Summer of Code projects

If you are interested in, or have questions about any of these projects, please contact the nominated developer.

Why contribute to OpenSSH?

We are also a well established software project (10 years old in September) with high aesthetic and quality standards for new code. This will provide a rewarding learning experience for someone new to free software development. Students looking for academic projects should be aware that a number of high quality papers have been written about various aspects of OpenSSH's implementation.

Contributor requirements

Source code is managed using CVS and bugs are tracked using Bugzilla. Familiarity with these tools is desirable, but not absolutely required. Final code submissions must compile without warnings and function properly on OpenBSD and Linux (at least).

After the initial application, selection and project startup, contributors will be asked to communicate on the public mailing lists and bugtracker. Contributors should expect debate and criticism of their proposals and implementation, but they can also expect this debate to be constructive.

Finally, contributors are not on their own - their mentors and the OpenSSH mailing list community are good sources of advice, ideas and wisdom.

Available projects

Project	Complexity	Contact
Renovate sftp(1)	low-medium	djm
Improve testing infrastructure	medium	djm
Performance improvements	medium	djm
Something else...		djm

Renovate sftp

sftp(1) is, at present, primarily an interactive program with a relatively weak commandline interface. It lacks essential features that are present in scp(1), especially support for recursive operations.

This project would seek to make sftp(1) a drop-in replacement for commands that use scp(1) by implementing a compatible commandline syntax. This would require implementation of recursive uploads/downloads, as these are essential features supported by scp(1). The new features will need documentation (manual pages) and regression tests.

Further projects might include:

• Final bugfixing and integration of the tab-completion patch for interactive mode
• Implementation of inter-file pipelined transfers. Currently sftp will pipeline uploads and downloads but requires a pipeline flush and server round- trip for each new file. This causes a slowdown when transferring many small files.
• An implementation of an rsync-like operation in sftp.
• Automatic tuning of pipeline settings based on network conditions.

Improved testing infrastructure

OpenSSH currently has a set of regression and interoperability tests, but these are by no means rigourous. They do not achieve 100% coverage of the available configuration and command-line options, let alone code coverage. There is no support for protocol-level testing, especially fuzzing.

The goal of this project would be to improve testing by

1. Improving test coverage of options (commandline and configuration file)
2. Establishing measurement of code coverage
3. Developing a means to perform protocol-level testing, and using it to implement basic fuzz tests.
4. Development of protocol-level feature and RFC-compliance tests.

Performance improvements

There are three aspects of ssh(1)/sshd(8) performance that are particularly interesting: CPU usage, memory usage and network performance. Some attention has been given to CPU and network performance, but more work could be done.

This project would examine and make quantifiable improvements to one or more of these performance aspects. A necessary precursor to this work will be development the some performance measurement standards and tools that will be used to assess the outcome.

This is probably the most open-ended of the suggested projects.

Something else...

We are open to new ideas and opportunities to improve OpenSSH, so feel free to suggest something. If you would like some inspiration, you might also want to look at the portable OpenSSH bugtracker, as many enhancement requests are listed there.