We present an approach to context-free parsing which builds up a meta-theory from scratch on this subject. >From this meta-theory one can derive any parsing algorithm known to the author (among many others, e.g. Earley's, Tomita's, LL, LR(k) etc.) in a constructive way as well as new algorithms having promising properties. Conclusions from this theory are not only, that all algorithms stem from a single source regardless of the way they have been originally invented, but also open a way to derive specialized algorithms suited for particular demands in practice.