(From Lognet 91/4)
Bill Gober has commented that certain combinations of little words parse, whereas when written separately, they do not parse. This is because the preparser (its lexer, in this instance) decides that a combination, e.g. lemi , belongs in the LE lexeme, whereas if written le mi, it would be lexed as a sequence of two lexemes, and one that the grammar cannot handle. Doing something about this would require doing something about combinations that are intended to parse differently. E.g., lepo is in the lexeme LEPO, but le po is an instance of LE PO, a quite legitimate sequence of two lexemes. In the grammar, the LEPO operator takes a sentence as an operand, whereas PO expects a PREDA. Thus Le po preda parses as Le (po preda) in which po preda is an "inflected" predicate which is distinct in meaning from its base term preda. This is an important distinction in the grammar.
Also Bill was concerned because NB3 said that GA was a member of the PA lexeme, where now, that is, in the 4th Edition language, it is a separate lexeme to allow for its added usage in deferring first arguments. Hence Gahu? does not parse, whereas Nahu? does.
Thirdly, it has not been made clear that JI can take bare predicates as operands (and these may have sutori arguments) whereas older, pre-4th Edition Loglan required JIO. JI + predicate phrases, like JIO phrases, are closed with the GUI lexeme. JI + argument phrases require no closure (though a PREDA-final argument may require a pause-comma or a gu). This last problem arose because an oversight in the grammar did not allow ne preda pe (JI) le prede, ji le predi (though le preda pe le prede, ji le predi parsed successfully). This has been corrected.
There seems also to have been a misunderstanding about the role of ‘.’ in Loglan. It has no role in the grammar, as a full stop is indicated by either a long pause, or a new sentence beginning with an ‘I’ word or Nao. When LIP was asked to parse continuous text, a period was added to serve as a textual signal to the parser when to stop and parse. Hence fragments with one or more periods between quotes or parentheses would not be expected to parse. Thus the Loglandic equivalent of a full stop is permitted to occur within quotations and parenthetic expressions in text provided they are typed without periods. —RAM