Content models for Part Div and NonStruct

[davidcarlisle]

The current schema follows ISO 32005 in specifying the content models of structure elements. This leaves Part, Div and NonStruct allowing a fairly large set of child elements.

The PDF 2.0 spec (ISO 32000-2) specifies that the allowed content of each of these is context sensitive and is the allowed content of the parent.

RelaxNG could model this but at some cost in the complication of the schema, essentially adding a new pattern for each context, so for example

Ruby currently allows a NonStruct child (which then allows anything).

this could be changed to allow a NonStruct.Ruby pattern which would be the same pdf2:NonStruct element but with a content model matching that of Ruby.

This would have to be repeated for all uses of all these elements.

I think probably I should do this, as it would give better validation at the cost of making the schema perhaps less readable (and quite a bit longer) but open to suggestions (@u-fischer @zauguin)

[u-fischer]

Yes, it is a bit painful, but I think one have to check the "transparent" rules. Btw: they are not always transparent, in a few cases they do not allow an tag as child directly. e.g. math is not allowed in any of them.

[davidcarlisle]

@u-fischer Hmm Formula allows Div so table 365 in 14.8.4.4 of 32000-2 says

A structure element with the type of Div shall inherit the containment
requirements and limitations of its parent element.

which means that a Div in Formula has the containment rules of Formua, which allows math

This seems to be an error in 32005 which does not list math as allowed in Div ?

[u-fischer]

hm, I'm not completly sure where I got the rules for math from, perhaps that is my own interpretation as the matrix was missing that. But e.g. the more or less official matrix https://pdfa.org/resource/iso-32005/ says e.g. for Part (which has also such a general text in 32000) that it can have Em or Strong as children. And neither Part nor Div can contain direct content items. (I can't find a version which contains the current state after the discussions in the TWGs.)