the unintuitive semantics of “Key words for use in RFCs to Indicate Requirement Levels” (RFC 2119)

Jonathan Kew wrote:

    I'm not opposed to an informative comment in the WOFF spec
    <http://people.mozilla.com/%7Ejkew/woff/woff-spec-latest.html>,
    noting that UAs MAY provide users with means to examine the metadata

Ah, but each use of the key word “MAY”, as RFC 2119 
<http://tools.ietf.org/html/rfc2119> defines it, actually presents two 
absolute requirements. The phrase “implementations MAY do x” is a 
convenient shorthand for the phrase “implementations MUST interoperate 
with implementations that do x; implementations MUST also interoperate 
with implementations that do not do x”.

Including a normative key word in a non‐normative (informative) comment 
is, at best, useless and, more probably, a source of confusion. I oppose 
such inclusion.

(The canonical version of RFC 2119, “Key words for use in RFCs to 
Indicate Requirement Levels” <http://www.rfc-editor.org/rfc/rfc2119.txt> 
(plain text), has a derivative that is more useful 
<http://tools.ietf.org/html/rfc2119> (HTML). The specification of WOFF 
as Jonathan Kew maintains it 
<http://people.mozilla.com/%7Ejkew/woff/woff-spec-latest.html> is not 
the specification of WOFF which the W3C received as a submission on 8 
April 2010 <http://www.w3.org/Submission/2010/SUBM-WOFF-20100408/>; 
neither of those are a resource through which the W3C might publish a 
standards‐track specification of WOFF <http://www.w3.org/TR/WOFF>; none 
of those are another resource through which the W3C might publish a 
standards‐track specification of WOFF <http://www.w3.org/TR/woff>.)