Question about test/csl/number_PlainHyphenOrEnDashAlwaysPlural.txt

[jgm]

This test has an item with "page": "Michaelson-Morley".
The expected test results show that the processor is expected to recognize that this is not really a page range, and hence (a) not use the plural form "pages" and (b) not turn the - into an en-dash.

Since this is all undocumented in the spec, I want to ask what the algorithm is. How do we decide whether something is really a page range? And, is this really needed? Who would put Michaelson-Morley as a page range, and if they did, might they not want it to be interpreted as one?