I think it is very close to ref and we should reflect that in the design. Both take a label and cross-reference it. ref requires the element to be referenceable because the default form requires a supplement and numbering. But the page form doesn't need that so it can have other requirements.

So, I think perhaps the best solution is not even ref.page, but rather ref(<label>, form: "page"). Similar to cite(.., form: ".."). And ref.form would be "normal" | "page". What do you think?

Okay I see what you mean, and I agree that is a nicer solution.

Just two questions on how ref should then work with forms.

• Do you think that the "page" form should allow the target to be an entry from the bibliography?

• Should the default supplement for the "page" form be none? Or, should it be similar to a figure's supplement where it is based on the kind and the active text language? So maybe the page element function has the argument supplement: auto added, which defaults to the equivalent of "page" in the set language. That way, users get (a perhaps more expected, as in, more inline with the "normal" form) "page X" (in their language) when using the "page" form.

  I think it makes a lot of sense to at least add a supplement argument, with default none, to page which is used as the supplement in the "page" form.

Do you think that the "page" form should allow the target to be an entry from the bibliography?

I'm not sure how useful it is, so I think I would skip it.

Should the default supplement for the "page" form be none? Or, should it be similar to a figure's supplement where it is based on the kind and the active text language? So maybe the page element function has the argument supplement: auto added, which defaults to the equivalent of "page" in the set language. That way, users get (a perhaps more expected, as in, more inline with the "normal" form) "page X" (in their language) when using the "page" form.

Tricky question. But I general agree that I would expect it to behave similarly to normal refs.

I've added page.supplement so page refs have similar behaviour to normal refs. I'm just not sure whether there's any point in implementing the Synthesize trait for Packed<PageElem>. I'm not sure what situation would require its supplement to be resolved from auto before show rules. It makes sense for headings or figures, but pages I'm not so sure.

I'm just not sure whether there's any point in implementing the Synthesize trait for Packed.

No, it doesn't make sense. Also since instances of PageElem won't exist anymore with #4840.

I think I've fixed everything up so that it all works now that #4840 has been merged. Though I might have missed some things that need to be done differently.

Can we use the translations from babel?

Can we use the translations from babel?

We should be able to. I didn't want to go through babel and try to add them to this PR though until after its been reviewed.

I am also not sure whether for other languages the following are the same as in English: page number after the page word, capitalisation of the page word (*), the grammar. Though I suppose this falls under #2485.

(*) In the current implementation, the page word "Page" is always capitalised. I don't know if this is the right default; for figures or equations it makes sense for the word to be capitalised. I guess ideally we would automatically detect if ref is called at the start of the sentence to decide on the capitalisation (but this is certain to cause issues for other languages and be annoying in some situations).

We should be able to. I didn't want to go through babel and try to add them to this PR though until after its been reviewed.

I think we could add translations now.

I'll have a look at doing that.

A couple of things from my attempt at adding translations, using the info from babel.

• For languages with Latin alphabets, I've made the first letter lowercase. The babel translation files appear to be inconsistent with the capitalisation, so I have no idea if this is correct for all these languages.
• he is RTL I think, and might be wrong in the files.
• tl has no translations at all.

Thank you! I was a bit unsure about this but it's true that if we change the reference system in the future in a larger way (which I think will be necessary), things will be highly breaking either way.