Is, for example, "pp. 1567–93" an acceptable way to denote pages 1567 through 1593? In what contexts must you always write it in long form "pp. 1567–1593" instead? Also, what about in the case of smaller numbers, e.g. "pp. 54–7," or more complicated ranges, e.g. "pp. 6, 23–5, 7, 67–8, 94–117"?
The Chicago Manual of Style, 15th Edition (2003), has very clear preferences, which it lists at section 9.64 (rules paraphrased from a table):
For ranges starting with a page number of 1 through 100 (or multiples of 100), use all digits of the end-range number: 3–10, 71–72, 96–117, 100–104, 1100–1113
For ranges starting with a page number of 101 through 109, 201 through 209, and so on, use only the changed part of the end-range number: 101–8, 1103–4
For ranges starting with a page number of110 through 199, 210 through 299, and so on, use two or more digits of the end-range number, as needed: 321–28. 498–532, 1087–89, 11564–615
But if three digits of a four-digit first page number change in the four-digit end-range number, use all four digits of the second number: 1496–1504, 2787–2816
Words into Type, Third Edition (1974), adopts a more flexible approach to inclusive numbering style:
Inclusive page numbers may be given in full ("413–415") or elided ("413–15"); either way, when type is set an en dash, not a hyphen, should be used. There are different styles of elision; whichever is chosen, the indexer should be consistent. The following is a suggested way: Omit from the second number the digit(s) representing hundreds, except when the first number ends in two zeros, in which case the second number should be given in full. If the next-to-last digit in the first number is a zero, only one digit is necessary after the en dash.
8–10, 22–23, 100–102, 107–9, 119–21, 133–34, 1074–76
Interestingly, however, the Words into Type "suggested way" of handling inclusive page numbers matches the Chicago method point for point.
Other styles, of course, differ. Thus, for example, the MLA Style Manual (as represented in the Koinonia Academy "MLA Citation Guide") rejects the Chicago rule that would render "pages 102 through 103" as "102–3":
- Treat inclusive page numbers in text citations and in the list of works cited as follows: 67–68, 102–03, 237–42, 389–421.
Similarly, the North Dakota Supreme Court Citation Manual, which is based on the Harvard Blue Book legal citation style guide insists that the last two digits of a page range should always be retained (though it doesn't offer an example along the lines of "102–03"):
Multiple pages, footnotes, sections, and paragraphs: When citing material that spans more than one page, give the inclusive page numbers, separated by a hyphen or dash. Always retain at least the last two digits, but drop repetitious digits.
*Proper:* 789-90. *Improper:* 789-790. *Proper:* 789-801. *Improper:* 1012-3. *Proper:* 1012-13.
This guide also accepts hyphens or en dashes as punctuation for inclusive page numbers.
I can't comment right now so I make this an answer even though it is only a note.
The Chicago Manual of Style rules are very good but I think that as long as you have less than a million pages it is always easier for a reader to interpret full ranges; shortened ones always require a little more intellectual work, even if you're always consistent and explain the rules to the reader.
They are often more elegant but I think especially on these days when people are used to clear and schematic thinking most readers would prefer the full version.
You just need to ensure that in ranges with several intervals the single intervals are clearly distinguishable, through proper separation.