What I see as being the problem here is an attempt to carry over AACR traditions into RDA. We have a long-standing tradition of using CCTs for complete works, and for many (but most definitely not all) selective collections of works. That tradition was developed for the predominant technologies of the time (book and card catalogs) and, I would argue, seems to have been designed to serve the needs of scholars more than casual library users. The CCTs served to arrange the catalog so that all of the resources by a given author would be sorted into groups: the complete works together, the complete works in a certain form together, the collections of selected works together, etc.
I do not believe that is really the function that RDA intends the Authorized Access Point to be. For complete works, I think perhaps someone could make an argument that the complete works of a creator make up a theoretical monolithic "work", and so each different published version would then constitute an expression of that theoretical work. The preferred title "Works" is clear and unambiguous, so it can serve reasonably well to identify the nature of the resource. When the users see the title "Works", that title tells them with a reasonable amount of certainty that the resource contains everything by the given creator. One instance of "Works" is more than likely to serve their needs just as well as any other instance of "Works"—and if it doesn't, we provide other information as well to lead them to the desired specific resource. So I could see some merit in an argument to ignore some of the words in RDA and use "Works" as the preferred title for any resource purporting to be the complete works, even if it has its own title that it has been published under.
But selected works are an entirely different matter. You can pretty much guarantee that one instance of "Selections" is not the same as another instance of "Selections", and will not meet the user's needs. What, then, is the point of assigning the CCT "Selections"? In order to have any meaning to the user, it's going to have to get qualified by the title that the collection really is named by. If we need to add that title as a qualifier, then why in the world are we using "Selections" to begin with? Consider the following:
1) The prevailing (mis-)interpretation of RDA would give us:
King, Stephen, 1947- . Short stories. Selections (1978)
King, Stephen, 1947- . Short stories. Selections (1982)
King, Stephen, 1947- . Short stories. Selections (2010)
King, Stephen, 1947- . Short stories. Selections (2015)
King, Stephen, 1947- . Short stories. Selections (2016)
What user is going to make sense of that? So, a slightly better suggestion, still not supported by RDA, would give us:
2) King, Stephen, 1947- . Short stories. Selections (Bazaar of bad dreams)
King, Stephen, 1947- . Short stories. Selections (Descendants)
King, Stephen, 1947- . Short stories. Selections (Different seasons)
King, Stephen, 1947- . Short stories. Selections (Full dark, no stars)
King, Stephen, 1947- . Short stories. Selections (Night shift)
Well, that at least gives the user the information they're really expecting. But it's filled with a whole bunch of other stuff. What RDA actually calls for is:
3) King, Stephen, 1947- . Bazaar of bad dreams
King, Stephen, 1947- . Descendants
King, Stephen, 1947- . Different seasons
King, Stephen, 1947- . Full dark, no stars
King, Stephen, 1947- . Night shift
RDA 6.2 is about identifying the works contained in the resource being described. In 188.8.131.52, we're referred to 184.108.40.206-220.127.116.11 if the collection has a title. The biggest problem here is that RDA says "commonly identified by a title or form of title in resources embodying that compilation or in reference sources" and catalogers are confusing the phrase "commonly identified" with the idea that "If the title is not supported by citations in two or more independent scholarly resources showing this to be the title that the world generally knows the collection by, then we can't assume anybody really knows it by this title—forget the fact that it has never been published under any other title." Yet that statement seems to be how many (most?) catalogers are interpreting RDA. If we can't assume the collection Bazaar of bad dreams (published in 2015) is "commonly identified" by its title, how can we assume that the novel Finders Keepers (also published in 2015) is "commonly identified" by its title? The guideline for choosing the preferred title of a single work is the same as for a collection: we choose "the title or form of title in the original language by which the work is commonly identified either through use in resources embodying the work or in reference sources." If 18.104.22.168 and 22.214.171.124 are not saying the same thing, please someone explain to me what the difference is!
In 126.96.36.199.1 and 188.8.131.52.2, we're told to use CCTs for collections of complete works or complete works in a single form, when the compilation does not have a collective title. For collections of some but not all, we're told to use use the titles of each individual work included; or, as an alternative, a CCT with "Selections". This alternative avoids having to give five, ten, or 300 individual preferred titles. When assigning a CCT to a resource that already has its own collective title, we are doing nothing to help the user identify the resource. All we are doing is assigning a title that will fit the AAP into a single kind of bibliographical listing. Such access points may certainly have a very good function; as VAPs I think they would be very valuable. But as AAPs, I fail to see their value. They identify the resource in a way that probably looks strange to the user; they obscure more than they enlighten.
Kevin M. Randall
Principal Serials Cataloger
Northwestern University Libraries
Proudly wearing the sensible shoes since 1978!