Setting Up Subversion for One or Multiple Projects
The history of versioning systems is quite extensive. They have proven to be effective during the many stages of project development, from source code management to documentation and publishing. In the Open Source community, the Concurrent Versioning System (CVS) has become the standard in the development process, being an essential tool to coordinate the efforts of tens to hundreds of developers around the world.
After many years' worth of success stories, however, CVS has begun to show problems, mostly related to security and to the lack of such important features as atomic commits. Recently, many CVS-replacement candidates have begun to appear on the scene. Some of them still are immature for critical deployments, while others propose radically new approaches, making them inadequate for a smooth migration.
Among this plethora of new players, Subversion is receiving the attention of many open-source developers due to its robustness, similarity to CVS and innovative architecture. Having recently marked its 1.0 version release, Subversion is being used in many open-source projects, including SpamAssassin, the Linux 1394 FireWire support project and the SILO Sparc Boot loader.
As with any new toy that comes into a system administrator's hands, security is the first concern when dealing with Subversion. The good news is Subversion has been developed in these hard times, in which few can be trusted. Subversion takes the smart approach of coupling with the Apache Web Server and delegating many security functions to Apache. This approach has many advantages:
Apache is a mature and well understood platform for handling security functionalities, such as authentication, access control, confidentiality and so on.
Apache is monitored strictly for security bugs: response times to a new vulnerability are very low.
Apache adopts many of the most recent standards, making it an always current platform.
Apache often is present in the established network infrastructure, so there's no need to open new network ports and adapt your firewall accordingly.
Apache's flexibility in log management eases the task of security auditing, which can be accomplished using well-known tools.
In this article I deal with a complex Subversion repository deployment and show how security concerns can be addressed from a system administrator's point-of-view.
Our task is to deploy a Subversion repository for our projects; it must be accessible both from our internal network and the Internet. Our organization already runs an Apache Web Server, so we will use it as our gateway to the repository. Thus, the Subversion configuration must meet the following requirements:
We want to host multiple projects in the repository, grouped as public and private.
We want our developers to have unlimited access the projects in which they are involved from anywhere around the world.
We want other people to have read-only access to our public projects.
Given these requirements, we must configure the Subversion server properly to manage authentication, access control, data confidentiality and integrity. But, in what form does the Subversion server come? There's no unique answer to this question, but a common strategy is to build the Subversion server as an Apache 2.0 shared module by extending the built-in mod_dav.so Apache module. In such a configuration, Apache takes care of many of the aforementioned security aspects, so you don't need to learn another configuration language. Simply adapt the familiar Apache configuration files to your new needs.
Even if Subversion requires Apache 2.0, this is a minor problem, because you are not required to migrate your current Apache installation to the 2.0 series. A simple and effective strategy consists of letting the Apache 1.3 Web Server proxy any Subversion HTTP requests to the Apache 2.0 Web Server. You can migrate your Apache 1.3 installation at a later time or never migrate if you aren't forced to.
Figure 1 illustrates the environment with which we are working. Subversion clients connect to the server from the Internet or from a trusted subnetwork. Here the term "trusted" means passwords won't be sniffed either because we trust users or because we adopted other countermeasures. The HTTP request--possibly sent over a secure channel--enters the server and contains DAV methods from the Delta-V extensions. Then, Apache 1.3 proxies the request to the Apache 2.0 Web Server, which begins the examination. The user issuing the request is authenticated via plain HTTP authentication or through the use of client-side certificates. Then, an access-control decision is taken and the access control rules are enforced. Accepted requests are passed to the Subversion module, which generates a response.
|Privacy Is Personal||Jul 02, 2015|
|July 2015 Issue of Linux Journal: Mobile||Jul 01, 2015|
|July 2015 Video Preview||Jul 01, 2015|
|PHP for Non-Developers||Jun 30, 2015|
|A Code Boot Camp for Underprivileged Kids||Jun 30, 2015|
|Comprehensive Identity Management and Audit for Red Hat Enterprise Linux||Jun 29, 2015|
- Privacy Is Personal
- PHP for Non-Developers
- Secure Server Deployments in Hostile Territory
- Linux Kernel 4.1 Released
- Django Templates
- Comprehensive Identity Management and Audit for Red Hat Enterprise Linux
- July 2015 Issue of Linux Journal: Mobile
- A Code Boot Camp for Underprivileged Kids
- Attack of the Drones
- Practical Books for the Most Technical People on the Planet