From: Russ Allbery (rra@stanford.edu)
Date: Fri Jun 01 2001 - 18:10:23 CDT
Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
> If a header fails to achieve what its users expect and intend it to
> achieve, then that is "harm". And this whole RFC 2119 argument is being
> grossly overplayed. Look at RFC 2822, which has been accetped by IETF,
> and identify the interoperability problems or the harm that will arise
> in the case of each of their MUSTs or SHOULDs. You will find that we are
> being excessively timid compared with what they have got away with.
RFC 2822 was retrofitted with RFC 2119 language near the very end of the
working group, and the uses of MUST and SHOULD in that standard were also
controversial on the mailing list. I think that in some cases they were a
mistake, and would rather that we not make the same mistakes.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>