On 12/19/2013 11:51 AM, wrote:
> A mundane question: given a specification with schemas and text
> descriptions of values, if these are in conflict which one is
> authoritative? IOW, would you ignore the text and use the schema
> values as declared, or modify the schema to match the text>?
A specification like that has ceded any claim to authority. People are
free to choose the interpretation that suits them at a given moment.
One resolution is to write into the text "if there are any conflicts,
the schema wins". I don't think it's helpful to write into the text "if
there are any conflicts with the schema, the text wins," since schema
users will just ignore that, and I don't know how to write into the
schema, "if there are any conflicts with the text, the text wins,"
although that might be what is wanted.
Of course this is generally the situation. What this question does for
me is point out a crucial flaw of relying overmuch on specifications,
which are nearly always internally inconsistent. As an implementer, I
would say the ultimate authority lies with the artifact, and with the
creators, maintainers and users of the artifact, since they have access
to what *is*, rather than what ought to be.
-Mike
