--On Friday, April 26, 2013 12:47 -0700 SM <sm(_at_)resistor(_dot_)net>
wrote:
...
I think you are right. Of course, individuals pushing drafts
to the ISE could do
the same thing, but that is probably out of scope for us.
The ISE could even point to your document as useful advice for
individuals pushing drafts in the Independent Stream.
This piece of the discussion interacts with another issue.
There has been pushback from time to time about either
Independent Submissions or Informational documents generally
containing normative statements of any kind. Personally, I
think that is an excess of caution that leads to convoluted
workarounds that don't help anyone, but others disagree.
We will probably find more out about this as the experiment runs
but my guess is that implementation statements will have much
less value if they say "we implemented something like what is
specified here" rather than "to the best of our understanding,
we have an implementation that conforms to this specification".
It is hard to the do latter, or to talk about conformance at
all, in the absence of normative and/or conformance statements.
Interoperability reports have much the same problem: if a
specification doesn't contain any normative statements, it
becomes difficult to even guess which of a pair of
implementations that don't interoperate is actually operating as
intended.[1]
So I recommend that we continue to concentrate this effort on
Standards Track protocol specification documents (as the I-D now
does) and start worrying about the various other cases after we
see where the current proposed experiment that leads us.
john
[1] The community has some history of settling that problem by
examining several implementations, figuring out which
interpretation is used by more of them than others, and then
changing the specs to match. That scenario has certainly not
been the tendency in recent years.