Rough Consensus & Running Code Reconsidered

ABSTRACT

An early slogan of W3C standardisation efforts was 'rough consensus and running code'. In practice, the consensus came first and the code was then written to implement it. XML 1.0 syntax has now been stable for more than three years. Its syntactic adaptability has convinced many very different observers that it is well-suited as the 'glue' between diverse nodes, whether for inter-process communication, for bridging non-congruent data structures, or for inter-network messaging. Some of the most promising applications of XML build bridges of these sorts between end points which could not be altered to adapt to one another. Newer ancillary specifications in the XML family, however, specify and enforce such rigorous data design in XML as effectively to dictate the implementation of the applications which depend on that data. This not only limits the usefulness of XML glue to those applications which can be implemented to a given data design, but in the real world it forces business which want to use XML in this way to rewrite around a data specification those processes which embody their expertise and express their core competence. What if, instead, design procedure using XML were to accept already running production code as the given and then to use the data structure to achieve the rough consensus required to bridge a given pair of processes? What if the next step in the refinement of an XML design were to abstract that data structure to accomodate any process, known or not yet known, which might have use of that data? This paper will examine how that might be done as a matter of rigorous and reproducible methodology, and what the resulting XML structures, and the code capable of handling them, might look like.