Use of RFC 2119 Language In Experimental RFCs

["James Kempf" <kempf@docomolabs-usa.com>]

Several years ago, I had a draft on track for experimental RFC returned, and
one of the comments was that it was using RFC 2119 language
(MUST/SHOULD/MAY) when, as an experimental RFC it shouldn't. I've now been
told that recently experimental RFCs are being approved with such language.
I've been encouraging my WGs to not use such language, in an effort to avoid
having drafts returned, but perhaps I'm mistaken.

Could you clarify what the policy is w.r.t. RFC 2119 language in
experimental drafts?

            jak