"Therefore, the speed of your server-side programming language is not important (or less important)." I view this statement as a prediction, in the scientific

"Therefore, the speed of your server-side programming language is not important (or less important)." I view this statement as a prediction, in the scientific sense: "If you re-implement a program in a faster language, but have the same database access, the user will not have any visible performance improvements." Like almost anything else, the answer is, "it depends". It is often true. However, I have falsified it in some specific instances. It is definitely possible to take a "scripting language", which ranges in the 20x-50x slower than C (a rough, but adequate, approximation of "optimal"), then layer it in a couple more layers of abstraction that may be really nifty and cool and awesome but also adds another 10x to the slowdown (or more... if you're not careful, this term can be unboundedly large), and get to the point that simply generating an HTML page takes a full second of a 3GHz's processor time. I have replaced things that did non-trivial database work with something in a compiled language and gone from "noticeably sluggish" to "essentially instant" without all that much effort. In one of those cases, significant effort had been expended trying to figure out why the database was so slow, presumably because of the prevalence of this meme, when even the first glance at the output of a profiling tool said that was not where the problem was. Performance is rarely the first thing to worry about, but I find it's something you do want to at least keep in mind while developing. If you do at least keep it in mind, I suspect the saying is generally true, and the raw speed of the underlying server language won't bother you much. But you are definitely much closer to the point where performance becomes problematic than you start at with a faster language, so you need to be that much more vigilant. People who are lulled into a false sense of security by the idea that the database is always the bottleneck can get blindsided by performance issues in their app. I think the meme overall is damaging. It's not true enough to make it a core, unquestioned assumption that you can rely on, and if it can't be that, then you might as well simply use the correct answer, which is to use a real profiler and look at the real data. I pretty much always through at least a basic profile pass at any code that isn't just brute scripting.