The Model document describes the override of the natural language
for a single 'text' or 'name' attribute as just inserting an extra
'naturalLanguage' value in front of the 'text' or 'name' attribute
value which qualified the next value only.
The Protocol document uses a more general mechanism that includes a
new "out-of-band" tag called a 'compoundValue' whose values indicates the
number of following values that are to be treated as a single value.
For the natural language override, the compoundValue would always
have the value 2, so that only the next 'text' or 'name' value is
combined with the naturalLanguage value.
We need to pick one method or the other. The first method would
require removing the 'compoundValue' from the Protocol. The second
method would require changing the description of the 'text' attribute
syntax to use the term "natural language override" (which I think is
a better term), instead of "natural language exception", and to remove
the example which shows just the two values.