I report here a mail conversation between Jenny and Dan, because it might interest other XKOS users.
Jenny: "We are making a Norwegian specification for classifications, based on XKOS, and have a question for you about xkos:next and xkos:previous. Are these intended to be used for semantic ordering (e.g. for time periods, the stone age before the bronze age, the iron age after the bronze age) or administrative ordering (e.g. alphanumeric ordering, B2 before B3, B after A)."
Dan: "Next and Previous were intended to be the immediate successor and predecessor in a sequence. This is in contrast to Succeeds and Precedes, which don’t imply the immediacy. The Succeeds and Precedes relations are transitive, whereas the others are not. In fact, they are anti-transitive, i.e., they never are transitive. If B is Next after A, and C is Next after B, then it is not the case that C is Next after A!
So, Next and Previous can be applied in any sequence where the immediate ones that come before and after are known. I think applying these to time epochs is problematic, because there’s always the possibility that an intermediate period can be identified. As for an alphanumeric ordering, as long as Next and Previous are well-defined given the data, then it should work.
In general, I have a feeling the “semantic ordering” idea is one where Next and Previous should only be used with great care. I can’t prove it, of course, but any time an ordering is based on meanings, it seems something new can be slipped in between any two adjacent items. This is sort of like the rational numbers with the natural ordering imposed. You can always slip another between any 2 you are given: if m, n, p, q are integers with nq != 0 and m/n < p/q, then m/n < (mq+pn)/2nq < p/q."
I report here a mail conversation between Jenny and Dan, because it might interest other XKOS users.
Jenny: "We are making a Norwegian specification for classifications, based on XKOS, and have a question for you about xkos:next and xkos:previous. Are these intended to be used for semantic ordering (e.g. for time periods, the stone age before the bronze age, the iron age after the bronze age) or administrative ordering (e.g. alphanumeric ordering, B2 before B3, B after A)."
Dan: "Next and Previous were intended to be the immediate successor and predecessor in a sequence. This is in contrast to Succeeds and Precedes, which don’t imply the immediacy. The Succeeds and Precedes relations are transitive, whereas the others are not. In fact, they are anti-transitive, i.e., they never are transitive. If B is Next after A, and C is Next after B, then it is not the case that C is Next after A!
So, Next and Previous can be applied in any sequence where the immediate ones that come before and after are known. I think applying these to time epochs is problematic, because there’s always the possibility that an intermediate period can be identified. As for an alphanumeric ordering, as long as Next and Previous are well-defined given the data, then it should work.
In general, I have a feeling the “semantic ordering” idea is one where Next and Previous should only be used with great care. I can’t prove it, of course, but any time an ordering is based on meanings, it seems something new can be slipped in between any two adjacent items. This is sort of like the rational numbers with the natural ordering imposed. You can always slip another between any 2 you are given: if m, n, p, q are integers with nq != 0 and m/n < p/q, then m/n < (mq+pn)/2nq < p/q."