This is a non-issue, or should be anyway, for philosophers. Language changes, at a speed and in a manner to be determined by its users on a case by case basis. In the face of perceived linguistic anarchy, prescriptivists are right that there are indeed objective "linguistic facts" -- that word X means "y" and not "z", or that string W is not well-formed -- but any imperatives (prescriptions) we may derive from same are what you might call "hypothetical" rather than "categorical." That is, while it is (as they say) a true fact that "dog" means "one of those [indicating Fido]" and not "one of those [indicating Garfield]", the only sense we can ultimately make of this sort of "semantic normativity" is that IF you wish to speak in the way that English speakers (as a matter of empirical fact) typically do, THEN (and now comes the normativity) you must say "dog" when speaking of Fido and his canine kin, and "assassinated U. S. President" when speaking of Garfield, Lincoln, et al. (A related point is that IF you wish to be understood by English speakers (or perhaps to keep from annoying them), THEN you must speak in the way that they typically do, or at least refrain from improvising freely.)
Or so we Davidsonians say (locus classicus: "A Nice Derangement of Epitaphs," now available in Truth, Language, and History). This means that in particular cases the issue is not whether there are linguistic facts, but whether our purposes are best served in this case by speaking this way or that. In general, useful distinctions are retained while useless ones fade away. The problem arises when some people let distinctions become obscured while others are still using them (and hoping to do so in the same way they had been doing up to now, by saying this and not that). The proper philosophical attitude to take here is again that it'll all come out in the wash (which is often, I find, the proper philosophical attitude to take).
So we should be neither hard-line prescriptivists nor hard-line descriptivists (as neither position is coherent). The sympathy I have for either is limited to the empirical fact that proponents of the other can be annoyingly dense. If language currently allows a useful distinction, then let's not just let it go out of sheer carelessness. Common courtesy suggests that if someone is using something, you don't just throw it out just because you can't be bothered not to (and maybe you'd like to use it too - try it and see). We all have our favorites, but I appreciate the semantic distinction between "jealous" and "envious", and of course I have a professional interest in keeping "imply" and "infer" straight. And some things are just eyesores (quotation marks for emphasis, possessive "it's"; again, we all have our favorites).
On the other hand, there's nothing quite like doctrinaire prescriptivists for cluelessness. They make me appreciate the existentialist notion of "bad faith": I don't have to use my judgment, there's a rule I can follow (and berate others sanctimoniously for not following)! It's like (and indeed I imagine in some cases this really is it) they want to be able to feel justified in inferring social class (and thus ultimate worth) directly and reliably from forms of speech - how can we distinguish Us from Them if you insist on talking in a way in which (even if there is no other reason not to do so) They do and We do not?
What's particularly galling is the (not at all unusual) cases in which prescriptivists are (ironically enough) just wrong. Again we all have favorite examples (mine is the ridiculous kneejerk rejection of singular "they," which is just fine, thank you). I ran across another today, which I don't think I'd ever seen before. Baldo is perhaps the most annoying and reliably unfunny comic strip I've ever seen, but I can't help sparing the 2 seconds it takes every day to read it. Today's jape features the pretentious Anglo (naturally) English teacher upbraiding car-parts shop employee Baldo:
Baldo: Can I help you?... the point being, of course, that what B. "should" have said was "May I help you?" Now of course this latter is a traditional greeting and perfectly idiomatic, employing a locution that people do in fact abuse in the form our man no doubt usually hears it (his response here is, verbatim, that my 4th-grade math teacher regularly gave to "Can I go to the bathroom?"). But Baldo's actual question makes perfect sense on its own terms. For if he were to return snark for snark, we might hear:
PAET (snarkily): I don't know – can you?
Baldo: I don't know either – tell me what you want, and I'll see if ... (wait for it) ... I can help you.At least that's better than "Whom should I say is calling?"