Re: [xwiki-devs] [Proposal] XE 3.0 and Roadmap leading to it

On 11/01/2010 12:50 PM, Gregory GUENEAU wrote: From the committers' point of view, it makes perfect sense for 3.0 to be the culmination of the 2.x releases, but I'm not sure the users understand this as well, so I'm extending the question to the users list. Traditionally, proprietary software is developed behind the curtains, and it's normal to have one big release every two years, with lots of new features and some bugs which get fixed in subsequent bugfix releases. Marketing comes mostly from the proprietary software world, so from a marketing point of view, this is the "normal" way releases work. In the open source world, since the development is done in the open, it doesn't make sense to stash code away in a repository and only release it once every two years. Still, most big projects use a release versioning scheme similar to the proprietary products, but with a slight difference which deeply changes the meaning. While proprietary products use only a number (3, 8, 2010...), with eventual bugfixes versions (SP2, 5.5), open source usually has 3 numbers in its versions, with the following meanings: The first number changes rarely, and when it does, it signals a critical change, like a complete rewrite of the codebase, a change of paradigm, or major new features. The second number is the one that actually identifies a release. The third number is the bugfix version. So, when we say that PHP 5.3.2 was released, we actually say that version 3 of PHP5, which is a different thing than PHP4, has been released again, giving the second bugfix release. KDE 4.5.1 means version 5 of KDE4 saw its first bugfix release. Another tendency is for open source software to linger towards a major release. For example, lots of software have a lot of releases before 1.0, going nearer and nearer: 0.6, 0.9, 0.9.9... And everybody knows that 0.x comes before 1.0, and it's not just a bugfix version of an imaginary 0.0 release. A different topic is that of agile development, with very short releases (2-3 weeks) for which traditional version number make no sense, since such a product would reach version 42 in a couple of years. Either an identifier, such as the SCM version number is used, or a continuous counter (v1.42) is used. Since the software evolves in a fluid manner, without a "breakthrough" version coming out of the regular releases, a "major" version is released when the current features are stable and they mix well together in a coherent product. Since releases come so frequently, it's normal for users to be split into two categories: those that follow the releases and know how the software is evolving, and those that only monitor the major releases, and which see all the continuous improvements at once. While XWiki Enterprise is used in the enterprise environment, and it is backed by commercial companies, it is a true open source project, following an open source release model, so I don't think that a proprietary product release scheme is suited. Our development/release style is closer to the agile development practice, albeit with mixed release types. In the future, once the core is more stable than it is now, we'd like to move even closer to a full agile release process, with 2-3 weeks between final releases. Thus, I believe that the last release versioning strategy is the best for XWiki Enterprise. Note that we're already using this strategy in a more obvious way for smaller modules (applications, skins, tools), where we do have 1.32 as a stable version number. Also note that although 1.9 was followed by 2.0, this was just a coincidence, and not a natural version evolution, since we also felt that the code was mature enough to receive a major number bump, but it could just as well have been followed by a 1.10. So, there are two different opinions here. One that 3.0 should present the groundwork/roadmap for the following 3.x releases, and one that 3.0 should be the culmination of the work done so far on the 2.x releases. The committers (with the exception of Ludovic) believe that the latter is the better one, and it is in accordance with what we've been doing so far. What do our users think? As a final remark, the XWiki Open Source projects are governed by the committers, so in the end the decision lies in the hands of the developers, but we're always open to the larger community. If no consensus is reached about when to release 3.0, we will continue releasing 2.x versions. What do XWiki users think? Is 3.0 as a culmination of the 2.x releases, with no major new features besides what 2.5 already provides, something the community expects? XWiki SAS is in the Gartner report, as a vendor. The XWiki Enterprise project is not in that report. I think the marketing direction that Gregory and Ludovic are supporting is better suited for the company offerings. Yes, these should be central in the future 3.x releases, but not in 3.0, which should be as stable as possible, without any "demo" features. -- Sergiu Dumitriu http://purl.org/net/sergiu/