On 2010-02-06 13:19, Bob Hinden wrote: > Doug, > > On Feb 5, 2010, at 3:59 PM, Doug Barton wrote: > >> On 2/5/2010 2:37 PM, Brian E Carpenter wrote: >>> Oh, OK, that is fine for conformance of course, but leaves things >>> open when you are talking about generating strings. If we want the >>> new recommendation to be a MUST, we may have to consider wording to >>> make it clear how widely it applies. Many existing specs may be >>> affected implicitly. >> Would something like this work? >> >> In the absence of a conflicting specification, ... MUST ... At the time >> of this writing the following specifications are known to conflict: <list> > > That's essentially the definition of SHOULD. It's what you do unless there > is a compelling reason not to do it.
[Digression: a sad fact of life is that this is what MUST means in practice. And SHOULD means "product managers who are short of programmers MAY leave this out".] I was rather thinking that we have to say MUST (as the IESG wishes) and then add something like: The following specifications, among others, are affected by this requirement: RFC 3986, ... I think this will demonstrate the scope of this change from SHOULD to MUST. Brian -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------