes a DVCS different from aCVCS . In Section 3 we summarize the anticipated benefits, and possible issues reported by several open-source projects that have made or are considering making a transition. We finally conclude in Section 4.

BZR ,3

2. Background
In this section we provide a brief description of both centralized and decentralized VCSs, and highlight the opensource projects examined. Note that as the variousDVCS
retrieved 2009/01/20 retrieved 2009/01/20 5 For example, the Perl Foundation reported that their transition to GIT took approximately 21 months, and involved significant manual work [9].
4 bitkeeper.com; 3 bazaar-vcs.org;

implementations are still undergoing rapid evolution, a detailed feature-by-feature comparison of the current tools would be quickly obsolete. Rather we compare the...