Using the formalism of [Höhfeld and Smolka1988] a constraint-based grammar would formally be considered as a definite clause specification using some specific constraint language. In their simplest form parsing and generation are then viewed as proof procedures that try to find answers for a given goal with respect to a given grammar.
Unless we specify what a definite clause specification intended to represent a natural language grammar, linguistically means, there is obviously no distinction between parsing and generation possible, and computational linguistics would just be constraint-logic programming. In the same sense as conventional programming languages are used just as computational means to specify algorithmic solutions of some problem domain, we adopt the view that CLP is just a formal and operational tool to specify grammatical theories. The theories itself have to give criteria for giving parsing and generation a different meaning which they intuitively seem to convey. It is then a matter of effectivity and efficiency whether they can be realized by the same process or whether two specialized processes have to be developed.
Clearly, those people who are interested in computational aspects of natural language cannot wait until the one and only grammatical theory has been found because grammar theory development certainly undergoes (scientific) evolution. The basic advantage of currently developed theories that follow the constraint-based view, however, is that they present an important homogeneous view with respect to the level of information they try to model that also fits very well with current developments in CLP. By taking into account this commonality when developing parsing and generation strategies it is possible to obtain efficient computability for a broad class of linguistic theories.
Although we do not want to make too many restrictions on the specific form of a constraint-based grammar we want to process, we have to make some assumptions about the representation of phonological and semantic information, and how linguistic objects can be combined to form larger objects.