I personally think Stack Overflow could stand to relax, just a little when it comes to questions that seek a library that fits a problem. For the most part, Google does a very good job of turning up candidates for common needs; I think we could excel in the land of caveats.
I also don't understand why people insist on writing 'good' and 'best' - as those words instantly trigger a subjective viewpoint. It's not as if we're going to tell you about a bad or worse library, just state your need.
If you're looking for an XML parser for Python, Google is your friend. If you want to compare and contrast libraries, your time is probably better spent in a lab than writing a question in most cases where just trying both yourself is something easily done.
Where Google fails you is where we could probably be a bit more helpful, if not excel. You need an XML parser for Python that can handle documents that are gigabytes in size with an alphabet soup of encodings and odd characters. To compound that, you have memory constraints, and need to be able to accomplish parsing in a certain amount of time.
That kind of question:
- Isn't going to attract very many valid answers
- Isn't going to overrun the site
- Is going to be helpful by filling a knowledge gap, however small
- Might just attract answers like "I don't know of anything, but this is what the code would look like .."
- Doesn't suck, like "What's the best jQuery image gallery plugin?" where the number of valid answers are the number of plugins that exist
We didn't get enough good examples of these types of questions early on to be able to identify ways in which they could work. Perhaps we can reconsider our definition of undesirable when it comes to these at some point in the future.
Still, the majority of these questions that you'll find on the site are not much more than lists, sorted by popularity as it was within a week or two of the question being asked. They aren't useless, but they're not great examples of the gaps we're trying to fill.
You've been sent in comments to dig into meta a bit more, I was one of the elected moderators that finally said enough is enough with these because they were becoming a constant source of work. Problems were:
- They attract answers that are barely more than a link, links break, the author is long since gone
- They attract barely more than link answers that are, at best, tangentially related to the question. People would seek out and try to answer these questions since they were seen as an easy way to get enough rep to lift new user restrictions
- They are chronic spam magnets
- They diminish the value of some of our harder to earn badges, at least in the eyes of some. Why did user1234 get a Good Answer badge for 120 characters that was mostly a link, where user2345 spent a half hour writing code that solved an interesting problem?
These could work in certain cases, provided that the questions scoped the number of possible answers to not much more than a few and ensured that answers would need to contain a reasonable amount of depth.
I don't think it's impossible, but It's not something we can easily just decide to allow one day and leave it at that. Until (and if) we work out a better way to ask library recommendation questions more freely - you can still do it - just state the objectives of your endeavor along with whatever relevant code you've got - if a library exists that is well suited, someone's bound to let you know.