SX:CE EOL; or, The Great Mistake
Posted on August 27, 2009
I’ve been exceedingly slow, despite prodding, to respond to the announcement of Solaris Express Community Edition’s (SX:CE) end-of life. I have been largely because I’ve warned against it many times before, and I shall explain my thoughts on the matter again…
The death of SX:CE represents the end of an era. The cornerstone of any distribution is its installer(s) and packaging. Consider Debian or Red Hat or SuSE. Fundamentally they are all the same; with all the “default” tweeks and installation preferences you can perfectly replicate the experience of one with another, the core difference lay in their packaging (RPM, apt-get, etc) and their installers (Anaconda, YaST, etc.). SX:CE has served a very special purpose of giving us the latest Solaris bits while maintaining the old tried and true distribution. Chucking SX:CE is closing the door on the ways and means we’ve used to manage systems for over a decade.
Now, you may argue that the Solaris PKG format, Jumpstart, and the legacy installer are terrible, horrible, ghastly things that deserve to die the death… and you may just have a point. But they have one thing that their successors, IPS, Caiman, AI, etc, don’t have…. maturity. I don’t bash Solaris often, so gather round, these opinions are well known by the authors of these technologies.
Caiman is a pretty installer, but its terribly immature and certainly not production ready. Come on, there isn’t even a text driven version of it! It’s GUI only! Ridiculous!
IPS is a novel packaging format, but its very young and very fragile. Implementing it in Python has been a constant thorn in the developers sides (that is to say, users give them an exceedingly hard time about it). It makes Sparse Zones impossible and installation of Zones on an IPS system involves re-pulling packages from the network repository, meaning that if the network or the packaging server is down, you can’t install zones. Plus, its a very painful system to create packages for, very unintuitive… novel, but unintuitive and frustrating to new packagers.
The Automated Installer (AI) will replace Jumpstart. Jumpstart itself is frock with problems. Its complicated and difficult to understand because its not integrated (meaning, with Jumpstart you manage NFS, TFTP, DHCP/PXE, plus the Media and various pre,post,finish scripts separately but to work in harmony, which takes a long time to get good at). Never the less, AI has so far shown to be immature and fragile. Conceptually I love the architecture of AI, because unlike Jumpstart it is a very integrated system which makes it simple and more straight forward to manage. Never the less, enterprise environments have done the hard work, trained their people, and know and trust Jumpstart despite its failings. AI isn’t nearly there.
What am I saying? OpenSolaris (meaning, Indiana) is a fabulous desktop/laptop distribution, but it is a horrific excuse for a server OS. I’ve warned against this for years. Sun has really focused on the laptop experience in order to whoo Linux users over to OpenSolaris, but all the while Linux users have been abandoning their PC/Linux laptops for Mac’s. We fought the wrong fight.
Untill the EOL announcement all this has been academic. We had both options available to us. Want Nevada with the old distribution? SX:CE. Want Nevada with the new distribution? OpenSolaris. Everyone is happy. There is balance in the universe…. no longer. So now the choice is less friendly, embrace OpenSolaris lock, stock and barrel or abandon it for the comfort and safety of Solaris 10.
My long held opinion is that SX:CE shouldn’t be killed, but instead it should become Solaris 11. OpenSolaris can continue to do its thing for desktop users, but give enterprise environments what they need to continue using the platform. One day, I hope, all the kinks and immaturity will be worked out of the new distribution, but it hasn’t happened yet, not by a long shot. I’ve held to a delusion that it wasn’t a big concern, really, because Oracle would come in and slap some sense into the roadmap, but no signs have yet suggested that to be the case. We can only hope that by year end something pans out.
My message to the powers that be at Sun… I love ya, I love Solaris, but don’t take this from us. Please, please just don’t.