The Unglamorous Stuff
Lots of posts on operations today. Bill de hÓra is a good starting point, where he outlines a contrast between Steve Loughran's and Dave Thomas' statements. I'm not so sure they're that far apart. Steve also says "If you are developing server-side applications, keeping the operations team happy must be one of your goals." I think that's what Steve means by splitting dev and deployment. If you're going to just throw an application at your ops people and expect it to be their problem, you're going to be disappointed. Dave T. does say something about making deployment issues "someone else's problem". Someone else's problem will become development's problem before too long. I don't think Dave was advocating the "throw an app at operations" approach though, he was advocating reducing the friction between groups.
Dare Obasanjo posts on operational concerns, more specifically, provisioning and power considerations, and James Robertson argues that small-scale operations don't have this problem. I side with Dare on this. I think that these considerations have a way of catching up to you very quickly. I know that power is always a consideration, and we now have full-time operations people who have to worry about that. Modern architecture is very much going in the direction of "don't worry, you can always throw another server at it", but you do have physical constraints. You don't want to be in a situation where you want to throw another server at the problem, only to find out that you don't have rack space, or your data center can't supply enough power to your rack. My takeaway from the original O'Reilly post is that these concerns will make you a victim of your own success. I can't speak to the point of whether MS or OSS has a better story here. I honestly haven't seen a great story from either perspective yet.