Martin Haspelmath has put online a paper (to appear in Language Documentation & Conservation 2017) that describes “some principles that one might use for taking decisions [concerning language names] when there are variant forms in use, or when one feels that none of the existing names is appropriate,” adding that the principles “arose from work on Glottolog, a English-language database of the world’s languages” (and how did I not know about Glottolog?). Here’s the abstract:
This paper discusses eleven principles of language naming, which may be relevant to language documenters in case a language does not have a stable name yet: (i) Language names (like city names) are loanwords, not code-switches; (ii) Names of non-major languages are not treated differently from names of major languages; (iii) Each language has a unique name; (iv) New language names are not introduced unless none of the existing names is acceptable for some reason; (v) Language names that many speakers object to should not be used; (vi) Language names in English are written with ordinary English letters, plus some other well-known letters; (vii) Highly unusual pronunciation values of English letters are not acceptable; (viii) Language names must be pronounceable for English speakers; (ix) Language names begin with a capital letter; (x) Language names may have a modifier-head structure; (xi) The usage of prominent authors is given substantial weight. Finally I note that it is not a principle that the English name needs to be close to the autoglottonym.
In general it’s quite sensible; it’s not always clear, however, whether he’s being descriptive (this is how we at Glottolog decided to do it) or prescriptive (this is how everybody should do it), and I got my back up at section 3 (“Each language has a unique name”):
Just as cities never have multiple names within the same language (and variation is exclusively handled by renaming, e.g. Leningrad > St. Petersburg, or Madras > Chennai), languages do not have multiple names (i.e. synonyms are not acceptable). Of course, sometimes multiple names are in current use, but this cannot be a general solution. When Madras was renamed to Chennai, the old form Madras was used for a while, but it became old-fashioned and is being used less and less. The situation is analogous with changes of language names; thus, when the language Papago was renamed to O’odham, there was never a period during which both names were acceptable to the community of linguists, or to speakers of American English. As soon as the new name was introduced, the old became obsolete. This does not necessarily imply that all members of the community obey the same norm and change their behaviour at the same time, of course, but the nature of the norm is that only one name is possible.
I do not see any way to read this, given the reference to “speakers of American English,” other than as a general, prescriptive statement: “There is One Right Way to refer to any language, and if you do not use it (after an appropriate transition period, during which you can be considered simply ill-informed) you are a Bad Person.” I reject this approach in toto and in the strongest terms; you can use whatever terms you like in your database, but don’t tell me how to use my own language. (Thanks for the very interesting link go to John Cowan.)