Re: consensus call sect 4.4

["C. M. Heard" <heard@pobox.com>]

Just to clarify, I am not going to take any position on the
proposed fix (other than as wordsmith), and I will happily
implement whatever consensus emerges.  But I do have some
comments to make.

On Wed, 5 Feb 2003, David T. Perkins wrote:
> The consequence of this has not been fully explored and
> described.  Until this has been done it is inappropriate to go
> forward with the below recommendation.

Ideally, I think we'd pick one of two paths:  we either issue
an update to RFC 2580 to explicitly exempt all items referenced
by compliances and agent-caps (i.e. groups and notifications
as well as objects) and ask that people do not import them,
or we issue an update to RFC 2580 to exempt nothing and require
that everything be in the IMPORTS statement.  It's pretty clear
that either rule would work, and both (I think) would be reasonably
consistent.

I did review the SNMPv2 mailing list archives from the late
1994 and early 1995 time frame when this issue was first
discussed.  Dave Perkins was the proponent of the proposal
to exempt items in compliances and agent-caps from the need
to be imported.  It was clear that his proposal was intended
to remove the requirement to import groups, notifications,
and objects, not just the latter.  My conclusion is is that
the present situation in RFC 2580 (exempting objects but not
mentioning groups and notifications) is the result of
incomplete edits in RFC 1904 (the predecessor of 2580).

One of the things that emerged from this bit of mailing list
archaeology was the point -- not discussed yet in this venue
-- that a module that contains ONLY a compliance or ONLY an
agent-caps would dramatically increase in size if those items
have to be imported.  There was an expectation that such modules
would be common -- compliances from third parties specifiying
requirements, and agent-caps from vendors.  I'm not sure how
things have actually worked out;  I've never seen a 3rd party
compliance statement, but the few agent caps I've seen have
indeed been in stand-alone modules.

Maybe folks want to consider these points in deciding what
to do about Section 4.4.


On Wed, 5 Feb 2003, Juergen Schoenwaelder wrote:
> >>>>> Wijnen, Bert (Bert) writes:
> 
> Bert> I would prefer to change the MAY in the line prefixed with -> in
> Bert> a RECOMMENDED (which is basically the same as a SHOULD, but yet
> Bert> sounds more relaxed). I would like to see if we have (rough)
> Bert> conensus on this. Can all members of this mreview list pls make
> Bert> their opinion known.
> 
> I am fine with RECOMMENDED.

I am somewhat amused by this, because I think RECOMMENDED sounds
stronger. But it's no big thing, because RECOMMENDED and SHOULD
mean exactly the same thing, according to RFC 2119.  Putting on
my wordsmith's hat, here's how I'd do the revisions:

   [ ... ]
   apparent oversight) does not mention those cases.  The exemptions are
-> widely seen as unhelpful because they make IMPORTS rules more
   complicated and inter-module dependencies less obvious than they
   otherwise would be.  It is therefore RECOMMENDED that external
   symbols referenced by compliance statements and capabilities
   statements be listed in the IMPORTS statement.

Note the s/sometimes/widely/ where the arrow is and deletion of
unnecessary text at the end of the last sentence.

//cmh