Edit: Mark Trapp wrote a pretty good answer on Programmers about a very similar issue.
questions that ask how to do something (where the "recommendation" per se is not the point of the question) are on-topic, questions that ask what will do something (where the "recommendation" is the point of the question) is off-topic
However, that mostly applies to the Programmer SE site itself.
On Stack Overflow itself there's another question with a similar answer:
The problem with these questions is that unless you are very specific about your requirements all you get is either a list of people's favourite technology or an argument in comments about which technology is "better".
The truth is that these days, unless you have a really specific problem, there can be little to choose from between technologies. This means that your question is either "not constructive" or "too localised".
So I'm inclined to agree with the other answers here now that those are generally off topic. I've updated the FAQ.
Looking at meta.stackoverflow I came across these questions:
The core of the argument can really be summed up by the blog post I've been adding as comments to a lot of these types of questions (and part of the answer to that question), at least ones that name specific tech:
The only reason I've been lax on shutting them down with a vengeance is that there usually is some kind of detail. At this point in time, I kind of feel that "which tech can do X, Y, and Z" does have an answerable question and is useful to people. And a similar question that's "which can can do X, Y, and W" isn't an exact dupe of the first because the requirements are different. So like ChrisF said in his answer to the second question, it's all about being specific.
If there are no constraints than it's not a real question. But where that line crosses from answerable to not isn't completely well defined. Like if they're looking for something that's "easy" or "fast" it's sometimes hard to answer that aspect of it.