/Link allows /Formula as a child, but /Reference does not.

[faceless2]

A recent MathML examples from David had this sample

<math>
  <mfrac>
    <mrow id="numerator">
      <mi>x</mi><mo>+</mo>
      <mtext style="display:inline">
        <b>bold </b> <a href="#denominator">link to <math><mi>x</mi></math></a>
      </mtext>
    </mrow>
    <mrow id="denominator">
      <mi>y</mi><mo>+</mo>
      <mtext style="display:inline">
        <i>italic </i> <a href="#numerator">link to <math><mi>y</mi></math></a>
      </mtext>
    </mrow>
  </mfrac>
</math>

When converting to PDF/UA-2 the <a href='#denominator'> is tagged using a /Reference, because in PDF/UA-2 9.2.5.20:

the Reference structure type should be used for intra-document targets

However this leads to invalid content because /Formula is not a valid child of /Reference according to ISO32005.

It is a valid child in /Link, which means - according to the above suggestion - if the href had not been an intra-document target, it would have been valid. /Reference can also contain /Figure.

I'm pretty sure this isn't intentional. Suggestion: Add /Formula as an allowed child of /Reference

[u-fischer]

shouldn't the title say Formula instead of Figure?

[faceless2]

@u-fischer Thank you, fixed.

[mrbhardy]

Proposed Solution: Resolve the Reference child relationships to match Link (Matthew to check if there's anything that would not make sense). Reuse TWG agrees.

32005

[davidcarlisle]

May I ask what is the proposed solution exactly here? (trying to see necessary changes in a validator)

The proposed solution seems to indicate that the allowed children of Link and Reference should be the same (which makes sense) but currently Reference allows Link but not Reference and Link allows Reference but not Link.

Is the proposal that the content models are the same except that Reference allows Link and Link allows Reference, or is the proposal that they share a content model and both allow nesting of both, or that neither is allowed to be nested in either?

Reference
    0..n NonStruct
    0..n Private
    0..n Note
    0..n Lbl
    0..n Em
    0..n Strong
    0..n Span
    0..n Link
    0..n Annot
    0..n FENote
    0..n BibEntry
    0..n Figure
    0..n Artifact
    0..n content item

Link
    ∅* DocumentFragment
    ∅* Part
    0..n Art
    0..n Div
    0..n Sect
    ∅* Aside
    ∅* BlockQuote
    0..n NonStruct
    0..n Private
    ∅* P
    0..n Note
    0..n Code
    ∅* Hn
    ∅* H
    ∅* Title
    0..n Sub
    0..n Lbl
    0..n Em
    0..n Strong
    0..n Span
    0..n Quote
    0..n Reference
    0..n Annot
    ∅* Form
    0..n Ruby
    0..n Warichu
    0..n FENote
    ∅* L
    0..n BibEntry
    ∅* Table
    ∅* Caption
    0..n Figure
    0..n Formula
    0..n Artifact
    0..n content item

[car222222]

Just a note that: David’s question needs an answer; Any “proposed solution” should be applied NOW!

[car222222]

This got assigned to Duff, but I sm unsure why!

[mrbhardy]

We discussed this in the last TWG meeting, because it was a comment on the last draft. We agreed to the change. The proposal was to match the containment for each and allow link and reference.

[car222222]

Did this get changed, at least in the table in Clause 5 of 32005?

[mrbhardy]

Yes, already fixed in the latest 32005 (you can check this yourself @car222222, since it was shared to the reuse group).