[Ietf-not43] MUST/SHOULD protocol/service schism

Andrew Newton anewton@ecotroph.net
Wed, 22 Jan 2003 14:02:56 -0500


In rewriting the requirements draft so that MUST/SHOULD language is not 
used for service descriptions, a problem has been made more evident.

First, the current draft actually rarely uses the word "protocol", 
instead using "service" in places where we would naturally want to say 
protocol.  So if we went through the draft and strictly changed the 
capital case in those places, the draft actaully won't specify too much 
of anything.

So I went back and changed the word "service" to "protocol" in places 
where the requirement was about a protocol issue.  However, now the 
draft has a schizophrenic feel.  Here are a few examples:

    The protocol MUST provide the capability of searching for
    registrants, within limits set by the policy of an operator, by exact
    name match or a reasonable name subset.  This search must comply with
    Section 3.2.8.

    The service must provide a mechanism to distribute this search across
    all applicable domain registries and registrars.  The protocol SHOULD
    have a means to narrow the scope of a search to a specific TLD.  The
    protocol MUST be able to specify an empty result set should the
    search yield no results.

and

    The protocol MAY allow the ability, within the limits of the
    operator, to list all domains hosted by a specific nameserver given
    the fully-qualified host name or IP address, if applicable, of the
    nameserver.  The service may provide a mechanism to distribute this
    search across all applicable domain registries and registrars.


I do not think this makes it clear that the service requirements are 
OPTIONAL depending on the policy of an operator.  It is confusing.

I'd like to restructure the document so that service model descriptions 
are separated from the client/server protocol interactions.  I think by 
placing them in separate sections, it would be clearer.  The MUST/SHOULD 
language on the service descriptions would be lowercased or removed 
altogether.

Please let me know what you think.  I'd rather get out a good document 
than one that can't be understood by anybody.

-andy