Keith Moore wrote:
well, maybe the thing to do is to _discuss_ amendments rather than
a revised document, with the understanding that the amendments won't
be incorporated into the document until (just before) last call.
note that the amendments can still be useful in their own right, for
people who are very familiar with the current spec - those people want
to know the diffs from the previous version.
Some memos incorporate such a list of changes as an appendix, and
that is indeed useful to those familiar with the previous document.
Perhaps the thing to do would be to first compile a list of potential
changes, then revise that list based on discussion. Changes could
be folded in (with items from the list placed in an appendix) as
consensus is reached, which makes reviewing the draft for internal
consistency easier than dealing with document + changes; alternatively
that could be postponed until the end.
The thing to avoid would be the situation which 2821/2822 sought to
address, namely that 821/822 were modified by so many other documents
that it was nearly impossible for an implementor to build according