Believe me, I'd be the first in line to shut down recommendation questions if there was a real problem, or if there was a good reason to, but I'm not seeing either here.
How are we doing right now in handling these types of questions? Are the current policies sufficient or are there more concrete policies that would be best implemented?
It's hard to tell how we're handling these types of questions without a significant amount of data demonstrating a problem. You've cited two questions:
They were both asked by the same person in rapid succession relatively about the same topic and both are absolutely terrible. If these two questions are demonstrative of an alleged problem with book recommendation questions, then I don't see it.
The biggest issue with their closing is that they were closed as "not constructive" instead of what they are: not real questions. The "not constructive" close reason seems to be used more and more as the "I don't care why it's closed, just close it" reason, which dilutes its meaning and causes ambiguity about why questions don't belong.
The problem with these questions aren't that they are asking for a book, it's that they don't describe anything about the problem the user is having, or rather why the user is asking the question, to afford a useful response. They're like asking, "tell me what book will teach me everything and anything about Java." Where to begin? Questions must narrow the scope down to a specific problem, otherwise, they don't belong here.
Because the two cited questions are incomplete in this respect, they're not real questions. They'd be "not constructive" if they had asked something like:
- "What's your favorite way to study for Exam 486?"
- "Exam 487 really sucks, how am I supposed to study for this stupid exam?"
- "How does Exam 486 make you feel (as a programmer)?"
- "What are some tips and tricks for overcoming my inability to take Exam 487 seriously?"
- "My co-worker/buddy/priest/cat says X book is the best for Exam 486. Help me prove him/her/it wrong."
Without expertise in the subject matter, how do people (especially moderators who might have to act on a flag) know if there is a singular "canonical" book on the question?
That's not the guideline for books, and it's not up to anyone (much less moderators) to determine ahead of time (or during the act of moderation) whether there is a single canonical book for a subject.
Instead, mimicking the guidelines in "Good Subjective, Bad Subjective", and as I described in "What's up with all the book recommendation flags?", the important things are:
the topic asked about is capable of providing a canonical book on the subject. The application of this criteria requires common sense. Java doesn't count. PHP doesn't count. NLP for Big Data probably does. F# or any other boutique language probably does. If someone's really in doubt whether something counts, do an Amazon.com search for the subject. If there's over 10,000 hits, it doesn't count.
the questions sufficiently describes a situation or problem space that invites converging on a small number of correct answers. "What are some good books on F#?" sucks. "Coming from a purely C# background, I'm trying to add F# to my repertoire. What is the definitive [read: canonical] book for getting up to speed with idiomatic F#?" doesn't, or at least sucks a lot less. If in doubt, wait. If the question attracts dozens of answers, that's a problem. If it settles on 3–5, awesome. If there's a dozen but only a couple of really stand-out answers, cull the herd and protect the question if necessary.
Neither of these guidelines require domain-specific knowledge of the question; just an understanding of how to ask a good question, which is something anyone moderating (diamond or otherwise) should have a good handle on.
What happens to topic areas covered by a small number of resources now (such as the latest language or library) down the road, when it is more popular and prevalent? Might resource requests, perhaps, be too localized since the available (and perhaps even what is considered canonical) change over the course of time?
Irrelevant: Stack Exchange is designed to handle changing answers, even rapidly changing answers, through frictionless editing. Answers for pretty much any subject or problem change over time.
Problems arise trying to keep questions up to date when there are dozens of posts to maintain, which is why there are guidelines and technical features that kick in once a question starts getting over 15-30 answers. Per the guidelines for book recommendation questions, this is a non-issue for them.
So unless there's some other reason, not mentioned, to shut them down now, I don't see anything that's cause for a policy change.