Over at Language Log, Professor Geoffrey Pullum gives the lie to the canard that descriptivists think that there are no rules in English, presenting a compact summary of the punctuation of relative clauses:
There are two major ways in which a relative clause may function. One is that a relative clause may be a fully integrated modifier of the noun in a noun phrase, often providing some sort of semantic restriction on the reference of that noun. Thus person can be used to denote the entire class of human beings, while person who has been unsuccessful denotes only the smaller subset of those who have failed at something. The underlined part is what The Cambridge Grammar of the English Language calls an integrated relative clause. They are often called "restrictive" relative clauses, or "defining" relative clauses.
The other major function for relative clauses is to serve as a parenthetical interruption of the main flow of a sentence, contributing supplementary information about someone or something immediately after it is referred to in the main content. Thus You can talk to John if you like just says that if you want you can talk to John, but You can talk to John, who has more experience, if you like adds some supplementary (and definitely secondary) information about John's experience level. This kind of relative clause is the one that CGEL calls a supplementary relative clause.
There are all sorts of differences between the two, but the one that is crucial here is that supplementary relatives must be separated off with commas and integrated ones must not be.
The erroneous use of whom as a subject also comes in for attention.
The comments on the post are well worth your attention, particularly when they veer off into the mistaken belief that “[u]ltimately, all a comma is is a breath or short pause.”
What some people, many of them my students, have difficulty in grasping is that the comma functions in two ways. In some cases, as in the supplementary relative clauses that Professor Pullum describes, or in appositive clusters, the commas are essential. That is a rule. But there are also commas with which writers try to indicate pauses mimicking the rhythms of spoken English. They are discretionary.
Note to readers: I have usually tried in this blog to indicate an extended quotation by boldfacing the text to distinguish it from my own comments. But some readers have found the boldface type difficult to read, and in this case the original text contains boldface type. So I am experimenting here with putting a block quotation in a different font to set it off. What do you think?