Stein, an open source engineer with Google and chairman of the Apache Software Foundation, will be disclosing the new service officially at his talk "A Google Service for the Open Source Community," scheduled for 1:45 p.m. PDT today.
I sat down yesterday with Stein and Google's open source program manager Chris DiBona, who describe the service as similar to SourceForge.net and other community hosting projects, but not designed to compete with those projects.
Stein says, "We really like SourceForge, and we don't want to hurt SourceForge" or take away projects. Instead, Stein says that the goal is to see what Google can do with the Google infrastructure, to provide an alternative for open source projects.
DiBona says that it's a "direct result of Greg concentrating on what open source projects need. Most bugtrackers are informed by what corporations" and large projects need, whereas Google's offering is just about what open source developers need.
Stein says that Google's hosting has a "brand new look" at issue tracking that may be of interest to open source projects, and says "nobody else out there is doing anything close to it." At the same time DiBona and Stein say that Google's hosting offering will not have some features present in SourceForge.net and other code repositories that open source projects and enterprise customers might want.
With the new service, Stein says Google was able to "cut out a lot of heavy structure" and apply Google's full text search to just the features that open source projects may need. "Rather than doing queries through that [heavy] structure, we can just full text search across it all. It provides a really powerful mechanism for issue tracking, but keeps it really simple."
The other main feature for Google Code hosting, according to Stein, is a "massively scalable Subversion repository." Stein says Google rebuilt Subversion to store data in Big Table, a massively scalable, highly available storage technology used in Google.
Stein says that the company will have all the Google projects on there, but they're not going out there to get projects to move. As a precaution, Stein also says that Google has a list of SourceForge.net projects, to ensure that new projects will not encroach on existing projects' namespaces.
For example, it won't be possible to set up a Gaim project on Google Code hosting, unless there's an approval from the project owner on SourceForge.net. This will prevent any confusion or deliberate attempts at impersonating SourceForge.net projects.
Not yet feature-complete
The initial public release will not be feature-complete, but users will be able to sign up right away without an invitation, unlike some of Google's other new service launches.
In particular, Stein says that Google Code is missing file download at the moment, but that it's a high priority to add that feature. Unlike SourceForge.net, Stein says that the service will not have, and Google has no plans to add, Web site hosting for projects.
To sign up for the service, a project needs to be licensed under one of seven approved licenses: Apache license, Artistic License, GNU General Public License (GPL), Lesser General Public License (LGPL), Mozilla License, BSD license, or MIT license. DiBona says that Google is trying to make a statement about license proliferation by offering only a narrow set of license for projects to choose from.
DiBona and Stein describe the project as ideal for smaller open source projects, rather than larger projects with more complex needs, such as Apache or GNOME. However, they also say that larger projects are welcome.
One of the most discussed topics at OSCON this year has been open data -- the ability for users to get their data out of a program or service and use it elsewhere. Stein says that Google understands the importance of being able to move data. "We don't have those [migration features] in there now, but that's something we intend to [have] ... we intend to do it soon after launch."