--On tirsdag, september 25, 2001 11:29:28 +0900 Jiwoong Lee <porce(_at_)ktf(_dot_)com>
wrote:
James and Ken,
Quite interesting remarks you have made. I agree to Ken's syllogistics
and I love James' consideration on corporeal world practice.
Let's begin at the beginning. What is the requirement level for
exceptions of "MUST NOTs" ?
First, in RFC terms, MAY has the equal value to MAY NOT. In fact, there
is no definition on "MAY NOT" in RFC 2119.
Exceptions to MUST NOTs means there are cases which are not prohibited.
The semantics gives me the implication that the appripriate requirement
level is placed between "SHOULD NOT" and "MAY".
MUST NOT means that you never do it, and any implementation that does it is
nonconformant to the spec.
If you find that you have to do it in some case, you have found a bug in
the spec.
Note that English is a slippery language, and constructs like "MUST NOT do
X unless Y, Z and W occurs" and its semantic equivalents are quite common.
Most specs have bugs.
Harald