Jon Udell writes about the scalability myth. How true! I wish I had a dollar for every time I saw a reasonable solution get rejected because of misplaced or erroneous beliefs about scalability and performance.
Jon's article reminds me of an article I wrote awhile back, "The Performance Presumption." I wrote it as a reaction to what seems to be a common misperception in the middleware world: that "high performance" automatically means "high quality" and thus "the best answer." That's just an extreme case of premature optimization. Depending on the specifics of the problem at hand, other qualities such as configurability, flexibility, manageability, deployability, versionability (if that's even a word), and others often matter more.