Lisa Dusseault schrieb:
You and I have had this discussion on the WebDAV mailing list, now we
can repeat it here :)
Such "MAY" constructions convey useful information: It's telling other
agents that they need to be prepared to handle more than one possible
behavior. You could nearly always translate such a sentence to read
"The other party MUST be prepared to handle [set of behaviors]" but
sometimes that doesn't make the spec language quite as readable.
Well, I think we loose readability this way. There's an XML grammar, and
recipients are expected to process messages that conform to this
grammar. That's it. No need to throw in RFC2119 keywords for each of the
aspects, in particular if it isn't done consistently.
In this case it seems unneeded but inoffensive; in other cases it's
actually quite useful, and I don't think that the pattern need generally
be avoided.
It seems to me that opinions about how to use RFC2119 differ. Some claim
that if a spec doesn't use RFC2119 terminology, it's not normative. Some
say it should only be used carefully, and only if it's essential for
interop.
For instance, the W3C just removed lots of them from the 3rd edition of
XML 1.0 (see
<http://www.w3.org/TR/2006/PER-xml-20060614/PER-xml-20060614-review.html>).
It would be nice if the best way to use it could be clarified.
Best regards, Julian
