Now that were a few months into 2007, I still think that the prediction—if judged by the metric of whether its likely to come true—was a lousy one. While Solaris is an awfully compelling OS, and while Im convinced that the OpenSolaris effort is for real, I think that OpenSolaris has about as much a chance of pushing Linux to the sidelines this year as Linux has of knocking Windows off the mainstream desktop. Thats not to say that either of these scenarios couldnt happen eventually, but 12 months is a pretty tight timeline.
Measures of accuracy aside, the prediction scores pretty well as a piece of writing, because I find that my thoughts often return to it, particularly when Sun makes a move that strikes me as either beneficial or deterimental to the forecasts eventual fulfillment.
Sun appears to have taken a step in the right direction recently when the firm hired Ian Murdock, the "ian" of Debian GNU/Linux, to fill the fanciful-sounding role of Chief Operating Platforms Officer. In a blog entry on the topic, Murdock described his new job as "head(ing) up operating system platform strategy," which sounds like a good perch from which to address my No. 1 Solaris peeve: software packaging.
Im holding out hope because the Debian distribution that Murdock helped found sports an excellent software packaging system that makes my life a lot more pleasant as I conduct the many software installations and updates and OS patching and testing that fill a product reviewers work (and often home) hours. A case in point: I have reviewed significantly fewer test releases of Solaris (since OpenSolaris was born, thereve been many) than I wouldve liked, mostly because that means downloading a few GB of CD images, and running through the installer program.
In contrast, when Im following the test cycle of Debian or of its popular child distro, Ubuntu, I can track the process continuously, and the cost of reviewing the latest code is typing a few words at a command line, with which I direct my test machine go fetch and install all the latest software from the network repositories Ive selected. Its not just for testing, either—if the system Im using is for production, I choose stable repositories instead, and install security updates or new applications using the same tools.
While there are a couple good Solaris volunteer packaging efforts out there, I think itll take a software management system overhaul to bring Solaris software tools up to the level to which Linux users, administrators and developers have become accustomed. Without an overhaul of this sort, I cant see OpenSolaris overtaking Linux.
Theres actually already an OpenSolaris distribution, called Nexenta, that combines the Solaris kernel with all the userland applications of Ubuntu Linux, including Debians slick package management system. I see a lot of potential in the approach, but the project could really benefit from more support from Sun. So far, however, its been important to Sun that OpenSolaris distributions sink or swim on their own, and theres been no indication whether the firm might someday bring Debians packaging tools into canonical versions of Solaris. Whether or not Sun opts to pursue a Debian tools tack for Solaris, itd make sense to contribute some resources to the Nexenta project.
Another, and potentially better route to Linux-challenging packaging tools for Solaris could be Conary, the software management tool that organizes rPath Linux and a host of derivative distributions based on rPaths rBuilder platform. Despite all my good experiences with Debians software tools, Ive found that creating and managing new packages is quite a bit easier with Conary. Again, itd be smart for Sun to stow its "sink or swim" concerns and to devote some resources toward building a conary-based distribution of OpenSolaris.
None of this will likely be enough to turn the open source operating system world on its head in the next eight months, but if Sun keeps making smart moves around OpenSolaris, such a reversal is certainly within reach.