As far as I understood it, this usage of deprecated would be semantically wrong. Deprecated to me means that this claim is not, was never and never will be true. Normal however means that his claim has been true at some place in time or, more generally, is true in some non-default context. I suppose these redirects are somewhat true, they are just not the main (preferred) truth.

Is there any page which clearly explains how the ranks should be used in general? If not, I think the usage should be clarified before anyone works on this task. Personallay, I use preferred for the statements which are the most important and relvant in this context (for example the first publication) and normal for other ones. The deprecated rank is used relaly rarely by me, for example when I want to give a increment to normal statements.

https://www.wikidata.org/wiki/Help:Ranking

https://www.wikidata.org/wiki/Help:Ranking doesn't mention how they're supposed to apply to external identifiers though, so it's still not very clear. You could argue that a redirecting identifier "represents outdated knowledge" (because the object it originally pointed to is no longer considered a separate object) or that it doesn't (because it still points to a valid object, just not the one it originally pointed to). :/

@MGChecker: I don't think using preferred for the most important and relevant is sufficient to distinguish redirects. For example, a site might have one identifier which covers almost all of an author's works, another identifier which only has a single work where someone typoed the author's name, and a third identifier which is a redirect to the first identifier. It would make sense for the first identifier to be preferred, as that's clearly the main object with almost all of the information, making it the most important and relevant, but that doesn't provide a distinction between the second identifier (which points to a different object with different information) and the third (which is just an alias for the first object).

Either way, even if people decide to handle redirecting identifiers in a different way, I still think deprecated statements should be less prominently displayed.

@adrianheine: By the way, a number of people are already using (and telling other people to use) the deprecated rank for redirected identifiers, see for example https://www.wikidata.org/wiki/Wikidata:Project_chat#One_item.2C_two_IMDb_identifier_.28P345.29 and https://www.wikidata.org/wiki/Wikidata:Project_chat#How_to_make_.22unique_value.22_constraint_violations_report_ignore_deprecated_values_.3F - If you think it's semantically wrong, maybe you'd like to start an on-wiki discussion about it. Personally I don't think people will stop using it that way though, unless they're given another way (e.g. another rank) to mark statements as being true but no longer current (it's not just redirected identifiers either, I've also seen people mark historical population values and old software versions as deprecated).

@Nikki: using it for old (but correct) population values is definitely wrong for the semantics we give to ranks. For identifiers I would also lean towards marking the one that is in use as preferred and the others as normal. That is closest to the intended use of ranks.

I started a new section on the project chat page: https://www.wikidata.org/wiki/Wikidata:Project_chat#The_meaning_of_the_deprecated_rank

I would tend to agree that the current best-known ID of an external identifying page should be set as preferred.

Maybe this points to the need for a new semantic rank of "outdated". This would help fix overloading of "normal" (which presently means both "old, but valid" and "another view, possibly in the minority, but certainly not 'wrong' [which is 'deprecated']").

Outdated views of the data could also be qualified to show that the notion that some data was true (or thought to be true?) at some point in time, but I'm not sure that fits better as a qualification than the ranking methodology. Point in time is useful for some of these cases but not all of them, I think.