Erik Nordmark <[EMAIL PROTECTED]> wrote: > I think the RFC 2119 upper case terms are more exact and allows > distinguishing between regular english use of these terms, such as "an > implementor should consider the performance ...", and rather exact > statements about the requirement levels in the specification.
That's one of the reasons I don't like capitalizing those words. What if I'm reading a spec in which "must" is usually capitalized, but in one sentence it is lowercase. Am I required to obey that sentence or not? At that point I think we're making distinctions that don't exist. > Thus it shouldn't be a search and replace everything with upper case. If I were persuaded to capitalize those words, I would manually check each instance, and if it shouldn't be capitalized, I'd reword the sentence to avoid that word altogether. So in the end, those words would be capitalized everywhere they appear. Then I'd keep my master copy uncapitalized and unpaginated, and use a script to add the capitalization, pagination, and internet-draft/RFC boilerplate. > Also, there appears to be "should not" even though the "not" terms are > not listed in the 2119 referencing text. Thanks, I'll revise the list of terms to match RFC 2119. AMC
