Ryan, good to see you working on it..
I of course agree with Yves and Fredrik on the interpretation. The main reason being that it really is NOT open to interpretation..
The Constraint does not say anything on when the attribute is OPTIONAL.
The attribute is OPTIONAL as per the schema and the "Contains:" normative and exhaustive enumeration.
The Constraint is a continuation of that definition and it specifies cases when the attribute becomes REQUIRED in contrast with its initially being described as OPTIONAL.
It is set apart because it cannot be expressed in a 1.0 xsd.
Now about the "verbiage".. There is an important logical distinction between "IF" and "IFF" (shorthand for "if and only if").
Whenever IFF is used it means a biconditional, conditional valid in both directions, a definitory equivalence. The only valid reason ever for that attribute becoming REQUIRED is that the Document has at least one <target> element. In old philosophical speech this means that having a <target> element is both necessary and sufficient for trgLang being REQUIRED on the Document.
It is also true that when the Document does NOT have a <target> the trgLang attribute is NOT REQUIRED.
But NOT REQUIRED is equivalent with OPTIONAL not with MUST NOT be set or similar (forbidden).. I guess this was where you got confused and thought that it might mean forbidden. It does not and cannot to be sure :-)
In the modern logical sense, you can say that the attribute being REQUIRED is coextensive with any Document having a <target>.
The closest to plain English is this. IFF statement means that the IF statement is true in both directions.
IFF = <->
IF = ->
(p<->q) -> ((p->q) & (q->p))
((p->q) & (q->p)) -> (p<->q)
(p<->q) <-> ((p->q) & (q->p))
The IFF statement is stronger because it implies both IF statements, while any of the IF statements alone does not say anything about the other way round.
I made every effort to have the IFFs used properly throughout the spec. When you are testing them, you can always decompose them into two IF statements with the antecendent (p) and consequent (q) switched as shown above..
I hope this helps in reading the spec in general