on constraint (of attribute alternation) duplication

Introduction

There are 6 uses of <attList org="choice"> in the Guidelines:

(See below for glossary of column headings.)	element	attributes	works?	+cS?
`<anyElement>`	`@include`, `@except`	NO	no	[1,2]
`<classRef>`	`@include`, `@except`	yes	no
`<dataRef>`	`@key`, `@name`, `@ref`	yes	no
`<moduleRef>`	`@include`, `@except`	yes	yes
`<moduleRef>`	`@key`, `@url`	yes	yes
`<relation>`	`@active`, `@mutual`	yes	yes	[3]

element = The element specification that contains the choice=org. (There are none in other specifications.)
attributes = The attributes that are in alternation with one another.
works? = Whether or not the patterns in tei_all.rnc correctly enforce the alternation — i.e., whether or not the Stylesheets work correctly in this case.
+cS = Is there an added <constraintSpec> that also enforces this disjunction?
note = Note anchors, if any.

Major Problem (not this ticket)

In at least one case of elementSpec//attList[@org eq 'choice'] it does not work. IIRC, it also never works in cases of over-riding an attribute defined in a class. But those are problems for the Stylesheets repo, not this repo.

Secondary Problem (this ticket)

I think we should consistently have an additional <constraintSpec> in all cases or (better yet) not have it in any cases.

Tertiary Finding

(This ticket, but only because there were other problems; this one is a small, corrigible error I would have just fixed in dev without discussion. But since the discussion here might result in “just delete it” …)

The <constraintSpec ident="not-except-and-include"> in <moduleRef> is misplaced. (It is a child of the "prefix" attribute definition; it should be the child of either the "except" or "include" attribute definition.)

Notes

[1] @martindholmes actually has a (not unreasonable) use case for allowing both attributes, anyway. [2] See also #2357. [3] There is an interesting annotation in an XML comment in the <constraintSpec>: “this constraint is pointless in RELAX NG land, where the org=choice works. It is useful in DTD land, where the attList/@org has no effect. It looks to me like it is useful in W3C XML Schema land, too, which I find suprising, as I thought XSD could express a disjuntion like that w/o difficulty. (But I may be reading the XSD incorrectly.) — Syd, 2018-05-01”.

TEIC / TEI