processContents="skip" for other attributes of common mathml elements

[Programmierschmied]

In my organization, when creating schemas that include MathML formulas, we aim to reference variables using elements that contain values. To this end, we intend to utilize our custom namespace for referencing IDs of other elements within the same document. However, I've noticed that the content checks seem to be bypassed in this scenario. This implies that a validator is unable to confirm if such references are legitimate and pertinent to the given document.

For improved accuracy and reliability in validation processes, it would be more preferable to set 'processContents' to 'lax'. This setting would ensure that all elements within the schema are evaluated for their content, thereby preventing any potential misinterpretations or errors caused by omitted or incorrect references

Affects following lines:

<xs:attributeGroup name="CommonAtt">
<xs:attribute name="id" type="xs:ID"/>
<xs:attribute name="xref"/>
<xs:attribute name="class" type="xs:NMTOKENS"/>
<xs:attribute name="style" type="xs:string"/>
<xs:attribute name="href" type="xs:anyURI"/>
<xs:attributeGroup ref="m:CommonDeprecatedAtt"/>
-<xs:anyAttribute namespace="##other" processContents="skip"/>
+<xs:anyAttribute namespace="##other" processContents="lax"/>
</xs:attributeGroup>

[davidcarlisle]

Hmm yes I guess so.

The xsd version is (mostly) generated from the RelaxNG. but there are already some custom cleanups/fixes so adjusting this to use lax here seems reasonable (I think)

I was planning to update the MathML3 XSD over the holiday season for #498 which has been open for an embarassingly long time, I'll probably look to adjust this at the same time.

[davidcarlisle]

Note I guess this is effectively raised against the MathML3 schema but I've added the MathML4 label so I get reminded to adjust the mathMl4 schema generation so it doesn't revert when that is regenerated. (I probably won't touch the MathML3 schema generation pipeline, just hand fix the final schema)

[Programmierschmied]

Thank you, this sounds great!
Also I did take in account, that the attribute xref has no type and could be xs:idref or xs:idrefs. But there I'm unsure if this was done intentionally for any reason.

[davidcarlisle]

It was

xref IDREF #IMPLIED

in the mathml2 dtd (and the relaxNG derived from that)

for mathml3 we switched round and developed with RelaxNG and derived DTD and XSD from that, and, as you say, it's just unconstrained text.

It's all a long time ago and I'd have to search the records to check if that was oversight or by intention. Either way I don't think I want to change the MathML3 schema on that point after all this time, but we should consider this for MathML4.