This release of SDML introduces a new feature: constraints. This is
provided by permitting "false" to be used in the consequents of a
rule. Since it is clearly ridiculous to infer that false is true,
firing such a rule causes SDML to generate a special
self-contradictory assumption and backtrack to investigate other
possibilities. Thus, a rule whose consequents are false acts as a
constraint - it specifies that the antecedents must NOT hold.

This release also incorporates some improvements to the backtracking
algorithm that is used when assumptions cannot be resolved in one
partition (including when a constraint fails). This algorithm
attempts to find resolutions of assumptions in earlier partitions
which enable all the rules in the rulebase to be fired.

Specifically this release fixes some bugs which caused the
backtracking algorithm to go wrong under some rare circumstances, and
it introduces some optimisations to speed up backtracking.

A map colouring example is available to illustrate the constraint and
backtracking features of SDML.