Overview of tersmu architecture
===============================

Input: Lojban text, an instance of the 'text' production.

Step 1: Input is parsed into an abstract syntactic form.
    The core of this is the pappy-generated module Tersmu
    but some preprocessing is first done by
	(pappy-generated) Morphology and Morph,
	    which split the input into lojban words,
	and by ParseText.hs,
	    which implements some imperfect ad hoc handling of indicators and
	    frees.
    The result of this stage is an instance of JboSyntax.Text

Step 2: The abstract syntax is converted to logical form.
    This is the job of JboParse. The algorithm is necessarily baroque.
    A little monadic voodoo is required, using a stack of state and
	continuation monad transformers. The definitions and associated code are
	in ParseM; JboParse aims to be relatively clean and readable without
	extensive knowledge of Haskell.
    The result is a list of (fragments and) propositions in the generalised
	predicate logic defined abstractly in Logic and specialised to lojban
	in JboProp.

Output: For now, all we know how to do with a JboProp is print it.
    JboShow has the print routines.
    The output syntax should be considered provisional, subject to change.