WORKING PAPER
  
             REGISTRY - A NEW FEATURE FOR ISO RECOMMENDATIONS
  
In nine years of existence, ISO/TC97 has been presented with substantial
evidence of the complexity of information processing standardization.
Here, and also at the national level, it is seen that there are two classes
of results:
  
1.  Complete agreement upon a standard.
  
    Here the standard will be incomplete.  It will allow ambiguous
    interpretation, and it will be insufficient for utilization as a
    system building block.
  
2.  Agreement upon a standard is not attainable.
  
    Here the ponderous mechanics force the work to go on for a number of
    years.  Meanwhile, the standard is relatively unknown to those that
    have need for it in their design; more deviations and alternatives
    are introduced into practice.  These additional complications, with
    the resulting investments to protect, make it increasingly difficult
    to reach agreement on a single standard.
  
Between a single standard and an uncontrolled multiplicity of practices there
can be the compromise of a small and controlled number of practices.  These
are to be considered not as opposite in nature, but rather as practices that
are identical except for controlled deviations, mechanically and logically
convertible.
  
There is a feasible method to accomplish this within the framework of
ISO/TC97.  It is exemplified by the ESCape mechanism utilized by TC97/SC2
to achieve code extension and expansion.  It is time for TC97 to examine
this principle as it applies to other areas of its work.
  
The ESCape mechanism. as set up originally in ISO R646, was to provide
for additional controls and graphics.  However, the device was conceived
to be more general than this.  ESCape is being utilized now to create
system compatibility with other graphic sets, i.e., those used in typo-
graphic work.
  
The concept is of even broader utility.  ESCape mechanism may be used
to identify, for example, such items as:
  
o  Different data and file structures.
 
o  Different notational conventions (i.e., variants in the hardware
   representation of ALGOL basic symbols).
  
The key to successful usage of ESCape is the concept of "Registry", which
would be a category of lesser force than an ISO Recommendation, but
nevertheless must be controlled by the ISO.  This control must be exercised
under a rigorous set of procedures, however, to protect the integrity of
the registry mechanism.
  
Because disagreement on reaching a Draft Recommendation is often localized
to a small portion of the technical content, registry has the following
advantages:
  
1.  Ratification of the core of agreement (possibly with a single preference
    indicated for the required remainder) can occur earlier.  This allows
    earlier official dissemination, which prevents growth of unnecessary
    contrary usage.
  
2.  Similar usage that is not anticipated or met by an existing Recommenda-
    tion can be adjoined to the Recommendation by registry in a shorter time
    than by revision, which nevertheless can proceed under normal procedures.
  
3.  The flexibility of accommodation to practices in a state of development
    more advanced than the ISO Recommendation lessens the probability of
    rejection of the totality of a Recommendation.  Variety is thereby
    reduced and eventual convergence is facilitated.
  
4.  Self-identification of conventions, media and structures that are not in
    accord with ISO Recommendations (existing practices in particular) enables
    coexistence with those ISO Recommendations with minimum difficulty.
    Possibilities of system interchangeability are enhanced.
  
5.  This same self-identification of variations permits and encourages gradual
    and gentle conversion to the specifications of the ISO Recommendation.
  
It is not attempted to propose registry procedures in this paper.  TC97/SC2
and the ECMA have been studying this matter carefully and can submit proposals.
These can then be expanded to cover the more general case of all TC97 work.
However, some overall requirements can be stated:
 
1.  There must be an official body with authority to control the registry
    and issue the comprehensive list periodically.
  
2.  This body must have the authority to make technical decisions to
    accept or reject practices proposed to have a registry number.
  
3.  There should be an appeal upon denial.
  
4.  There must be a published procedure available to bodies that wish to
    propose candidates.
    
5.  This procedure must be available only to such bodies that have overall
    authority in their domain.
 
6.  New practices proposed must receive the same full agreement as an ISO
    Recommendation (although this can occur in a shorter time).  Even so,
    acceptance cannot be indiscriminate;  only one practice in a specific
    area or for a specific function is acceptable (i.e., only one ESCape
    sequence shall be registered to indicate shift to half duplex).
  
7.  Existing practices may be registered at the technical option of the
    registering body, subject to demonstration of volume of usage and need
    (i.e., the EBCDIC code should be registered, but that of the IBM 360
    should not).
  
8.  For any particular group of practices, the ESCape mechanism shall be
    unambiguously and singly defined as to location, effect, and character-
    istics (i.e., the mechanism to define the disk format, encoding and
    density shall be found at only one position and must conform to ISO
    Recommendations).
  
9.  Practices may be registered only if they have isomorphic or homomorphic
    relationships to the corresponding ISO Recommendation, together with a
    method of returning to a state of conformity to that Recommendation.
  
In conclusion, registry is an almost mandatory mechanism in any field
with a rate of technical development and change that exceeds the rate of
development and approval of Recommendations.  It is a key to the formation
of large and interconnected systems.
 
                                                  R. W. Bemer