Interesting fact about the Opensolaris development cycle

Opensolaris has the reputation of being the unstable tree of Solaris. Nowhere near of production quality. Like the development builds of Linux. But the reality is: There is nothing like a experimental tree in Solaris. When you want to do development you essentially branch the code, doing development in your own branch, offering BFU packages for early testing and only after testing you put new code into the mainline code by updating your branch to the actual state of the mainline, working through all correction nescessary by changes in the mainline code and then put the whole stuff back to the mainline. The development follows an important dogma: FCS quality all the time. FCS stands for First Customer Shipment - the first non-beta version of a product. The idea behind this dogma is, that you don´t integrate code into the mainline with known gaps, errors and problems. It must be possible to take Nevada, doing quality control on it and make Solaris 11 FCS out of it. The article “FCS Quality all the time” written by Barbara Corwin shows a nice consequence of thid dogma. As a part of “eat your own dogfood” some of the production servers ran at OpenSolaris 45 for almost 8 months:

And it's worked. From August 2006 until around March of this year, we ran Build 45 on the infrastructure machines (note that the current build in progress is Build 100). We finally hit some system problems this Spring that were fixed in later ON builds, so we started upgrading. And in general, we want to set up a regular rolling upgrade plan to keep more recent builds on these machines.