Protocol Testing BOF (PROTTEST)

Reported by Peter Desnoyers/Midnight Networks


Purpose

The BOF started off with a description of the possible goals of the
session -- determining guidelines and definitions related to testing,
for testers, implementors, RFC writers, or some combination of the
above.

After a brief discussion of how to structure a testing guide (by
splitting into definitions and methodology, a la BMWG) the discussion
turned to protocol testability, with an emphasis on what could be done
in the future to prevent certain problems from recurring.


Discussion

A statement of the problem which this effort was trying to solve was
proposed by Mike O'Dell and agreed on:



     There is a consensus that [Internet Protocol] specifications
     would benefit from clarity, rationale, precision, and
     completeness.  There exists a body of knowledge, tools, and
     experience that can be brought to bear on this problem.
     To help alleviate this problem it is proposed to produce a
     guidelines document which collects this knowledge and presents
     it in a way useful to RFC writers.



Some mention was made of the possible use of formal methods in the
specification of future Internet Protocols, such as SDL as used by the
ITU. There was little support for this approach at this time, however.

Further discussion concerned the organization and content of the
document.  There was some discussion of whether the document should be
produced as an Informational or standards track RFC; the opinion of the
Area Directors was that a standards track document would be much more
useful.  The approach for the document was described as a ``Strunk and
White'' for RFC authors, giving definitions, guidelines, and a list of
heuristic rules to consult to help avoiding common mistakes in
specification.



Action Items

It was decided to charter a working group to pursue this effort.  Art
Mellor volunteered to chair, and Peter Desnoyers will be the document
editor.  After some discussion the mailing list for the group was named
`stdguide,' as the content had strayed somewhat from the original topic
of protocol testing.  A charter will be written for the group and posted
to this list for discussion.